Cisco Nexus Dashboard Fabric Controller Verified Scalability

Verified Scale Limits for Release 12.1.3b

This section provides verified scalability values for various deployment types for Cisco Nexus Dashboard Fabric Controller, Release 12.1.3b.

The values are validated on testbeds that are enabled with a reasonable number of features and aren’t theoretical system limits for Cisco Nexus Dashboard Fabric Controller software or Cisco Nexus/MDS switch hardware and software. When you try to achieve maximum scalability by scaling multiple features at the same time, results might differ from the values that are listed here.

Nexus Dashboard System Resources

The following table provides information about Server Resource Requirements to run NDFC on top of Nexus Dashboard. Refer to Nexus Dashboard Capacity Planning to determine the number of switches supported for each deployment.

Cisco Nexus Dashboard can be deployed using number of different form factors. NDFC can be deployed on the following form factors:

  • pND - Physical Nexus Dashboard

  • vND - Virtual Nexus Dashboard

  • rND - RHEL Nexus Dashboard

Table 1. Server Resource Requirements to run NDFC on top of Nexus Dashboard
Deployment Type Node Type CPUs Memory Storage (Throughput: 40-50 MB/s)
Fabric Discovery Virtual Node (vND) – app node

16 vCPUs

64 GB

 550 GB SSD

Physical Node (pND)

(PID: SE-NODE-G2)

2 x 10-core 2.2GHz Intel Xeon Silver CPU

256 GB of RAM

4 x 2.4 TB HDDs

400 GB SSD

1.2 TB NVME drive

Physical Node (pND)

(PID: ND-NODE-L4)

2.8GHz AMD CPU

256 GB of RAM

4 x 2.4 TB HDDs

960 GB SSD

1.6 TB NVME drive
Fabric Controller Virtual Node (vND) – app node 16 vCPUs 64 GB 550 GB SSD

Physical Node (pND)

(PID: SE-NODE-G2)

2 x 10-core 2.2GHz Intel Xeon Silver CPU

256 GB of RAM

4 x 2.4 TB HDDs

400 GB SSD

1.2 TB NVME drive

Physical Node (pND)

(PID: ND-NODE-L4)

2.8GHz AMD CPU

256 GB of RAM

4 x 2.4 TB HDDs

960 GB SSD

1.6 TB NVME drive
SAN Controller

Virtual Node (vND) – app node

(with SAN Insights)

16 vCPUs

(with physical reservation)

64 GB

(with physical reservation)

 550 GB SSD

App Node (rND)

(with SAN Insights)

16 vCPUs

(with physical reservation)

64 GB

(with physical reservation)

 550 GB SSD

Data Node (vND) – Data node

(with SAN Insights)

32 vCPUs

(with physical reservation)

128GB

(with physical reservation)

3 TB SSD

Data Node (rND)

(with SAN Insights)

32 vCPUs

(with physical reservation)

128 GB

(with physical reservation)

3 TB SSD

Physical Node (pND)

(PID: SE-NODE-G2)

2 x 10-core 2.2GHz Intel Xeon Silver CPU

256 GB of RAM

4 x 2.4 TB HDDs

400 GB SSD

1.2 TB NVME drive

Physical Node (pND)

(PID: ND-NODE-L4)

2.8GHz AMD CPU

256 GB of RAM

4 x 2.4 TB HDDs

960 GB SSD

1.6 TB NVME drive

Scale Limits for NDFC Fabric Discovery

Table 2. Scale Limits for Fabric Discovery Persona and Nexus Dashboard

Profile

Deployment Type

Verified Limit

Fabric Discovery

1-Node vND (app node)

100 switches

Fabric Discovery

3-Node vND (app node)

200 switches

Fabric Discovery

5-Node vND (app node)

1000 switches

Fabric Discovery

1-Node pND

100 switches

Fabric Discovery

3-Node pND

1000 switches

Scale Limits for NDFC Fabric Controller

Table 3. Scale Limits for Fabric Controller Persona and Nexus Dashboard

Profile

Deployment Type

Verified Limit

Fabric Controller

1-Node vND (app node)

50 switches

Fabric Controller

3-Node vND (app node)

100 switches

Fabric Controller

5-Node vND (app node)

400 switches for Easy Fabrics1

Fabric Controller

5-Node vND (app node)

1000 switches for External Fabrics2

Fabric Controller

1-Node pND

50 switches

Fabric Controller

3-Node pND

500 switches for Easy Fabrics1

Fabric Controller

3-Node pND

1000 switches for External Fabrics2

1 Easy Fabrics include Data Center VXLAN EVPN fabrics and BGP fabrics.

2External Fabrics include Flexible Network fabrics, Classic LAN fabrics, External Connectivity Network fabrics, and Multi-Site Interconnect Network fabrics. Both managed and monitored mode are supported.

Table 4. Scale Limits for Switches and Fabrics in Fabric Controller

Description

Verified Limit

Switches per fabric

200

Physical Interfaces per NDFC instance1

30000

1 Supported scale for 1-node vND is 2500 physical interfaces.

Table 5. Scale Limits For Provisioning New Data Center VXLAN EVPN Fabrics (also referred to as "Greenfield" Deployment)

Description

Verified Limit

Fabric Underlay Overlay

Switches per fabric

 200
Overlay Scale for VRFs and Networks1

500 VRFs, 2000 Layer-3 Networks

or

2500 Layer-2 Networks
VRF instances for external connectivity

500

IPAM Integrator application

150 networks with a total of 4K IP allocations on the Infoblox server

ToR and Leaf devices

A Data Center VXLAN EVPN fabric can manage both Layer-2 ToR switches and leaf switches. Maximum scale for this sort of fabric is 40 leaf switches and 320 ToR switches.

