Cisco Collaboration System 10.x Solution Reference Network Designs (SRND)

Cisco Unified Communications Directory Architecture

Figure 16-3 shows the basic architecture of a Unified CM cluster. The embedded database stores all configuration information, including device-related data, call routing, feature provisioning, and user profiles. The database is present on all servers within a Unified CM cluster and is replicated automatically from the publisher server to all subscriber servers.

Figure 16-3 Cisco Unified CM Architecture

By default, all users are provisioned manually in the publisher database through the Unified CM Administration web interface. Cisco Unified CM has two types of users:

• End users — All users associated with a physical person and an interactive login. This category includes all Unified Communications users as well as Unified CM administrators when using the User Groups and Roles configuration (equivalent to the Cisco Multilevel Administration feature in prior Unified CM versions).
• Application users — All users associated with other Cisco Unified Communications features or applications, such as Cisco Attendant Console, Cisco Unified Contact Center Express, or Cisco Unified Communications Manager Assistant. These applications need to authenticate with Unified CM, but these internal "users" do not have an interactive login and serve purely for internal communications between applications.

Table 16-3 lists the application users created by default in the Unified CM database, together with the feature or application that uses them. Additional application users can be created manually when integrating other Cisco Unified Communications applications (for example, the ac application user for Cisco Attendant Console, the jtapi application user for Cisco Unified Contact Center Express, and so forth).

Based on these considerations, Figure 16-4 illustrates the default behavior in Unified CM for user-related operations such as lookups, provisioning, and authentication.

Figure 16-4 Default Behavior for User-Related Operations for Unified CM

End users access the Unified CM User Options page via HTTPS and authenticate with a user name and password. If they have been configured as administrators by means of User Groups and Roles, they can also access the Unified CM Administration pages with the same credentials.

Similarly, other Cisco features and applications authenticate to Unified CM via HTTPS with the user name and password associated with their respective application users.

The authentication challenge carried by the HTTPS messages are relayed by the web service on Unified CM to an internal library called Identity Management System (IMS). In its default configuration, the IMS library authenticates both end users and application users against the embedded database. In this way, both "physical" users of the Unified Communications system and internal application accounts are authenticated using the credentials configured in Unified CM.

End users may also authenticate with their user name and a numeric password (or PIN) when logging into the Extension Mobility service from an IP phone. In this case, the authentication challenge is carried via HTTP to Unified CM but is still relayed by the web service to the IMS library, which authenticates the credentials against the embedded database.

In addition, user lookups performed by Unified Communications endpoints via the Directories button communicate with the web service on Unified CM via HTTP and access data on the embedded database.

The importance of the distinction between End Users and Application Users becomes apparent when integration with a corporate directory is required. As mentioned in the previous section, this integration is accomplished by means of the following two separate processes:

• LDAP synchronization

This process uses an internal tool called Cisco Directory Synchronization (DirSync) on Unified CM to synchronize a number of user attributes (either manually or periodically) from a corporate LDAP directory. When this feature is enabled, users are automatically provisioned from the corporate directory in addition to local user provisioning through the Unified CM administration GUI. This feature applies only to End Users, while Application Users are kept separate and are still provisioned via the Unified CM Administration interface. In summary, End Users are defined in the corporate directory and synchronized into the Unified CM database, while Application Users are stored only in the Unified CM database and do not need to be defined in the corporate directory.

• LDAP authentication

This process enables the IMS library to authenticate user credentials of LDAP synchronized End Users against a corporate LDAP directory using the LDAP standard Simple_Bind operation. When this feature is enabled, End User passwords of LDAP synchronized End Users are authenticated against the corporate directory, while Application User passwords and passwords of local End Users are still authenticated locally against the Unified CM database. Cisco Extension Mobility PINs are also still authenticated locally.

Maintaining and authenticating the Application Users internally to the Unified CM database provides resilience for all the applications and features that use these accounts to communicate with Unified CM, independently of the availability of the corporate LDAP directory.

Cisco Extension Mobility PINs are also kept within the Unified CM database because they are an integral part of a real-time application, which should not have dependencies on the responsiveness of the corporate directory.

The next two sections describe in more detail LDAP synchronization and LDAP authentication, and they provide design best-practices for both functions.

Note As illustrated in the section on Directory Access for Unified Communications Endpoints, user lookups from endpoints can also be performed against a corporate directory by configuring the Cisco Unified IP Phone Services SDK on an external web server.

LDAP Synchronization

Synchronization of Unified CM with a corporate LDAP directory allows the administrator to provision users easily by mapping Unified CM data fields to directory attributes. Critical user data maintained in the LDAP store is copied into the appropriate corresponding fields in the Unified CM database on a scheduled or on-demand basis. The corporate LDAP directory retains its status as the central repository. Unified CM has an integrated database for storing user data and a web interface within Unified CM Administration for creating and managing user accounts and data. When LDAP synchronization is enabled, the local database is still used, and additional local end-user accounts can be created. Management of end-user accounts is then accomplished through the interface of the LDAP directory and the Unified CM administration GUI. (See Figure 16-5.). Accounts for application users can be created and managed only through the Unified CM Administration web interface.

The user account information is imported from the LDAP directory into the database located on the Unified CM publisher server. Information that is imported from the LDAP directory may not be changed by Unified CM. Additional user information specific to Cisco Unified Communications is managed by Unified CM and stored only within its local database. For example, device-to-user associations, speed dials, call forward settings, and user PINs are all examples of data that is managed by Unified CM and does not exist in the corporate LDAP directory. The user data is then propagated from the Unified CM publisher server to the subscriber servers through the built-in database synchronization mechanism.

User information synchronized from the LDAP directory can be converted to local user information so that the user information then can be edited locally on Unified CM. Local end users can be added manually using the Unified CM administration GUI. During an LDAP sync, a local end user is converted to an active LDAP user, and if a user with the same user ID is found in LDAP, the locally configured data is replaced with data from the directory.

Figure 16-5 Enabling Synchronization of User Data

When LDAP synchronization is activated, only one type of LDAP directory may be chosen globally for the cluster at any one time. Also, one attribute of the LDAP directory user is chosen to map into the Unified CM User ID field. Unified CM uses standard LDAPv3 for accessing the data.

Cisco Unified CM imports data from standard attributes. Extending the directory schema is not required. Table 16-4 lists the attributes that are available for mapping to Unified CM fields. The data of the directory attribute that is mapped to the Unified CM User ID must be unique within all entries for that cluster. The attribute mapped to the Cisco UserID field must be populated in the directory and the sn attribute must be populated with data, otherwise those records are skipped during this import action. If the primary attribute used during import of end-user accounts matches any application user in the Unified CM database, that user is not imported from the LDAP directory.

Table 16-4 lists the attributes that are imported from the LDAP directory into corresponding Unified CM user fields, and it describes the mapping between those fields. Some Unified CM user fields might be mapped from one of several LDAP attributes.

Table 16-5 contains a list of additional attributes that are imported by the DirSync process and copied into the Unified CM. The attribute msRTCSIP-PrimaryUserAddress is populated in AD when Microsoft Lync is used. This table is included for completeness.

In addition to the direct mapping of directory attributes to local user attributes, other characteristics of the synchronized users are determined by settings on the LDAP directory synchronization agreement. Access control group membership of users created through LDAP synchronization is directly configured in the LDAP directory configuration setting. Further user capabilities are determined by the feature group template selected. The selection of a feature group template on an LDAP directory synchronization agreement is optional. The feature group templates allow administrators to define user characteristics, including home cluster selection, IM and Presence capabilities, mobility features, services profiles, and user profiles. The user profiles allow administrators to define a universal line template which is considered for automatic creation of directory numbers for LDAP synchronized users by Unified CM.

The synchronization is performed by a process called Cisco DirSync, which is enabled through the Serviceability web page. When enabled, it allows one to 20 synchronization agreements to be configured in the system. This number is reduced to 10 if more than 80,000 users are synchronized. An agreement specifies a search base that is a position in the LDAP tree where Unified CM will begin its search for user accounts to import. Unified CM can import only users that exist in the domain specified by the search base for a particular synchronization agreement.

