What is the Maximum MTU size for Jumbo Frames??

Recently on Twitter I had this discussion with Michael Webster @vcdxnz001, Patrick Carmichael  @VMCarmichael and Josh Odgers @Josh_odgers  about the Jumbo Frames topic which was just came out of the TCP/UDP joke.

And the answer is Yes, this one is easy "Its 9000".

Most likely people will give this answer when asked the question on Jumbo Frames.

Now  some might go little deeper and keeping in mind the over head and other "surcharge" by VMware vmkernel they may say 8972 to be precise. 

That is also a valid answer but not close to the information received from Intel.

Other vendors might have different opinions so worth to investigate/research with them ;-)

Now read this extract from the same article:


"The maximum MTU setting for jumbo frames is 16110. This value coincides with the maximum Jumbo Frames size of 16128."


Broadcom and other vendors might have different supportability on the MTU size used with Jumbo Frames.

vSphere supports only 9000. People are using mostly 1500 MTU as there are blogs out there which says there is no performance gain in using MTU of 9000 or if there is, its very marginal gain.

So hopefully demystifying the myth (in terms of Eric Sloof @esloof) here about maximum size of MTU available.

Care and share !!

Cheers.

Disable/Enable ABC for NFS Troubleshooting on NetApp

I was reading some issue with NetApp NFS troubleshooting and came across the KB from NetApp to disable flow Control on the Array.

This new Flow control setting is listed under a RFC 3465 which is described as ABC (Appropriate Byte Counting (ABC)).

Now this particular settings takes care of TCP Congestion Control with Appropriate Byte Counting (ABC) settings on the array.



As per this document  Storage Best Practice Document (TR-3749.PDF) page 25 Flow control means:


 FLOW CONTROL
Flow control is a low-level process for managing the rate of data transmission between two nodes to prevent a fast sender from overrunning a slow receiver. Flow control can be configured on ESX/ESXi servers, FAS storage arrays, and network switches. For modern network equipment, especially 10GbE equipment, NetApp recommends turning off flow control and allowing congestion management to be performed higher in the network stack. For older equipment, typically GbE with smaller buffers and weaker buffer management, NetApp recommends configuring the endpoints, ESX servers, and NetApp arrays with the flow control set to "send."

And further reading on the Congestion Management with Flow control - TR-3802.pdf page 23 , it mentioned other technologies to use such TCP windowing, increased switch buffering, and end-to-end QoS which may reduce the need for simple flow control throughout the network. Now I think along with that you can add the setting for ABC as per the RFC mentioned above. You can read the RFC to get more idea on how exactly it works on the TCP stack.

Fro ONTAP 8.1 Netapp has  released the guide with the following command:


ip.tcp.abc.enable

(Enables/disables the use of Appropriate Byte Counting in TCP Congestion Control following RFC
3465. Valid values for this option are on or off. The default value for this option is on.)


Now if you need further information on how to set this up on the affected Array and I would suggest to get in touch with the vendor.

Now this may or may not improve the overall situation but its worth giving a try.


Please share if you can.