Enabling TOE in a VMware VM for the effect of improved networking performance

Deathmage

Hey guys,

I'm curious if anyone here has ever enabled this command:

"netsh int tcp set global chimney = enabled"

It essentially enables TOE, RSS, Checksum/Segmentation offload.

iBrokeIT

I would be interested in hearing the use case for enabling these settings.

joelsfood

TOE can be tricky on VMs. at times it will make performance worse, not better

https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009517

Deathmage

Well I found this nice little book at Barnes and Noble hidden in a corner on Advanced Windows Setting in Windows Server 2008 and 2012. Since I'm always tinkering at home on how to tweak the piss out of Windows Server, I figured what the heck...

I'm right now transcribing them to my blog as I test them out, here is a graph I've made so far of the settings and their benefits:



[TH]Server role
[/TH]
[TH]Checksum offload
[/TH]
[TH]Segmentation offload
[/TH]
[TH]TCP offload engine (TOE)
[/TH]
[TH]Receive-side scaling (RSS)
[/TH]


File server

X

X

X

X



Web server

X

X

X

X



Mail server (short-lived connections)

X





X



Database server

X

X

X

X



FTP server

X

X

X





Media server

X



X

X




If your hardware supports TOE, RSS, Checksum/Segmentation offload, then you must enable that option in the operating system to benefit from the hardware’s capability. You can enable TOE by running the following command:


"netsh int tcp set global chimney = enabled"

Deathmage

joelsfood wrote: »

TOE can be tricky on VMs. at times it will make performance worse, not better

https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009517

Well that is interesting indeed. Will need to monitor the server that this was applied too. I just did it to the file server, will watch the graphs and compare them to the baselines from before the change. If it effects it, I can simply turn it off. Was curious of it's benefits maybe with the SQL server.

Always looking to increase performance, but it's not a perfect science.

iBrokeIT

Just because you can toggle a setting doesn't necessarily mean you should unless there is a strong need or use case for doing so. By doing a bunch of stuff like this you are increasing the complexity of your environment and you will have to spend time ruling this stuff out when something goes wrong all for a marginal gain in performance at best.

Deathmage

iBrokeIT wrote: »

Just because you can toggle a setting doesn't necessarily mean you should unless there is a strong need or use case for doing so. By doing a bunch of stuff like this you are increasing the complexity of your environment and you will have to spend time ruling this stuff out when something goes wrong all for a marginal gain in performance at best.

Indeed, I tested it at home in the home-lab and I saw differences. Testing it in Production now. Was curious though if others used it.

iBrokeIT

How much of a difference did you see? What were your numbers before and after? How did you measure the performance difference? (genuinely interested)

Deathmage

I'm home now, but when I get back to work, I can post the baseline screenshots.