In Figure 16-6, two synchronization agreements are represented. One synchronization agreement specifies User Search Base 1 and imports users jsmith, jdoe and jbloggs. The other synchronization agreement specifies User Search Base 2 and imports users jjones, bfoo, and tbrown. The CCMDirMgr account is not imported because it does not reside below the point specified by a user search base. When users are organized in a structure in the LDAP directory, you can use that structure to control which user groups are imported. In this example, a single synchronization agreement could have been used to specify the root of the domain, but that search base would also have imported the Service Accts. The search base does not have to specify the domain root; it may specify any point in the tree.

Figure 16-6 User Search Bases

To import the data into the Unified CM database, the system performs a bind to the LDAP directory using the account specified in the configuration as the LDAP Manager Distinguished Name, and reading of the database is done with this account. The account must be available in the LDAP directory for Unified CM to log in, and Cisco recommends that you create a specific account with permissions to allow it to read all user objects within the sub-tree that was specified by the user search base. The sync agreement specifies the full Distinguished Name of that account so that the account may reside anywhere within that domain. In the example in Figure 16-6, CCMDirMgr is the account used for the synchronization.

It is possible to control the import of accounts through use of permissions of the LDAP Manager Distinguished Name account. In this example, if that account is restricted to have read access to ou=Eng but not to ou=Mktg, then only the accounts located under Eng will be imported.

Synchronization agreements have the ability to specify multiple directory servers to provide redundancy. You can specify an ordered list of up to three directory servers in the configuration that will be used when attempting to synchronize. The servers are tried in order until the list is exhausted. If none of the directory servers responds, then the synchronization fails, but it will be attempted again according to the configured synchronization schedule.

Synchronization Mechanism

The synchronization agreement specifies a time for synchronizing to begin and a period for re-synchronizing that can be specified in hours, days, weeks, or months (with a minimum value of 6 hours). A synchronization agreement can also be set up to run only once at a specific time.

When synchronization is enabled for the first time on a Unified CM publisher server, user accounts that exist in the corporate directory are imported into the Unified CM database. Then either existing Unified CM end-user accounts are activated and data is updated, or a new end-user account is created according to the following process:

1. If end-user accounts already exist in the Unified CM database and a synchronization agreement is configured, all pre-existing accounts that have been synchronized from LDAP previously are marked inactive in Unified CM. The configuration of the synchronization agreement specifies a mapping of an LDAP database attribute to the Unified CM UserID. During the synchronization, accounts from the LDAP database that match an existing Unified CM account cause that Unified CM account to be marked active again. If accounts from LDAP match an existing Unified CM account that is not marked as an LDAP synchronized account, then these accounts are ignored.

2. After the synchronization is completed, any LDAP synchronized accounts that were not set to active are permanently deleted from Unified CM when the garbage collection process runs. Garbage collection is a process that runs automatically at the fixed time of 3:15 AM, and it is not configurable.

3. Subsequently when changes are made in the corporate directory, the synchronization from Microsoft Active Directory occurs as a full re-synchronization at the next scheduled synchronization period. On the other hand, the Sun ONE directory products perform an incremental synchronization triggered by a change in the directory. The following sections present examples of each of these two scenarios.

Note Once users are synchronized from LDAP into the Unified CM database, deletion of a synchronization configuration will cause users that were imported by that configuration to be marked inactive in the database. Garbage collection will subsequently remove those users.

Account Synchronization with Active Directory

Figure 16-7 shows an example timeline of events for a Unified CM deployment where LDAP Synchronization and LDAP Authentication have both been enabled. The re-synchronization is set for 11:00 PM daily.

Figure 16-7 Change Propagation with Active Directory

After the initial synchronization, the creation, deletion, or disablement of an account will propagate to Unified CM according to the timeline shown in Figure 16-7 and as described in the following steps:

1. At 8:00 AM on January 1, an account is disabled or deleted in AD. From this time and during the whole period A, password authentication (for example, Unified CM User Options page) will fail for this user because Unified CM redirects authentication to AD. However, PIN authentication (for example, Extension Mobility login) will still succeed because the PIN is stored in the Unified CM database.

2. The periodic re-synchronization is scheduled for 11:00 PM on January 1. During that process, Unified CM will verify all accounts. Any accounts that have been disabled or deleted from AD will at that time be tagged in the Unified CM database as inactive. After 11:00 PM on January 1, when the account is marked inactive, both the PIN and password authentication by Unified CM will fail.

3. Garbage collection of accounts occurs daily at the fixed time of 3:15 AM. This process permanently deletes user information from the Unified CM database for any record that has been marked inactive for over 24 hours. In this example, the garbage collection that runs at 3:15 AM on January 2 does not delete the account because it has not been inactive for 24 hours yet, so the account is deleted at 3:15 AM on January 3. At that point, the user data is permanently deleted from Unified CM.

If an account has been created in AD at the beginning of period A, it will be imported to Unified CM at the periodic re-synchronization that occurs at the beginning of period B and will immediately be active on Unified CM.

Account Synchronization with Sun ONE

Sun ONE products support incremental synchronization agreements and use a different synchronization timeline than Microsoft Active Directory. The synchronization makes use of the Persistent Search mechanism defined by an IETF Draft and supported by many LDAP implementations. Figure 16-8 shows an example of this synchronization timeline for a Unified CM deployment with LDAP Synchronization and LDAP Authentication both enabled.

Figure 16-8 Change Propagation with Sun ONE

The example in Figure 16-8 involves the following steps:

1. An account is deleted from the corporate directory at 8:00 AM on January 1, which causes an incremental update to be sent from the LDAP server to Unified CM. Unified CM sets its corresponding copy of the data to inactive. Because LDAP authentication is configured, the user will be unable to log in via password as soon as the LDAP server has deleted the record. Also, the PIN may not be used for login at the moment the Unified CM record is marked inactive.

2. During period B, the user’s record is still present in Unified CM, albeit inactive.

3. When the garbage collection runs at 3:15 AM on January 2, the record has not yet been inactive for 24 hours. The data remains in the Unified CM database until the beginning of period C on January 3, when the garbage collection process runs again at 3:15 AM and determines that the record has been inactive for 24 hours or more. The record is then permanently deleted from the database.

Accounts that are newly created in the directory are synchronized to Unified CM via incremental updates as well, and they may be used as soon as the incremental update is received.

Automatic Line Creation

For users created during LDAP synchronization, Unified CM can automatically create directory numbers. These auto-generated directory numbers are either based on information found in the directory and defined based on a mask to be applied to the phone number found in the directory, or the numbers are taken from directory number pools defined on the LDAP synchronization agreement. If a mask is defined on the synchronization agreement, then to allow for variable length +E.164 directory numbers to be generated, the following rules apply:

• If the mask is left empty, then Unified CM takes all digits and also a leading "+" (if present) from the directory.
• X is used as a wildcard character in the mask.
• A wildcard matches on digits and "+".
• Wildcards in the mask are filled from the right.
• Unfilled wildcards in the mask are removed.

Table 16-6 shows some examples.

As an alternative to creating directory numbers based on information from LDAP, directory numbers for new users can also be taken from predefined number pools. Each pool is defined by a start and end number. Directory number pools support +E.164 numbers. Up to five pools can be defined. Numbers are assigned from the first pool until all numbers of that pools have been assigned. Number assignment then starts to take numbers from the next pool.

Automatic Line Creation is enabled only if both of the following conditions are met:

• A Feature Group Template is assigned in the directory synchronization agreement, and
• A Universal Line Template is selected in the User Profile selected in the Feature Group Template.

This Universal Line Template defines the characteristics for all directory numbers that are automatically created for users added through LDAP synchronization.

Design Considerations

The calling search space defined in the Universal Line Template determines the class of service of devices using any of the auto-generated directory numbers. This implies that all directory numbers created through the same LDAP synchronization agreement share the same class of service, and thus if directory numbers for multiple sites and multiple classes of service need to be auto-generated, then multiple LDAP synchronization agreements (one per site and class of service) need to be configured. For each of these synchronization agreements, disjunct LDAP filters need to be defined, each exactly matching on only the users belonging to one of the site-specific and class-of-service-specific user groups. This mapping from LDAP attributes to site and class of service groups can be challenging unless the group membership based on site and class of service is explicitly encoded in few LDAP attributes (potentially even in a custom attribute). Also, the maximum number of supported LDAP agreements is limited, which limits the number of distinct user groups for which directory numbers can be created automatically.

Automatic creation of directory numbers applies only to users created during LDAP directory synchronization. Adding, changing, or updating the Universal Line Template for a given LDAP synchronization agreement will not create directory numbers for already existing users and will not change the settings of already existing directory numbers.