Endpoint Locator2
Endpoints 100000

VXLAN EVPN Multi-Site Domain

Sites

30

Virtual Machine Manager (VMM)3

Virtual Machines (VMs)

5500

VMware Center Servers

4

Kubernetes Visualizer application

Maximum of 160 namespaces with maximum of 1002 pods

1 Supported scale for 1-node vND is 250 VRFS and 1000 networks.

2 Supported scale for 1-node vND is 1 instance of endpoint locator with 10000 endpoints.

3 Supported scale for 1-node vND is 1 VMware Center Server and 1000 VMs.


Note

Refer to the following table if you are transitioning a command line interface (CLI) configured Cisco Nexus 9000 series switches based VXLAN EVPN fabric to NDFC.

Table 6. Scale Limits For Transitioning Existing Data Center VXLAN EVPN Fabric Management to NDFC (also referred to as "Brownfield Migration")

Description

Verified Limit

Fabric Underlay and Overlay
Switches per fabric 200
Physical Interfaces 11500
VRF instances 400
Overlay networks 1050
VRF instances for external connectivity 400
Endpoint Locator
Endpoints 50000
IPAM Integrator application 150 networks with a total of 4K IP allocations on the Infoblox server

Virtual Machine Manager (VMM)

Virtual Machines (VMs)

5500

VMware Center Servers

4

Kubernetes Visualizer application

Maximum of 160 namespaces with maximum of 1002 pods

Scale Limits for Cohosting NDFC and other Services

Table 7. Scale Limits for Cohosting Nexus Dashboard Insights and NDFC

Profile

Deployment Type

Verified Limit

Nexus Dashboard Insights and Nexus Dashboard Fabric Discovery

3-Node pND

50 switches, 1050 records/sec

Nexus Dashboard Insights and Nexus Dashboard Fabric Controller

3-Node pND

50 switches, 1050 records/sec

Scale Limits for IPFM Fabrics

Table 8. Scale Limits for Nexus Dashboard and IPFM Fabrics

Profile

Deployment Type

Verified Limit

Fabric Controller

1-Node vND

35 switches (2 spine switches and 33 leaf switches)

Fabric Controller

3-Node vND

120 switches (2 spine switches, 100 leaf switches, and 18 Tier-2 leaf switches)

Fabric Controller

1-Node pND

35 switches (2 spine switches and 33 leaf switches)

Fabric Controller

3-Node pND

120 switches (2 spine switches, 100 leaf switches, and 18 Tier-2 leaf switches)
Table 9. Scale Limits for IPFM Fabrics

Description

Verified Limit

NBM Active Mode Only

NBM Passive Mode Only

Mixed Mode

NBM Active VRF

NBM Passive VRF

Switches

120

32

32

32

Number of flows

32000

32000

32000

32000

Number of End Points (Discovered Hosts)

5000

1500

3500

1500

VRFs

16

16

16

16

Host Policy - Sender

8000

NA

8000

NA

Host Policy - Receiver

8000

NA

8000

NA

Host Policy - PIM (Remote)

512

NA

512

NA

Flow Policy

2500

NA

2500

NA

NBM ASM group-range

20

NA

20

NA

Host Alias

2500

NA

2500

NA

Flow Alias

2500

NA

2500

NA

NAT Flows

3000

3000

3000

3000

RTP Flow Monitoring

8000

8000

8000

8000

PTP Monitoring

120 switches

32 switches

32 switches

32 switches

Scale Limits for NDFC SAN Controller

Table 10. Scale Limits for SAN Zones

Description

Verified Limits

Zone sets

1000

Zone

16000
Table 11. Scale Limits for Nexus Dashboard and SAN Controller Persona

Profile

Deployment Type

Verified Limit

Without SAN Insights

With SAN Insights

SAN Controller

1-Node vND (app node)1

80 switches, 20K ports

40 switches, 10K ports, and 40K ITs

1-Node vND (data node)

80 switches, 20K ports

80 switches, 20K ports, and 1M ITLs/ITNs 2

1-Node pND (SE)

80 switches, 20K ports

80 switches, 20K ports, and 120K ITLs/ITNs

SAN Controller

3-Node vND (app node)

160 switches, 40K ports

80 switches, 20K ports, and 100K ITs

3-Node vND (data node)

160 switches, 40K ports

160 switches, 40K ports, and 240K ITLs/ITNs

3-Node pND

160 switches, 40K ports

160 switches, 40K ports, and 500K ITLs/ITNs

SAN Controller on Linux (rND)

(Install Profile: Default)

1-Node vND

80 switches, 20K ports

40 switches, 10K ports, and 40K ITs

SAN Controller on Linux (rND)

(Install Profile: Large)

1-Node vND

80 switches, 20K ports

80 switches, 20K ports, and 1M ITLs/ITNs2

SAN Controller on Linux (rND)

(Install Profile: Default)

3-Node vND

160 switches, 40K ports

80 switches, 20K ports, and 100K ITs

SAN Controller on Linux (rND)

(Install Profile: Large)

3-Node vND

160 switches, 40K ports

160 switches, 40K ports, and 240K ITLs/ITNs

1 App nodes have fewer features than data nodes. For example, the lun and fc-scsi.scsi_initiator_itl_flow features are not supported in the app ova, whereas those features are supported in the data ova. Therefore, you would have to install the data ova in order to use the lun or fc-scsi.scsi_initiator_itl_flow features.

2 1 million flows is the maximum number supported. If other features are enabled that consume resources, 1 million flows will not be stable in all situations. NDFC consumes more resources per flow when processing telemetry from a larger number of devices. Watch flow counts and node memory usage (1 minute averages above ~105GB starts to show instability).


Note

ITLs - Initiator-Target-LUNs

ITNs - Initiator-Target-Namespace ID

ITs - Initiator-Targets