Using Nimble for VDI on a Server That Has Only Two 10GB NICs

[jlehman123]

Anyone setup Nimble to an IBM HS22 7870AC1 blade server w/ a two port 10GB NIC (or similar setup?) I ended up using the two NICs for both iSCSI & normal data & we're currently testing. Does anyone see any red flags by doing it this way? We run vdi only, off of these HS22 blades. I didnt want to split the two NICs, say one for data & one for iSCSI (best practice;) because we would loose redundancy. Thanks. Jason

[etang40]

Hi Jason,

Wen Yu, Tech Marketing guru at Nimble covered this very topic of having only 2 x 10Gbps NICs for both iSCSI and other network traffic on his blog.

http://www.supersonicdog.com/2013/04/24/lacp/

You'll find the topic covered around halfway through the post.

Hope this helps

-Eddie

[jlehman123]

Thanks Eddie.

Based off of that article, it sounds like we're already doing most of what is suggested. I just have to confirm, w/ our network admin. I'd still love to see an official Nimble/Vmware best practice guide, when only two 10GB NICs are available in a server. I appreciate your time in pointing me to this article.

This did help.

Jason

[etang40]

Wen Yu, our tech marketing guru on VMware is working on a new revision of the BPG with the 2 x 10GbpsE configuration for guest and storage traffic as you recommended.  Stay tuned!

-Eddie

[jlehman123]

That's great news, can't wait to see his new revision & compare it to what I've done.

As an update for my tests...

I was successfully able to put 56 virtual machines running Windows XP on one server that has only two 10GB NICs. This server will use those two NICs for iSCSI traffic (Nimble) as well as normal data traffic. This is not a best practice from Nimble, but I've been told by several Nimble employees, that Nimble will still support us. Just beware that if there are bandwidth issues; this will be a major red flag to them.

I had to stop at 56 VMs because that's where the RAM on the server got to 93% utilized.

We monitored the network bandwidth w/ Paessler PRTG (sorry not sure how to monitor the network I/O at this time.)

At no time what so ever, did the two 10GB NICS go anywhere near their bandwidth limit of 10GB.

During some of the highest processes, moving the VM's from our current storage to the Nimble arrary; I saw spikes in the 600 - 800 MB range.

Not even enough to max a 1GB connection, though I wouldn't recommend trying this w/ a 1GB connection(s).

Jason

[wen35 · Accepted solution]

Hi Jason, you could use esxtop with "n" option to monitor network traffic (I had an older post that has more details on this subject: http://www.supersonicdog.com/2012/09/05/esxtop5min/)

Specific to network monitoring, you could look at MbTX/s (transmit) & MbRX/s for each vmnic; for further breakdown of vm vs. iSCSI traffic, you could look at the corresponding throughput for VM portgroup & vmk1 & vmk2 portgroups.  Hope this helps.

[jlehman123]

Thanks Wen,

That link/article is great!

This will help me with network monitoring.

Jason