The Universal Line Template allows administrators to define call forward unregistered destinations and either to select voicemail as the forward destination or to define an explicit destination. To reach endpoints in remote sites from registered endpoints in case of WAN failure, the call forward unregistered destination for the remote site's phones must be set to the PSTN alias (+E.164 number) of the remote phone. This cannot be achieved with Universal Line Template settings because this would require defining the call forward unregistered destination to be set based on the assigned directory numbers (potentially with a mask applied).

Security Considerations

During the import of accounts, no passwords or PINs are copied from the LDAP directory to the Unified CM database. If LDAP authentication is not enabled in Unified CM, the password for the end user is managed by using Unified CM Administration. The password and PIN are stored in an encrypted format in the Unified CM database. The PIN is always managed on Unified CM. If you want to use the LDAP directory password to authenticate an end user, see the section on LDAP Authentication.

The connection between the Unified CM publisher server and the directory server can be secured by enabling Secure LDAP (SLDAP) on Unified CM and the LDAP server. Secure LDAP enables LDAP to be sent over a Secure Socket Layer (SSL) connection and can be enabled by adding the LDAP server into the Tomcat trust store within the Unified CM Platform Administration. For detailed procedure steps, refer to the Unified CM product documentation available at http://www.cisco.com. Refer to the documentation of the LDAP directory vendor to determine how to enable SLDAP.

Design Considerations for LDAP Synchronization

Observe the following design and implementation best practices when deploying LDAP synchronization with Cisco Unified CM:

• Use a specific account within the corporate directory to allow the Unified CM synchronization agreement to connect and authenticate to it. Cisco recommends that you use an account dedicated to Unified CM, with minimum permissions set to "read" all user objects within the desired search base and with a password set never to expire. The password for this account in the directory must be kept in synchronization with the password configuration of the account in Unified CM. If the service account password changes in the directory, be sure to update the account configuration in Unified CM.
• All synchronization agreements on a given cluster must integrate with the same family of LDAP servers.
• Stagger the scheduling of synchronization agreements so that multiple agreements are not querying the same LDAP servers simultaneously. Choose synchronization times that occur during quiet periods (off-peak hours).
• If security of user data is required, enable Secure LDAP (SLDAP) by checking the Use SSL field on the LDAP Directory configuration page in Unified CM Administration.
• Ensure that the LDAP directory attribute chosen to map into the Unified CM UserID field is unique within all synchronization agreements for that cluster.
• The attribute chosen as UserID must not be the same as that for any of the Application Users defined in Unified CM.
• The LDAP attribute sn(lastname) is a mandatory attribute for LDAP Synchronization of users.
• An existing account in the Unified CM database before synchronization is maintained only if an account imported from the LDAP directory has a matching attribute. The attribute that is matched to the Unified CM UserID is determined by the synchronization agreement.
• Administer end-user accounts through the LDAP directory's management tools, and manage the Cisco-specific data for those accounts through the Unified CM Administration web page.
• LDAP Synchronization is supported only with Microsoft NT LAN Manager (NTLM). Kerberos and NTLMv2 are not supported.
• For AD deployments, the ObjectGUID is used internally in Unified CM as the key attribute of a user. The attribute in AD that corresponds to the Unified CM User ID may be changed in AD. For example, if sAMAccountname is being used, a user may change their sAMAccountname in AD, and the corresponding user record in Unified CM would be updated.

With all other LDAP platforms, the attribute that is mapped to User ID is the key for that account in Unified CM. Changing that attribute in LDAP will result in a new user being created in Unified CM, and the original user will be marked inactive.

Additional Considerations for Microsoft Active Directory

A synchronization agreement for a domain will not synchronize users outside of that domain nor within a child domain because Unified CM does not follow AD referrals during the synchronization process. The example in Figure 16-9 requires three synchronization agreements to import all of the users. Although Search Base 1 specifies the root of the tree, it will not import users that exist in either of the child domains. Its scope is only VSE.LAB, and separate agreements are configured for the other two domains to import those users.

Figure 16-9 Synchronization with Multiple Active Directory Domains

In Figure 16-9, each of the domains and sub-domains contains at least one domain controller (DC) associated to them, and the three synchronization agreements each specify the appropriate domain controller. The DCs have information only on users within the domain where they reside, therefore three synchronization agreements are required to import all of the users.

When synchronization is enabled with an AD forest containing multiple trees, as shown in Figure 16-10, multiple synchronization agreements are still needed for the same reasons listed above. Additionally, the UserPrincipalName (UPN) attribute is guaranteed by Active Directory to be unique across the forest and must be chosen as the attribute that is mapped to the Unified CM UserID. For additional considerations on the use of the UPN attribute in a multi-tree AD scenario, see the section on Additional Considerations for Microsoft Active Directory.

Figure 16-10 Synchronization with Multiple AD Trees (Discontiguous Namespaces)

Unified CM sends a default LDAP search filter string to AD when performing the synchronization of accounts. One of the clauses is to not return accounts that have been marked as disabled in AD. An account marked disabled by AD, such as when failed login attempts are exceeded, will be marked inactive if synchronization runs while the account is disabled.

Unified CM Multi-Forest LDAP Synchronization

A Unified CM deployment using a multi-forest LDAP infrastructure can be supported by using Active Directory Lightweight Directory Services (AD LDS) as a single forest view integrating with the multiple disparate forests. The integration also requires the use of LDAP filtering (see User Filtering for Directory Synchronization and Authentication). For full details, refer to the document on How to Configure Unified Communication Manager Directory Integration in a Multi-Forest Environment, available at

http://www.cisco.com/en/US/products/sw/voicesw/ps556/products_configuration_example09186a0080b2b103.shtml

LDAP Authentication

The LDAP authentication feature enables Unified CM to authenticate LDAP synchronized users against the corporate LDAP directory. Application users and locally configured users are always authenticated against the local database. Also PINs of all end users are always checked against the local database only. This authentication is accomplished with an LDAPv3 connection established between the Identity Management System (IMS) module within Unified CM and a corporate directory server, as shown in Figure 16-11.

Figure 16-11 Enabling LDAP Authentication

To enable authentication, a single authentication agreement may be defined for the entire cluster. The authentication agreement supports configuration of up to three LDAP servers for redundancy and also supports secure connections LDAP over SSL (SLDAP) if desired. Authentication can be enabled only when LDAP synchronization is properly configured and used. LDAP authentication configuration is overridden by enabling SSO. With SSO enabled, end users are always authenticated using SSO, and LDAP authentication configuration is ignored.

The following statements describe Unified CM's behavior when authentication is enabled:

• End user passwords of users imported from LDAP are authenticated against the corporate directory by a simple bind operation.
• End user passwords for local users are authenticated against the Unified CM database.
• Application user passwords are authenticated against the Unified CM database.
• End user PINs are authenticated against the Unified CM database.

This behavior is in line with the guiding principle of providing single logon functionality for end users while making the operation of the real-time Unified Communications system independent of the availability of the corporate directory, and is shown graphically in Figure 16-12.

Figure 16-12 Authenticating End User Passwords, Application User Passwords, and End User PINs

Figure 16-13 illustrates the following process, adopted by Unified CM to authenticate an end user synchronized from LDAP against a corporate LDAP directory:

1. A user connects to the Unified CM User Options page via HTTPS and attempts to authenticate with a user name and password. In this example, the user name is jsmith.

2. If the user is a local user, the password is checked against the local database.

The following steps apply only to LDAP synchronized users:

3. If the user is an LDAP synchronized user, Unified CM issues an LDAP query for the user name jsmith, using the value specified in the LDAP Search Base on the LDAP Authentication configuration page as the scope for this query. If SLDAP is enabled, this query travels over an SSL connection.

4. The corporate directory server replies via LDAP with the full Distinguished Name (DN) of user jsmith (for example, "cn=jsmith, ou=Users, dc=vse, dc=lab").

5. Unified CM then attempts to validate the user's credentials by using an LDAP bind operation to pass the full DN and password provided by the user.

