Virtualization and single points of failure in small businesses

unnamedplayer

Hey guys,

I was just thinking about something. I had a small business network (around 50-60 users) and they had a few servers (2 DCs [one acting as a file server also] and an Exchange server). Long story short, these machines were on their last legs and they finally got some money to go ahead and upgrade the hardware. Now, virtualization sounds attractive; buy one new powerful piece of hardware instead of 3.

I was just thinking though, I wonder how well this strategy works in small businesses where there aren't a whole lot of servers to consolidate. I was thinking about situations like say a PSU failure. Now all the servers are down instead of one. And I know there are redundancy options for PSUs, but there are other scenarios like say a motherboard failure, or the resident "IT" guy spills his coffee on the machine

So, I guess I just wanted to get some feedback from some of the experts in this forum regarding how you generally approach such situations. Is it common to also buy another server as a backup? Now I've bought 2 servers instead of 3?

Anyway, I'm greatly interested in your all's thoughts.

SteveO86

Buy a very expensive piece of hardware or 2-3 mid-range pieces and create a cluster.

Best way to avoid that in my opinion.

Or get some type of 24x7x4 support on the single piece of hardware.

But 1 server running VM's is a single point of failure,

it_consultant

Yeah, its a single point of failure, however keeping VMDK based backups will mitigate this risk pretty well. Consider the last time you really had a bad hardware failure, they are not that common. I comfortably run entire networks on one ESX server but I did explain that if the thing busted, we would have to wait until a new piece of hardware to show up to recover them.

Don't go cheap on hardware or support and make sure you invest in a good ESX backup solution and you should be fine.

rsutton

I have a small business client running a couple servers on VMWare and have had two outages in the past year where all servers went down to a failure of the VMWare server. You need to have a failover solution unless your client does not care about down time.

undomiel

The expense of a cluster would be too prohibitive. He might as well just replace the servers then. The way we do it for a few of our clients is that their servers are consolidated onto one beefy server. The VMs are backed up nightly and if the server goes down we have a lower powered general use server back at the office that we bring on site and run the VMs from that. They might just be limping along on it but it gets them operational until we repair the physical server. Going with two servers will let you have a backup server for those times of emergency plus you can have both of them in production by splitting the VMs across them.