Project Planning $ Suggestions

[chi-ltd]

We are planning on increasing our VM host count and decommissioning our physical boxes..

Budget is a major factor, currently c15k for the servers and storage.

Plan A (my option):
Buy 2x more hosts.
32gb ram
2x12core cpu's
essentials basic
as much local storage as possible - at least 3tb.
buy a NAS 4-5tb
migrate hosts from old server to new
use 2 hosts @ live site and the 3rd (currently @ live & decommissioned upgraded @ DR site)
use veeam backup for live hosts and store to nas
use veeam replication for DR and re-ip and VLAN them
both live servers this will run approx 4-5 hosts each and split the load of crital servers. they include exchange, sql, rds, wds, oracle, dc's, ad, file/print.
we also have a tape drive which we plan to continue to use for the new physical veeam server.
no nas at DR just local storage

Plan B (my IT partners option):
Buy 2x more hosts.
32gb ram
2x12core cpu's
vsphere essential plus or standard
buy an entry level SAN (HP or DEll)
no nas

I know for HA the SAN is the route to go but from a budgeting exercise and automated HA availabilities plan B is best.

Suggestions?

[incognito]

check out the vmware virtual san appliance... I've had tremendous success with this.

pm me if you want to know more about it.

[J1mbo]

but at what cost. It's not the question I realise, but I think with 24 cores and just 32GB RAM the systems are quite unbalanced. Also Essentials IIRC is 6 sockets of <= 6 cores per socket?

Note that with Veeam only the SOURCE hosts are licensed, so you could have (say) 3x hosts at the primary site, with VMware Essentials (or Plus - and I would suggest exercising downgrade rights and running 4.1 to avoid vRAM limits) and a second site with Essentials only (1 to 3 hosts) then buy Veeam Essentials (plus) too.

NAS (NFS) or SAN (iSCSI) is unimportant; the availability features of the solution (redundant PSUs, fans, controllers preferably) are what matters in that they should be aligned to your needs. Backups and replicas are great security BUT, be honest would you ever want to use them? Yes you can recover, but nothing is in isolation.., so if you needed to restore an exchange database server (say) what about the way Active Directory has moved since? At the simplest point, what if the computer account password was renewed between backup and restore points?

In other words, assume the backup either won't work or will create problems. Much better to buy the availability you need in the storage solution.

Just my 2p! HTH!

[chi-ltd]

And heavily rely on the SAN not failing?
The veeam backups would surely provide a reliable way of a) restoring to a backup server (not HA) and b) use veeam to replicate the guests to DR & re-ip??

[J1mbo]

For replication, yes of course. I'd suggest separating servers onto their own VLAN (if not already done so) hence the re-IP bit can be alleviated in a complete fail-over scenario; instead the entire network is moved (create new router interface as part of the failover then presumably it will publish via whatever routing protocol you're using between sites). That could be further streamlined by using a LAN extension (bridging) service for that VLAN at both sites, if bandwidth permits, and a three-way routing failover instance. But anyway, that's off-topic.

For backups, yes Veeam will restore. The point is that the entire domain cannot be consistent since all the backups are at different times. There is always a chance Windows machine accounts may have been renewed between backup of member server and DC, for example. If that is the case between two domain controllers, it can be particularly fiddly to resolve. Flows between servers, for example an application and it's database backend may similarly be inconsistent. Indeed image level backup is specifically not supported for certain software such as Exchange 2010 database servers.

The point is storage can be reliable with redundant PSUs/fans/controllers/hot-spares. Storage systems like EqualLogic PS series for example if you want a single box, although personally I'm not keen on iSCSI, but that too is an aside. The point is there is no single point of failure in those boxes, so rely on the storage - yes, absolutely.