6. If the LDAP bind is successful, Unified CM allows the user to proceed to the configuration page requested.

Figure 16-13 Authentication Process

Design Considerations for LDAP Authentication

Observe the following design and implementation best-practices when deploying LDAP authentication with Cisco Unified CM:

• Create a specific account within the corporate directory to allow Unified CM to connect and authenticate to it. Cisco recommends that you use an account dedicated to Unified CM, with minimum permissions set to "read" all user objects within the desired search base and with a password set to never expire. The password for this account in the directory must be kept in synchronization with the password configuration of the account in Unified CM. If the account password changes in the directory, be sure to update the account configuration in Unified CM. If LDAP synchronization is also enabled, you can use the same account for both functions.
• Enable LDAP authentication on Unified CM by specifying the credentials of the aforementioned account under LDAP Manager Distinguished Name and LDAP Password, and by specifying the directory subtree where all the users reside under LDAP User Search Base.
• This method provides single logon functionality to all end users synchronized from LDAP. They can then use their corporate directory credentials to log in to the Unified CM User Options page.
• Manage end-user passwords for LDAP synchronized users from within the corporate directory interface. Note that the password field is no longer displayed for LDAP synchronized users in the Unified CM Administration pages when authentication is enabled.
• Manage end-user PINs from the Unified CM Administration web pages or from the Unified CM User Options page.
• Manage Application User passwords from the Unified CM Administration web pages. Remember that these application users facilitate communication and remote call control with other Cisco Unified Communications applications and are not associated with real people.
• Enable single logon for Unified CM administrators by adding their corresponding end user to the Unified CM Super Users user group from the Unified CM Administration web pages. Multiple levels of administrator rights can be defined by creating customized user groups and roles.

Additional Considerations for Microsoft Active Directory

In environments that employ a distributed AD topology with multiple domain controllers geographically distributed, authentication speed might be unacceptable. When the Domain Controller for the authentication agreement does not contain a user account, a search must occur for that user across other domain controllers. If this configuration applies, and login speed is unacceptable, it is possible to set the authentication configuration to use a Global Catalog Server.

An important restriction exists, however. A Global Catalog does not carry the employeeNumber attribute by default. In that case either use Domain Controllers for authentication (beware of the limitations listed above) or update the Global Catalog to include the employeeNumber attribute. Refer to Microsoft Active Directory documentation for details.

To enable queries against the Global Catalog, simply configure the LDAP Server Information in the LDAP Authentication page to point to the IP address or host name of a Domain Controller that has the Global Catalog role enabled, and configure the LDAP port as 3268.

The use of Global Catalog for authentication becomes even more efficient if the users synchronized from Microsoft AD belong to multiple domains, because it allows Unified CM to authenticate users immediately without having to follow referrals. For these cases, point Unified CM to a Global Catalog server and set the LDAP User Search Base to the top of the root domain.

In the case of a Microsoft AD forest that encompasses multiple trees, some additional considerations apply. Because a single LDAP search base cannot cover multiple namespaces, Unified CM must use a different mechanism to authenticate users across these discontiguous namespaces.

As mentioned in the section on LDAP Synchronization, in order to support synchronization with an AD forest that has multiple trees, the UserPrincipalName (UPN) attribute should be used as the user ID within Unified CM. When the user ID is the UPN, the LDAP authentication configuration page within Unified CM Administration does not allow you to enter the LDAP Search Base field, but instead it displays the note, "LDAP user search base is formed using userid information."

In fact, the user search base is derived from the UPN suffix for each user, as shown in Figure 16-14. In this example, a Microsoft Active Directory forest consists of two trees, avvid.info and vse.lab. Because the same user name may appear in both trees, Unified CM has been configured to use the UPN to uniquely identify users in its database during the synchronization and authentication processes.

Figure 16-14 Authentication with Microsoft AD Forests with Multiple Trees

As shown in Figure 16-14, a user named John Doe exists in both the avvid.info tree and the vse.lab tree. The following steps illustrate the authentication process for the first user, whose UPN is jdoe@avvid.info:

1. The user authenticates to Unified CM via HTTPS with its user name (which corresponds to the UPN) and password.

2. Unified CM performs an LDAP query against a Microsoft Active Directory Global Catalog server, using the user name specified in the UPN (anything before the @ sign) and deriving the LDAP search base from the UPN suffix (anything after the @ sign). In this case, the user name is jdoe and the LDAP search base is "dc=avvid, dc=info".

3. Microsoft Active Directory identifies the correct Distinguished Name corresponding to the user name in the tree specified by the LDAP query. In this case, "cn=jdoe, ou=Users, dc=avvid, dc=info".

4. Microsoft Active Directory responds via LDAP to Unified CM with the full Distinguished Name for this user.

5. Unified CM attempts an LDAP bind with the Distinguished Name provided and the password initially entered by the user, and the authentication process then continues as in the standard case shown in Figure 16-13.

Note Support for LDAP authentication with Microsoft AD forests containing multiple trees relies exclusively on the approach described above. Therefore, support is limited to deployments where the UPN suffix of a user corresponds to the root domain of the tree where the user resides. AD allows the use of aliases, which allows a different UPN suffix. If the UPN suffix is disjointed from the actual namespace of the tree, it is not possible to authenticate Unified CM users against the entire Microsoft Active Directory forest. (It is, however, still possible to use a different attribute as user ID and limit the integration to a single tree within the forest.)

User Filtering for Directory Synchronization and Authentication

Unified CM provides an LDAP Query Filter to optimize directory synchronization performance. Cisco recommends importing those directory user accounts that will be assigned to Unified Communications resources. To allow for enterprise-wide UDS based service discovery, all users assigned to Unified Communications resources on any cluster in the enterprise need to be imported to all clusters in the enterprise. Differentiation between local and remote users is achieved by the Home Cluster setting on the Feature Group Template associated with the LDAP synchronization agreement that is used. When the number of directory user accounts exceeds the number supported for an individual cluster, filtering must be used to select the subset of users that will be associated on that cluster. The Unified CM synchronization feature is not meant to replace a large-scale corporate directory.

In many cases, a unique search base is all that is needed to control which accounts are synchronized. When a unique search base is not available, a custom LDAP filter might be required. The information in the following sections addresses both methods that can be used to optimize directory synchronization. When any mechanism is used to limit the accounts imported into Unified CM, the default directory lookup configuration will list only those directory entries that exist in the Unified CM database. For directory lookup to access the entire directory, you must configure Unified CM to utilize an external web server. Details of this configuration are not discussed here but are discussed in the Unified CM product documentation available at

http://www.cisco.com/en/US/partner/products/sw/voicesw/ps556/tsd_products_support_series_home.html

Optimizing Unified CM Database Synchronization

The Unified CM Database Synchronization feature provides a mechanism for importing a subset of the user configuration data (attributes) from the LDAP directory store into the Unified CM publisher database. Once synchronization of a user account has occurred, the copy of each user's LDAP account information may then be associated to additional data required to enable specific Unified Communications features for that user. When authentication is also enabled, the user's credentials are used to bind to the LDAP store for password verification. The end user's password is never stored in the Unified CM database when enabled for synchronization and/or authentication.

User account information is cluster-specific. Each Unified CM publisher server maintains a unique list of those users receiving Unified Communications services from that cluster. Synchronization agreements are cluster-specific, and each publisher has its own unique copy of user account information. Only those users who will be assigned Unified Communications resources should be synchronized with Unified CM. The following is a partial list of common reasons why the entire set of users defined in the LDAP directory should not be imported into the Unified CM cluster:

• Importing users who will not be assigned Unified Communications resources can increase directory synchronization time.
• Importing users who will not be assigned Unified Communications resources can slow Unified CM searches and overall database performance.
• In many cases, the number of user accounts in the LDAP directory store far exceeds the total user capacity of the Unified CM database.

Unified CM has no enforced limit on the number of accounts that may be added to the system. Cisco recommends limiting the number of users to twice the supported number of endpoints. There might be cases where accounts are needed for applications, and some designs might require additional accounts.

Cisco recommends using the control mechanisms described here to minimize the number of user accounts imported, regardless of the LDAP database size. This will improve the speed of the first and subsequent periodic synchronizations and will also improve manageability of the user accounts.

Using the LDAP Structure to Control Synchronization

Many deployments of LDAP directories use the Organizational Unit Name (OU) to group users into a logical order and sometimes hierarchical order. If the LDAP directory has a structure that organizes users into multiple OUs, then it often is possible to use that structure to control the groups of users imported. Each individual Unified CM synchronization agreement specifies a single OU. All active accounts under the specified OU, even within sub-OUs, are imported. Only those users in the OU are synchronized. When multiple OUs containing users are required in a cluster, multiple synchronization agreements are required. When an OU contains users that will not be assigned Unified Communications resources, Cisco recommends omitting those OUs from the directory synchronization.

