Tuesday, July 29, 2008

Virtualization Fun with AoE

I think I will put fun in the title but I know our team would think of other descriptive words to put in its place. It has definitely been quite a learning experience, I will give you that much. A little background on what we had in mind with this.

The time had come to start thinking about replacing our aging servers with some more youthful, more vibrant servers but instead of just swapping out one for one we really started to think about the benefits of virtualizing our environment would give us. So the plan was made and everything was ordered but putting those pieces together of is not as easy as expected.

Here is what we have hardware wise:

Dlink DGS-3100-24
Custom Untangle Server
Dell Poweredge 1950 III
Dell R805 Dual Quad Core Xeon, 16GB RAM
Dell R805 Dual Quad Core Xeon, 16GB RAM
Coraid VS21
Coraid SR1521 (Raid 10 w/ 1TB drives)

We have the SR1521 on a separate VLAN for storage, the VS21 is on the storage VLAN and also a production VLAN, the poweredge is the management server and is on a management VLAN as well as the production VLAN, the R805 are on all VLANs, management, storage, production and lab. When we are done this will server around 30 internal users and around the same amount of users externally. We are trying to virtualize all of our servers and a few desktops for right now.

For the software side of the equation we looked at XenServer and possibly Vmware but we ended up with Virtual Iron because of the promise we saw in the solution. As of right now the storage is not fully supported with their software but we have been working with them to try and get things working. In the meantime we didnt want to sit around so we tried to find another solution that would work as well. We looked up OpenQRM and again things are looking very promising, AoE is supported but the Coraid boxes themselves are not specifically working as you would think but we are trying to work with them so that hopefully in the future it will be with now problems.

So right now we are not really doing too well with a solution that will support our setup but there is hope. The next step after this would be just using straight Xen for the solution, we'll see...

Once this is working then we'll start looking for more ways to increase redundancy and availability