Skip to main content
NetApp Knowledge Base

Impact of changing MTU size  on cluster interfaces from 1500 to 9000

Views:
1,052
Visibility:
Public
Votes:
0
Category:
fas-systems
Specialty:
network
Last Updated:

Applies to

  • NetApp purchased Cluster and/or MetroCluster switches
  • Ethernet switches

Answer

  • Impact to interfaces when changing the MTU from 1500 to 9000 after upgrading to N5k switches

    • When configuring clients for jumbo frames, you should verify certain configurations, such as the TCP window size of the client, the MTU size of the client, storage system, and any intermediate subnet.

  • Will configuring the N5k switch to MTU 9000 cause and outage?

    • The guidelines for configuring clients for jumbo frames are as follows:

  1. The jumbo frames on the client and on your storage system should be configured.
  2. To find out how to configure jumbo frames on your client, check the network adapter documentation for your client.
  3. The TCP window size on the client should be enlarged depending on the MTU size.
  4. The minimum value for the client's window size should be two times the MTU size, minus 40, and the maximum value can be the highest value your system allows. Typically, the maximum value you can set for your client's TCP window is 65,535. If your storage system is configured to support jumbo frames and the client is not, the communication between the storage system and the client occurs at the client's frame size.
  5. The storage system and the UDP clients should be configured to have the same MTU size.
  6. UDP systems do not negotiate the MTU size. If your storage system and clients do not have the same MTU size, the storage system might send packets that the clients may not be able to receive.
  7. If the storage system and the client are on different subnets, then the MTU size of any intermediate subnets should be checked.
  8. If the storage system and the client (both configured to use jumbo frames) are on different subnets and an intermediate subnet does not support jumbo frames, the intermediate router fragments the IP packets as a result of which the advantages of using jumbo frames are lost.
  •  Are the SAN interfaces of iSCSI, NFS and Cluster non-routable?

    • Before you enable jumbo frames on your storage system, jumbo frames must be enabled for the switch ports, client interfaces, and intermediate routers on the network. If your storage system and the client are on different subnets, the next-hop router must be configured for jumbo frames.

  • What is the steps to configure supporting Jumbo Frame? We have Cisco UCS and VMware vCenter infrastructure connecting to CDOT SAN.

    • Jumbo frames are larger than standard frames and require fewer frames.
    • Therefore, you can reduce the CPU processing overhead by using jumbo frames with your network interfaces.
    • Particularly, by using jumbo frames with a Gigabit or 10 Gigabit Ethernet infrastructure, you can significantly improve performance, depending on the network traffic.
  • Jumbo frames are packets that are longer than the standard Ethernet (IEEE 802.3) frame size of 1,518 bytes.
    • The frame size definition for jumbo frames is vendor-specific because jumbo frames are not part of the IEEE standard.
    • The most commonly used jumbo frame size is 9,018 bytes.
  • Jumbo frames can be used for all Gigabit and 10 Gigabit Ethernet interfaces that are supported on your storage system.
    • The interfaces must be operating at or above 1,000 Mbps.
  • You can set up jumbo frames on your storage system in the following two ways:
    1. During initial setup, the setup command prompts you to configure jumbo frames if you have an interface that supports jumbo frames on your storage system.
    2. If your system is already running, you can enable jumbo frames by setting the MTU size on an interface.

Additional Information

N/A

 

 

******************************************************* *******************************************************