The same technique may be used with AD, which defines containers. A synchronization agreement may specify a particular container in the directory tree and thereby limit the extent of the import.

Because there is only a limited number of synchronization agreements available, LDAP deployments with many OUs or containers can quickly exhaust this technique. One possible method to synchronize users in a multi-OU environment is to control the permissions assigned to the synchronization service account. Configure the synchronization agreement to a tree node that contains a mix of users, and then restrict the system account from read access to selected parts of the subtree. Refer to your LDAP vendor documentation on how to restrict this access.

LDAP Query

Additional control over filtering might be required for any of the following reasons:

• The LDAP directory has a flat structure that does not enable adequate control by configuration of the synchronization agreements. When the aggregate number of users that are imported by all the synchronization agreements is greater than the maximum number of users supported by the Unified CM cluster, then it is necessary to control the number of users imported through filters.
• You want to import a subset of user accounts into the Unified CM cluster, for administrative segmentation of users, to control a subset of users that have access and authentication to the cluster. Any account that is imported into a cluster has some level of access to the web pages and authentication mechanisms, which might not be desirable in some cases.
• The LDAP directory structure does not have an accurate representation of how users are going to be mapped into the Unified CM clusters. For instance, if OUs are set up according to an organizational hierarchy but users are mapped to Unified CM by geography, there might be little overlap between the two.

In these cases, the LDAP Query filter may be used to provide additional control over the synchronization agreements.

LDAP Query Filter Syntax and Server-Side Filtering

Unified CM uses standard LDAP mechanisms for synchronizing data from an LDAP directory store. It utilizes the Search mechanism, as defined by RFC 2251–Lightweight Directory Access Protocol (v3), to send a request to retrieve data from the LDAP server. Also defined by that mechanism is the ability to specify a filter string inside the Search message that is used by the LDAP server to select entries in the database for which to return data. The syntax of the filter string is defined by RFC 2254, The String Representation of LDAP Search Filters. This RFC may be used as a reference for constructing more complex filter strings.

The filter string is embedded within a Search message that is sent by Unified CM to the LDAP server and is executed by the server to select which user accounts will be provided in the response.

Simple Filter Syntax

You can configure a filter by specifying standard attribute names and values that are desired for those attributes. The attributes may also be specified by DN element instead of name. The filter string that is used by Unified CM in LDAP queries is stored internally in the ldapfilter table and is the string inserted into the Search message.

A filter is a UTF-8 formatted string that has the following syntax:

( attribute operator value )

or

( operator ( filter1 )( filter2 ))

Where filter1 and filter2 have the syntax shown in first line, and the operator is one of those listed in Table 16-7 . The attribute corresponds to an LDAP attribute that exists in the directory, operator is one of the operators listed in Table 16-7 , and value corresponds to the actual data value that is requested for the attribute.

An attribute specified in the filter can be any attribute that exists in the LDAP directory store, and it does not have to be one of the attributes that is understood and imported by Unified CM. The attribute is used only on the LDAP server to select data, and the corresponding entries will have a subset of their data imported into Unified CM.

Example 16-1 A Single Condition

(givenName=Jack)

The filter in Example 16-1 selects any user with a given name of Jack.

Example 16-2 Multiple Conditions May Be Joined with Logical Characters

(&(objectclass=user)(department=Engineering))

The filter in Example 16-2 selects all users in the engineering department.

Default Filter Strings

If no custom filter strings are defined, Unified CM uses a default LDAP filter string as follows:

• Default Active Directory (AD) filter string

(&(objectclass=user)(!(objectclass=Computer))(!(UserAccountControl:1.2.840.113556.1.4.803:=2)))

This default filter selects entries for which the object class is a user but not a computer, and for which the account is not flagged as disabled.

• Default SunONE and Oracle Directory Server filter string

(objectclass=inetOrgPerson)

This default filter selects all users for which the object class is inetOrgPerson.

• Default OpenLDAP filter string

(objectclass=inetOrgPerson)

• Default Active Directory Application Mode (ADAM) or Active Directory Lightweight Directory Services (AD LDS) filter string

(&(objectclass=user)((objectclass=Computer))(!(msDS-UserAccountDisabled=TRUE)))

Extending the Default Filter

Cisco recommends that you use the default filter string and append additional conditions to it. For example:

(&(objectclass=user)(!(objectclass=Computer))(!(UserAccountControl:1.2.840.113556.1.4.803:=2))(telephonenumber=919*))

This filter selects only users that have a prefix of 919 in their telephonenumber field. The synchronization agreement will import only users with an area code of 919. This example assumes all entries begin with an area code.

For the search filter, you may use any existing attribute or even a custom attribute that is defined in the LDAP directory store. The filter string controls which records are selected by the LDAP server to be returned to Unified CM, but the attributes that are imported are not affected by the filter string.

Custom LDAP filter strings can be up to 2048 characters long. Custom LDAP filters first need to be created, and then existing custom LDAP filters can be assigned to LDAP synchronization agreements. Different LDAP synchronization agreements can use different custom LDAP filters.

High Availability

Unified CM LDAP Synchronization allows for the configuration of up to three redundant LDAP servers for each directory synchronization agreement. Unified CM LDAP Authentication allows for the configuration of up to three redundant LDAP servers for a single authentication agreement. You should configure a minimum of two LDAP servers for redundancy. The LDAP servers can be configured with IP addresses instead of host names to eliminate dependencies on Domain Name System (DNS) availability.

Capacity Planning for Unified CM Database Synchronization

The Unified CM Database Synchronization feature provides a mechanism for importing a subset of the user configuration data (attributes) from the LDAP store into the Unified CM publisher database. Once synchronization of a user account has occurred, the copy of each user's LDAP account information may then be associated to additional data required to enable specific Unified Communications features for that user. When authentication is also enabled, the user's credentials are used to bind to the LDAP store for password verification. The end user's password is never stored in the Unified CM database when enabled for synchronization and/or authentication.

User account information is cluster-specific. Each Unified CM publisher server maintains a unique list of those users receiving Unified Communications services from that cluster. Synchronization agreements are cluster-specific, and each publisher has its own unique copy of user account information.

The maximum number of users that a Unified CM cluster can handle is limited by the maximum size of the internal configuration database that gets replicated between the cluster members. The maximum number of users that can be configured or synchronized is 160,000. With more than 80,000 users the maximum number of LDAP synchronization agreements is limited to 10, while with less than 80,000 users the total number of LDAP synchronization agreements is limited to 20. To optimize directory synchronization performance, Cisco recommends considering the following points:

• Directory lookup from phones and web pages may use the Unified CM database or the IP Phone Service SDK. When directory lookup functionality uses the Unified CM database, only users who were configured or synchronized from the LDAP store are shown in the directory. If a subset of users are synchronized, then only that subset of users are seen on directory lookup.
• When the IP Phone Services SDK is used for directory lookup, but authentication of Unified CM users to LDAP is needed, the synchronization can be limited to the subset of users who would log in to the Unified CM cluster.
• If only one cluster exists, and the LDAP store contains fewer than the maximum number of users supported by the Unified CM cluster, and directory lookup is implemented to the Unified CM database, then it is possible to import the entire LDAP directory.
• When multiple clusters exist and the number of users in LDAP is less than the maximum number of users supported by the Unified CM cluster, it is possible to import all users into every cluster to ensure directory lookup has all entries.
• If the number of user accounts in LDAP exceeds the maximum number of users supported by the Unified CM cluster and the entire user set should be visible to all users, it will be necessary to use the Unified IP Phone Services SDK to off-load the directory lookup from Unified CM.
• If both synchronization and authentication are enabled, user accounts that have either been configured or synchronized into the Unified CM database will be able to log in to that cluster. The decision about which users to synchronize will impact the decision on directory lookup support.

Note Cisco supports the synchronization of user accounts up to the limit mentioned above, but it does not enforce this limit. Synchronizing more user accounts can lead to starvation of disk space, slower database performance, and longer upgrade times.

