VM server to Core throughput

[groque]

Hi all,

Just wondering how other admins trunk VM servers into your core.

I am going to be using 2x 5400yl's would 1Gb to one 5400yl's and 1Gb to another work or should I put 2 per each switch.

I have 60 users in my office with 15 virtual servers. What would you guys recommend.

Thanks

[Mohammed Faiz]

Hi,

We're trunking in a pair of blade chassis that run in excess of 150 VMs. The blade chassis are using the HP virtual connect modules and at the moment we provide a pair of 8Gb trunk per chassis (plus smaller trunks for the service console traffic).
I would definitely start with 2Gb trunks and put some monitoring in place to see how heavily utilised they are.

[groque]

Thanks for your reply another question is how would you monitor your trunk ports would you use snmp traps or sflow monitoring to ensure that they provide enough bandwidth for end users to the vm servers

[Mohammed Faiz]

At the moment I use SNMP to poll in interfaces and log traffic historically. I'll also be looking at sflow now that we have PCM running.
It depends on how you want to monitor the trunks. If you want good real-time monitoring with the ability to determine source and end-point IPs for traffic flow then you want sflow.
If you want historical logging (e.g. looking back at the last 3 month's bandwidth usage) the SNMP based applications tend to be better at this.
Or you could have both :)

[groque]

Thanks for the response.

Cheers!