Looking into virtualization!

bigredshark · 10 Sep 2009 at 11:19

Don't shell out for a license and hardware for a dedicated DC! Just run in on the DB server, unless you have loads and loads of users...

kefkef · 10 Sep 2009 at 13:07

Have you guys done any virtualising using NetApps as the storage? Just looking at a proof of concept at the moment to shift a few trial servers onto VM's (DC, DHCP and 1 App server). Wondered if the NetApps would be a good match for this type of exercise?

POC involves 4 servers for ESX, 2 for Virtual Centre connected to two NetApp storage devices which replicate.

iaind · 10 Sep 2009 at 13:11

kefkef said:
Have you guys done any virtualising using NetApps as the storage? Just looking at a proof of concept at the moment to shift a few trial servers onto VM's (DC, DHCP and 1 App server). Wondered if the NetApps would be a good match for this type of exercise?

POC involves 4 servers for ESX, 2 for Virtual Centre connected to two NetApp storage devices which replicate.

That sounds like an absolutely Bizarre setup - a SAN for 3 servers is a bit mad

Why would you want 2 virtualcenter servers for 4 hosts?

bigredshark · 10 Sep 2009 at 13:14

kefkef said:
Have you guys done any virtualising using NetApps as the storage? Just looking at a proof of concept at the moment to shift a few trial servers onto VM's (DC, DHCP and 1 App server). Wondered if the NetApps would be a good match for this type of exercise?

POC involves 4 servers for ESX, 2 for Virtual Centre connected to two NetApp storage devices which replicate.

Makes most sense if you're using their vmware integration, which is very good, otherwise it's very expensive basic storage. That said it works well even as just simple storage but there are faster, cheaper solutions out there...

kefkef · 10 Sep 2009 at 13:40

should have said, we already have the NetApps (2 x 56tb FAS2050 arrays replicated) but was wondering if i should be thinking about seperate storage if it is more suited. the POC is what it says, a proof of concept, if it is proved (which we are fairly certain it will be) then there will be over 200 servers to be assesed for virtualisation. And it would be 6 servers in total to be virtualised as part of the POC (2 x DC's, 2 x DHCP and 2 App servers)

bigredshark · 10 Sep 2009 at 13:45

Really does depend on whether you want the advanced features netapp have for use with vmware...and even if you don't then if you have the spare storage you may as well use it. The Nexsan Satabeast/SASbeast products are great for cheap, basic storage for vmware in my opinion but no point buying new stuff if you have netapps which will do the job.

iaind · 10 Sep 2009 at 14:04

As above, if you've already got the netapps they will work fine - especially well with NFS apparently.

Still dont see a need for 2 VC servers though

kefkef · 10 Sep 2009 at 14:09

I was recommended having 2 as if the POC works for the initial trial (trialling HA and Vmotion aswell as just the virtualisation) then the next POC will be to test out Virtual Site Recovery and it was there that the two VC's should provide the resilience required.

iaind · 10 Sep 2009 at 14:33

Ah, didnt realise SRM was involved

Surely if your VC is a VM and your SAN is replicated, it wouldnt be necessary?

kefkef · 10 Sep 2009 at 14:52

It was recommended that we have the VC on dedicated hardware. Probably why i'm asking the questions as it's all new to me and not sure if they are trying to pull the wool over my eyes! (My techies and the VMware guys that is!!

)

morfmedia · 12 Sep 2009 at 20:31

Whilst you can have virtualcenter as a VM, there are several reasons it's wise not to. I'm sure a google will turn the answers up.

LordSplodge · 12 Sep 2009 at 21:32

morfmedia said:
Whilst you can have virtualcenter as a VM, there are several reasons it's wise not to. I'm sure a google will turn the answers up.

I was advised by a VMware trainer it is good practice to run Virtual Centre as a VM to gain advantages of VM functionality such as HA, Fault Tolerance and so forth.

Having your VC on physical tin doesn't make any sense to me unless your VM environment has no redundancy (local VMFS and only one ESX server for example)

We run our VC as a VM and so far DRS (it is on full auto at the moment) has moved it a few times. Although DRS won't be set to full auto on the production environment.