SAML Authentication

The actors in generic SAML authentication flow are:

• Client — Browser-based user client used to access the service
• SP — Application or service the user tries to access
• IdP — Entity performing the user authentication based on user credentials. The actual credentials and the actual authentication mechanism are hidden by the IdP. The IdP issues SAML assertions based on the authentication process result.

SAML defines a number of profiles to describe the use of SAML to solve typical use cases. The relevant profile used for SSO with Cisco Collaboration services is the web browser SSO profile of SAML V2.0.

The use case solved by this profile is the multi-domain web single sign-on, illustrated in Figure 16-16. In this use case, a user already has a login session with some web service (for example, travel.example.org) and is using this service. As part of the login process, a security context has been established for travel.example.org. If the same user now moves to another web service (for example, sales.example.de) and a business agreement exists between travel.example.org and sales.example.de that establishes a federated identity for the user between these services, then the user is able to access the web service sales.example.de without having to provide authentication credentials again. In this case the identity provider site (travel.example.org) asserts to the service provider site (sales.example.de) that the user is known, has been properly authenticated, and has certain identity attributes. The service provider site (sales.example.de) trusts this assertion based on the existing business agreement between the sites and grants access to the service.

Figure 16-16 Multi-Domain Web Single Sign-On

This description implies that the user first is authenticated by a web service and that this first web service then provides an identity assertion to enable the user to access the second web service. The web service accessed first (travel.example.org) acts as the IdP for SP sales.example.de. This is known as IdP initiated web SSO.

The more typical web SSO flow used with Cisco Collaboration Services is SP initiated web SSO, illustrated in Figure 16-17. In this case the user directly (without visiting an IdP first) tries to access a protected resource on an SP. The SP sends the user to the IdP to get authenticated, and then the user presents the authentication assertion received from the IdP to the SP to get access.

Figure 16-17 Web SSO Initiated by a Service Provider

The SAML web browser SSO profile provides a variety of options depending on whether the authentication is initiated by the IdP or SP and on how the messages are exchanged between IdP and SP. As mentioned above, Cisco Collaboration services use SP initiated SSO only where the SP sends a user to an IdP first to authenticate when the user is trying to access a protected resource and does not have an active session with the service provider. The IdP then builds an authentication assertion and sends the user back to the SP with that assertion.

The binding used for the messages exchange between IdP and SP for Cisco Collaboration services is the Redirect/POST binding, illustrated in Figure 16-18. Here an HTTP 302 redirect is used to send the SAML authentication request message from the SP to the IdP, and the authentication response from IdP to SP is sent using an HTTP POST message.

Figure 16-18 SP-Initiated SSO (Redirect/POST Binding)

The general steps of the SAML authentication flow are:

1. The user tries to access a service or resource by pointing the browser to the URL hosted on the application server. The browser at this moment does not have an active session with the service.

2. The SP realizes that the request originates from a client without an active session. Because HTTP is stateless, an active session can be detected by the SP only if the client sends a session cookie that has been issued by the SP earlier. Based on the SSO configuration, the SP now generates an SAML authentication request to be sent to the appropriate IdP defined as part of the SSO configuration. The SAML request contains information about the SP generating the request. This is required so that the IdP can identify the SPs sending SAML requests.

The SP does not communicate directly with the IdP to authenticate the user. Instead the SP redirects the browser to the IdP. The URL used for this redirect is taken from the IdP metadata exchanged earlier. The SAML request to be sent to the IdP is included in the redirect as a URL query parameter using Base64 encoding.

This redirecting HTTP 302 might look like this:

The encoded SAML authentication request shown above can be decoded to:

Among other details specifying authentication parameters and identifying the requesting SP, the above SAML authentication request also specifies the Assertion Consumer Service (ACS) URL. The ACS URL is the URL to which the SAML Authentication Response needs to be POSTed at the end of the authentication process.

3. The browser receives the redirect, follows the URL, and issues the corresponding GET to the IdP. The SAML request is maintained. The browser at this stage does not have an active session with the IdP.

4. After receiving the new request from a browser with no active session (browser is not sending a cookie issued by the IdP earlier), the IdP authenticates the user based on the pre-configured authentication mechanisms. Possible authentication mechanisms include user/password, PKI/CAC, or Kerberos. For user/password authentication, the IdP might push a form to the user to enter the credentials (for example, “200 OK” message with an IdP login form). For the actual authentication, the IdP might depend on back-end systems such as an LDAP server for user/password authentication.

One key point here is that the exchange of credentials for the purpose of authentication takes place between the IdP and the browser. The SP is not involved and does not see the credentials.

5. The browser provides further information required for the authentication process. For the user/password case, this would be a POST with the information. For other authentication mechanisms, other details would need to be sent to the IdP by the browser.

6. The IdP now checks and validates the provided credentials. The check could involve interactions with respective back-end systems (LDAP bind for user/password authentication against LDAP, communication with Kerberos server to validate ticket, and so forth).

Finally the IdP generates an SAML response for the SP. This response contains the SAML assertion documenting the result of the authentication process. The SAML assertion, in addition to the basic Yes/No information, also contains validity information and information about attributes describing the authenticated entity. At least the user ID of the authenticated entity has to be included in the well known attribute uid so that the SP can extract this information from the assertion to relate the authenticated entity to users existing in the local database.

The SAML assertion is signed by the IdP according to the SSO key information published in the IdP metadata. This ensures that the SP can verify the authenticity of the SAML assertion.

The IdP returns the SAML assertion to the browser in a hidden form in a 200 OK message. The hidden form instructs the browser to POST the SAML assertion to the Assertion Consumer Service (ACS) URL of the SP.

The IdP has to establish a security context so that future authentication requests from the same browser can be answered without going through the exchange of credentials. The IdP will then realize that it already has a valid session with the browser and will assert the authentication of the previously authenticated user without prompting for credentials again. This basically enables SSO against multiple SPs.

7. The browser follows the hidden POST received in the 200 OK message and POSTs the SAML assertion to the Assertion Consumer Service on the SP.

8. The SP extracts the SAML assertion from the POST and validates the signature of the assertion. This guarantees the authenticity of the SAML assertion and the IdP. The user identifier received in the SAML assertion in attribute uid is then used to decide whether the user is authorized to access the requested service. This is based on local access control configuration on the SP.

The SP grants access to the requested resource and sends back the content in a 200 OK message to the browser. The SP also sets a session cookie in the browser so that, for subsequent access requests from the same browser to the same SP, the SP does not have to initiate any more exchanges with the IdP. The IdP will be involved with additional requests from the same browser only after the SP session has expired.

Authentication Mechanisms for Web-Based Applications

When SSO is enabled for a collaboration service, any access to the respective service will be authenticated using SSO. As a fallback measure, a vanity or recovery URL also exists on the landing page. The vanity URL bypasses the SSO mechanism and provides access to all administrator GUIs. Access to the administrator GUI through the vanity URL is authenticated against the local user database. Access to the GUI through the vanity URL can be disabled on the CLI using the utils sso recovery-url disable command.

The vanity URL can be used as a recovery back door when there is an issue with the SAML infrastructure, such as when the IdP is unreachable or down, when there are metadata issues (for example, expired signing certificates), or when there are IdP configuration changes.

Collaboration services currently support the following user types:

• OS user

This user is specified during installation and has access to the CLI, the Disaster Recovery System (DRS) GUI, and the OS Admin GUI. Access to the CLI, DRS, and OS Admin GUI always is authenticated locally. The password is stored in the local database.

• Application user

These are functional users created and managed locally. Passwords are stored in the local database. Application users are not enabled for SSO. With SSO enabled, application users can get access to only the Admin GUI through the vanity URL on the landing page.

• Local end user

These users are created and managed locally. Passwords are stored locally. These users do not exist in the enterprise identity management system. If SSO is enabled, local end users cannot authenticate successfully. Support for local end users and LDAP synced users at the same time with SSO enabled will be reintroduced at a later time. Local end users and LDAP synced users without SSO enabled are still supported.

• LDAP synced end user

These users are managed in the corporate LDAP directory and are synchronized into the Unified Communications service through LDAP sync agreements. For every LDAP synced end user in the local database, there is a matching user in the corporate LDAP directory. If SSO is disabled, the passwords of LDAP synced end users are validated through an LDAP bind operation. With SSO enabled, the authentication of LDAP synced users is based on the authentication mechanism defined on the IdP, and authorization is based on local configuration. An LDAP synced end user has to have the proper rights assigned locally to be able to access the requested resource.

