Create a QoS (quality of service) profile for your tier-1 gateways to define limits on the traffic rates. You can specify the permitted information rate and the burst size to set the limitations. Any traffic that does not conform to the QoS policy, is dropped. QoS profiles can be set for both ingress and egress traffic, for all traffic types (unicast, BUM, IPv4/IPv6). You can choose to create a different profile for each tier-1 gateway.

Note:
  • Gateway QoS profile is supported only on tier-1 gateways.
  • QoS policies on tier-1 gateways apply only to north-south traffic and not to tier-1 gateway overlay segments or service interfaces.
  • The tier-1 gateways must be in active-standby mode with an NSX Edge cluster.
  • QoS profile is not supported on tier-1 gateways that are configured for distributed routing only.

Procedure

  1. With admin privileges, log in to NSX Manager.
  2. Select Networking > Networking Profiles.
  3. Click the Gateway QoS tab.
  4. Click Add Gateway QoS Profiles.
  5. Enter a name for the profile.
  6. Enter the commited bandwidth limit that you want to set for the traffic.
  7. Enter the burst size. Use the following guidelines for burst size.
    • B is the burst size in bytes.
    • R is the committed rate (or bandwidth) in Mbps.
    • I is the time interval in milliseconds, to refill or withdraw tokens(in bytes) from the token bucket. Use the get dataplane command from the NSX Edge CLI to retrieve the time interval, Qos_wakeup_interval_ms. The default value for Qos_wakeup_interval_ms is 50ms. However, this value is automatically adjusted by the dataplane based on the QoS configuration.
    The constraints for burst size are:
    • B >= R * 1000,000 * I / 1000 / 8 because burst size is the maximum amount of tokens that can be refilled in each interval.
    • B >= R * 1000,000 * 1 / 1000 / 8 because the minimum value for I is 1 ms, taking into account dataplane CPU usage among other constraints.
    • B >= MTU of SR port because at least the MTU-size amount of tokens need to be present in the token bucket for an MTU-size packet to pass rate-limiting check.
    Since the burst size needs to satisfy all three constraints, the configured value of burst size would be: 
    Max (R * 1000,000 * I / 1000 / 8, R * 1000,000 * 1 / 1000 / 8, MTU)
    For example, if R = 100 Mbps, I = 50 ms, and MTU = 1500, then 
    B >= max (100 * 1000,000 * 50 / 1000/ 8, 100 * 1000,000 * 1 / 1000/ 8, 1500) = 625000 in bytes
  8. Click Save.