Subscribe

Implementing extensible NetApp and VmWare ESX infrastructure

  Hello.

I have a cluster pair of NetApp FAS 3020c and a pair of Dell PowerEdge 1950.

My task is to implement extensible architecture which will be easy to extend

(add new servers) and robust.

Each filer has one ds14mk2 (423 GB SATA) ds14mk4FC (123 GB FC) with 2

spares.

Filers and servers are attached to gigabit cisco C2960G switch. Each filer

is attached with two cables into the switch.  On each filer vif is configured with

lacp.  Switch is also configured and runs EtherChannels.  Each server has

6 ethernet ports, only 4 of them were connected.  I use one port for ESX

Service Console, two ports are used to provide network connectivity for

virtual machines, and last cable is intended to handle VMotion and NFS

traffic.

There were no free ethernet ports on cisco switch, when servers were connected

to switch with 4 cables.  So i attached servers to filers directly, by doing this i

managed to free two ports on switch.  I made this to isolate NFS and VMotion

from switch. But i also have 8 additional licenses for ESX, so 4 additional servers may be

added to ESX cluster.  So i do not know how to setup my environment for additional

servers.

If i leave servers directly attached to filers i will not be able to directly connect

new servers directly because there are no free ethernet ports on each filer.  I

cannot be sure that there will be free ethernet ports on switch when new servers

will be avaliable.

If i attach servers back to switch my NFS and VMotion traffic will be flowing through

etherchannel with virtual machines network traffic.  This i think may significantly impact

network performance.

Re: Implementing extensible NetApp and VmWare ESX infrastructure

Phew....that's a ton of questions and topics to be honest. I think I'd start with looking at this guide on recommended NIC layouts for ESX....should help a bit at least (has options for various #'s of NICs on the ESX host).

http://www.networkworld.com/community/taxonomy/term/17475

From there you want to think through the single points of failure from the NetApp side and implement VIFs to handle them.

This is frankly a big topic and I might almost recommend getting someone from NetApp and/or VMware in there for a brief paid engagement to help with the architecture.