PIN-based authentication is always (even with SSO enabled) based on local configuration. Multiple collaboration services maintain individual PINs.

The following web services are enabled for SSO based on SAML IdP redirects:

• Cisco Unified Communications Manager Admin GUI
• Cisco Unified CM Self Care Portal
• Cisco Unified Communications Manager Serviceability GUI
• Cisco Unified Communications Manager Reporting Tool GUI
• Cisco Unified Communications Manager IM&P Admin GUI
• Cisco Unity Connection Admin GUI
• Cisco Unified Personal Communicator Assistant
• Cisco Unity Connection WebInBox

OAuth 2.0

OAuth is an open framework for authorization that provides delegated access to services on behalf of resource owners. The protocol allows authorization of clients to access resources without sharing the credentials used by the clients for authentication. OAuth essentially allows access tokens to be issued to clients by a central authorization instance. These tokens then are presented by the clients to the servers offering a service. The service providers then ask the authorization instance for validation of the presented tokens to verify the authorization. In this process the content of the authorization token, as well as the mechanism used to authenticate the client and issue the token, is completely transparent for the service provider.

To understand the operation of OAuth, a few definitions need to be introduced.

An OAuth Client is the entity that requests an OAuth access token from the OAuth Authorization service. Depending on the particular use case, an OAuth Client may be a Collaboration service (for example, the Collaboration Edge) or an end user client (for example, Jabber). If Collaboration Edge requests an OAuth token on behalf of a user, then Collaboration Edge acts as an OAuth client. In the case of a Jabber client login flow inside the enterprise, the Jabber client acts as the OAuth client.

Every OAuth client has an unique identifier, the OAuth client_id. This OAuth client_id uniquely identifies a client type. For example, Jabber for Windows and Jabber for Android use different client_ids, but all releases of Jabber for Windows use the same client_id unless concrete reasons mandate a changed client_id to enable the authorization service to differentiate between different client releases (for example, support variation in the OAuth exchange with different client releases). A set of client_ids is predefined for Cisco products and also for a third-party client.

For each given client_id an OAuth redirect_uri exists, to which the authorization service returns issued tokens. For all predefined client_ids, the redirect_uri is fixed to /ssosp/public/oauthcb.

When requesting authorization to a protected resource, an OAuth client might request a token with a particular scope. The scope indicates the range of services that an OAuth token can be used to access.

An OAuth Access Token is granted by the authorization service and is used by bearers (clients) for access to a protected resource. Typically access tokens are issued to a specific user and have a specific expiration time. Whenever an access token expires, the client must get a new access token.

OAuth supports a number of call flows, but only two are relevant for the Collaboration architecture:

• Implicit Grant Flow — An end user's client (for example, Jabber) acts as the OAuth Client and requests an OAuth token directly from the authorization service. This is the main grant flow used by collaboration clients accessing Unified Communications APIs.
• SAML Bearer Assertion Grant Flow — An entity (typically a service) uses an assertion issued on behalf of an end user to get an OAuth token that is associated with the end user. This grant flow is used by Collaboration Edge to get tokens on behalf of clients connecting from outside the edge.

OAuth tokens can be used in HTTP based flows by sending them in an Authorization header. OAuth tokens are transported in an Authorization header marked with the "Bearer" keyword.

SSO for Jabber and Other Endpoints

For users accessing any of the SSO enabled Web interfaces, authentication is based on the SAML exchange between the user's browser, the web server, and the IdP. After successful authentication, the web server (in case of accessing the Unified CM administration GUI this would be the administration application running on Unified CM) consults the local configuration to determine whether the authenticated user is authorized to access the given resource. For example, if user Bob provided valid credentials to the IdP and thus authenticated successfully, then Unified CM could still deny access to the Unified CM administration interface if Bob is not a member of the "Standard CCM Super Users" group. In this case Bob would only get a prompt indicating that he does not have the required privileges to access the system, instead of getting access to Unified CM administration.

Jabber clients and other endpoints require access to a number of collaboration interfaces (for example, Unified CM SIP, Unified CM CTI, and Unified CM IM&P SOAP) using SSO. After authenticating successfully once against the IdP, access to these collaboration interfaces has to be provided without requiring further authentication. Because all interactions with the IdP are browser based, all clients supporting SSO also need to support a mechanism to delegate the actual authentication process to a browser. The browser then hides the details of the actual SAML 2.0 authentication process from the client implementation (for example, the login procedure, the kind of credentials provided, and the authentication method used).

All Jabber platforms use embedded browser controls for SSO. These controls rely on underlying operating system browser technologies (see Table 16-8 ). Although on desktop operating systems cookies are shared between the embedded browser control used by Jabber and the underlying operating system browser, this does not guarantee cross-browser SSO. Even after successful authentication with the operating system browser against the IdP, Jabber using an embedded browser instance would also need to authenticate against the IdP to get a valid IdP cookie.

Jabber and other endpoints supporting SSO at startup need to get an OAuth authorization token to get access to the required collaboration APIs. The mechanism to acquire this OAuth authorization token is an OAuth Implicit Grant Flow, as shown in figure Figure 16-19.

Figure 16-19 OAuth Implicit Grant Flow

The components in this flow are the web browser (embedded browser instance), the collaboration service (Unified CM with IM and Presence, and Unity Connection), the authorization service, and the IdP.

In step 1 the client, via the embedded browser instance, accesses the Implicit Grant API by sending an HTTP GET request to the /authorize endpoint on the authorization server. This request at least has to indicate the desired response type (response_type) and also has to have client_id to identify the type of client. Clients can also indicate a scope to request a token (for example, scope=UnifiedCommunications).

Assuming that the client does not have an active session with the authorization service (no valid authorization service session cookie provided in the request), the authorization service initiates an SP-initiated SAML 2.0 redirect/POST authentication flow by returning a 302 response redirecting the browser to the IdP in step 2.

Browser and IdP will then exchange the messages required to authenticate the user. The message exchange in step 3 depends on the authentication method configured on the IdP.

If the SAML authentication succeeds, then as the last step of the SAML exchange the browser POSTs the SAML assertion to the assertion consumer service on the authorization service in step 4.

After authentication succeeds, the authorization service can then check whether the authenticated user is eligible for the requested scope. If that is the case, then the authorization service will grant an authorization token, which is returned in step 5 as fragment of a location sent back in an HTTP 302 message to the browser. With the 302 sent back, an authorization service session cookie also is returned to the browser.

In step 6 the browser follows the redirect and sends a GET to the OAuth callback service.

This service in step 7 returns a 200 OK with a web page indicating successful authorization. This returned web page might also contain JavaScript code to redirect the browser to some URL with some locally registered service type. This mechanism can be used to extract the granted OAuth token and make it available for the application that initiated the Implicit Grant Flow. This mechanism is not required for clients using embedded browser instances, in which case the client simply waits for the web transaction to finish and then extracts the authorization token as a fragment of the OAuth callback URL to which the client was last redirected.

The OAuth token obtained from the Implicit Grant flow can then be used when accessing resources. Figure 16-20 shows how an OAuth authentication token is used to provide authentication for a web service.

Figure 16-20 Using OAuth Token to Authenticate to a Web Service

The client browser presents the OAuth token in an HTTP Authorization header to the web service as part of the HTTP GET. In step 2 the web service asks the authorization service to validate the access token. The authorization service returns a 200 OK if the access token is valid. This 200 OK also includes the user ID for which the token initially was granted, so that the web service can verify that the token is not abused by a different user. If all checks succeed, then the service returns the requested resource to the client.

The login procedures of collaboration interfaces not based on HTTP (for example, IM&P SOAP and IM&P XMPP) are extended to support OAuth tokens. Clients using these interfaces present the OAuth token during the login process, and after validating the token with the authorization service, then access to the interface is granted.

OAuth tokens used in the Cisco Collaboration solution have a default lifetime of one hour (3,600 seconds). If an expired token is used to authorize a request, then the authorization service returns a 401 error indicating that the token is expired. The client then has to refresh the token by calling the /authorize endpoint again. If the embedded browser of the client now also passes a valid authorization service session cookie, then the authorization service will not initiate a new SAML authentication flow, but instead issues a new OAuth token right away. To avoid this forced OAuth token refresh, Cisco Collaboration clients initiate a token refresh after 75% of the token expiration time has elapsed. Note that OAuth token expiration does not affect session-based protocols such as XMPP and SIP.

Currently Cisco Unified CM with IM and Presence and Cisco Unity Connection run separate authorization service instances, which means that OAuth tokens achieved from the authorization service running on Unified CM cannot be used to access APIs on Unity Connection (for example, VMRest). Without sharing of OAuth tokens between Cisco Unity Connection and Unified CM with IM and Presence, clients requiring access to interfaces on both Unified CM with IM and Presence and Unity Connection must access the separate authorization services on Unified CM with IM and Presence and on Unity Connection to achieve separate authorization tokens.

SSO with Collaboration Edge

Single sign-on for clients outside the enterprise connected using mobile remote access through collaboration edge is achieved with Expressway-E and Expressway-C, as illustrated in Figure 16-21.

Figure 16-21 SSO with Collaboration Edge

Endpoints and collaboration clients registering through the collaboration edge can use SSO to access collaboration applications inside the enterprise. During endpoint registration, the SAML redirect/POST SAML authentication flow is initiated by Expressway-C (proxied by Expressway-E), and the browser on the client is redirected to a publicly accessible identity provider. This typically is an IdP proxy in the customer's DMZ acting as proxy for the IdP deployed inside the enterprise. The IdP proxy function essentially is only an HTTPS reverse proxy for the enterprise IdP. If the customer uses Identity as a Service (IDaaS), then the IdP used in the SAML exchange is located in the public Internet. In this case no IdP proxy in the customer's DMZ is required. Expressway-E and Expressway-C proxy only collaboration client requests for services on the collaboration applications. While the SAML authentication flow is redirected to an IdP proxy in the DMZ, the OAuth exchange to achieve a OAuth token passes through Expressway-C and Expressway-E, and Expressway-E requests the OAuth token as a proxy for the actual client. This is a variant of an OAuth SAML Bearer Grant flow as shown in Figure 16-22.

Figure 16-22 OAuth SAML Bearer Grant Flow with Collaboration Edge

To acquire an OAuth token, the browser sends an HTTP GET request to the /authorize endpoint on Edge. The /authorize endpoint on Edge is accessed using prefix encoding to refer to the customer domain. Edge in this description refers to a Expressway-C and Expressway-E pair implementing the collaboration edge.

Edge in step 2 now initiates an SP-initiated SAML 2.0 redirect/POST authentication flow by returning a 302 response redirecting the browser to the IdP. Edge also caches the authorization parameters from the client request because they are needed later in the actual OAuth proxy request.

Browser and IdP then exchange the messages required to authenticate the user. The message exchange in step 3 depends on the authentication method configured on the IdP.

If the SAML authentication succeeds, then as the last step of the SAML exchange the browser POSTs the SAML assertion to the assertion consumer service on Edge in step 4. Edge now still needs to exchange this SAML assertion for an OAuth token.

To achieve this, Edge in step 5 uses the /authorize_proxy endpoint on the authorization service. This request is authenticated using the username and password of an application user. The referenced application user has to have rights to access the AXL API on the authorization service on Unified CM and Unity Connection, respectively. The /authorize_proxy request contains all authorization parameters cached earlier.

The authorization service then can check whether the authenticated end user has the required privileges for the requested scope. If the authenticated user is authorized to access the requested scope, then the authorization service issues an OAuth token and in step 6 returns that token in a 200 OK.

Edge can then cache the token and still needs to return the token to the client. This in step 7 is achieved by returning a 302 to the client, redirecting the browser instance on the client to the OAuth callback on Edge. The target URL of the redirect contains the required information about the token, including the token itself and the token lifetime.

The embedded browser on the client in step 8 follows the redirection and accesses the OAuth callback resource on Edge. The 200 OK returned in step 9 finishes the SSO flow through Edge. The client can now extract the OAuth token from the final URL.

The acquired OAuth token is used to authorize all further client requests through Edge.

Understanding Session and Token Expiration Timers

Session and token expiration timers affect both the frequency of assertion and token renewals as the user experience. As described above, when receiving a request from a client, the service, the IdP, and the authorization service check whether the requesting client already has an active session. A session context is established by issuing session cookies. The lifetime of a session context on a Cisco Collaboration service is determined by the UC service session timer. A session on a Collaboration service ends if a client does not send further requests for longer than the UC service session timer. The same logic exists for sessions with the IdP and the authorization service. If a Cisco Collaboration service session expires, the Collaboration service initiates a new SAML authentication flow. As long as the IdP session has not expired yet, the IdP will issue a new SAML authentication token immediately. This also applies to the authorization service. The authorization service immediately issues an OAuth token without initiating a SAML authentication flow, as long as the browser requesting the OAuth token still has an active session with the authorization service. If no active authorization service session exists and the browser is redirected to the IdP for authentication, then the IdP would still silently authenticate the client and issue a new SAML assertion as long as the IdP session is still active.

Session and OAuth token expiration times interact as follows:

• When the OAuth token expires, the client will request a new token from the Authz service.
• If the Authz session has expired, the client will request a new Assertion from the IdP.
• If the IdP session has expired, the IdP will challenge the client to reenter credentials.

All three timer values impact network load and throughput. Whenever a timer expires, the client must signal over the network to get a new token or establish a new session.

The IdP session timer impacts user experience. If the user is redirected to the IdP due to OAuth token and Authz session expiration, and if the IdP session has also expired, the user will be challenged for credentials.

Recommended timer values are:

• OAuth access token timeout: 1 hour
• UC service session timeout: 30 minutes
• IdP idle timeout: 8 hours
• IdP session timeout: 48 hours

While the IdP timers are configured on the IdP, the UC service session timeout is configured on the individual UC services. The CLI command set webapp session timeout can be used on Unified CM, Unified CM IM and Presence, and Unity Connection servers to set the UC service session timeout for all web services running on a specific server. The UC service session timeout needs to be configured on each server individually, and it affects all web-based services including the authorization service. The OAuth token expiration timer is configured in the enterprise parameters on Unified CM and Unity Connection.

Decreasing the OAuth access token timeout to less than 40 minutes leads to an interesting situation in that, whenever a client tries to refresh the OAuth token after 75% of the OAuth token expiration, the default UC service session timeout of 30 minutes for the authorization service has not expired yet, so that the authorization service immediately issues a new token without initiating a new SAML authentication flow. While this minimizes the complexity of the reauthorization flow, the lack of re-initiating a new SAML authentication flow makes it impossible to de-authenticate currently logged in users.

Design Considerations for SSO

SAML SSO always has to be enabled or disabled for all nodes in a cluster. Either all nodes or no nodes in a cluster are enabled for SSO. Enabling SSO through the Admin GUI automatically enables SSO for all existing nodes at the same time. As part of this process, SP metadata for all cluster nodes is downloaded. Each cluster node needs to be represented on the IdP as an individual SP. If a node is later added to a cluster that is already in SSO mode, the metadata of this added node has to be imported into the IdP to complete the list of defined SPs on the IdP.

IdP metadata has to be imported into all SAML SPs. When SSO is enabled on the Admin GUI, the IdP metadata provided in the process is automatically imported on all nodes of the cluster.

The SP metadata does not include the optional ContactPerson information; therefore, the IdP will not be able to expose contact information for Cisco Collaboration SPs.

SAML SP can request signed assertions from the IdP by including WantAssertionsSigned in the SAML AuthnRequest. Currently Cisco Collaboration SPs do not send this information. This gives the IdP full control over assertion signing. Cisco recommends activating SAML Assertion signing on the IdP. Cisco products use their Tomcat certificate for signing SAML requests.

Availability of the IdP is a key requirement when using SSO. The IdP has to be deployed with full redundancy and fault tolerance. Essential for this kind of deployment is that the IdP is deployed with a single logical URL and that suitable load balancers and web server farms are deployed to make sure that the single IdP URL is highly available. The single IdP URL is included in the IdP metadata and is imported into all Cisco Collaboration SPs. Failure of a single element (for example, a single web server) should be invisible to the Collaboration service.

SAML requests and assertions are signed using SP and IdP certificates. The lifetime of these certificates has to be closely monitored to make sure that the SAML SSO mechanism continues to work.

SAML assertions contain validity information (NotBefore, NotOnOrAfter). To make sure that valid assertions are not rejected due to timing issues, it is essential that all services are synchronized using appropriate mechanisms such as Network Time Protocol (NTP).