Has anyone used VMWare for AFS item processing to consolidate servers? What pitfalls did you run up against?

I know we have several of our IT audit clients that have moved to virtual servers for some applications but not sure about AFS. From my perspective, anytime you can reduced the number of servers then the risk gets reduced. I would like to add to the question, does anyone know of any bugs and/or vulnerabilities with any virtual server software? Also, how are SAN being used in conjunction with these servers.

One limitation that we have found with Microsoft Virtual Server is that it is limited to a single processor per guest operating system. That is one of the reasons that we went with VMWare Virtual Infrastructure 3 and have been very happy with the performance and capabilities that it has given us. We have 23 production virtual servers running on 3 VMWare ESX servers. We have recently implemented a LeftHand Networks SAN and have moved 20 of our virtual servers onto the SAN. This allows us to use some of the advanced features of VMWare such as VMotion, VMWare HA, and VMWare DRS. Another advantage that the SAN gives us is the ability to replicate volumes to a remote SAN node to improve our DR plan.

Klint, what’s the recovery strategy if you lose your SAN? Do you duplicate the virtual environment at an alternate location? I agree in general that fewer servers to manage reduces risk; however, you have to consider any single points of failure you may create in the process, right? I can see scaling down the scope of the recovery, but would be very interested in performance issues and application availability.

Klint…didn’t realize you guys were so far into VMWare. Very good stuff. Jim..my bad…I should have qualified the statement of lower risk by saying, the risk can be mitigated even more by a replicated VM server at a remote location of some distance from the primary locations. You are RIGHT ON…

Jim, we have two means of recovery. One is that we use a product from Vizioncore called vRanger Pro that will take a snapshot of our VMDK’s, tar them up and store them in an alternate location. In addition to that, we have another SAN node in an alternate location and do asynchronous replication using the LeftHand product. Just the changed blocks are replicated on a daily basis, so it works fine over a T1 connection. We have also tried to eliminate any single point of failure by employing 3 ESX servers so that if any one physical box dies, we can run all the servers on the other two physical boxes. VMWare HA handles the transition of those servers from the failed machine to the other two servers. We have employed redundancy into our SAN environment as well using redundant switches, redundant UPS, redundant power circuits, ALB, etc. The SAN itself is run on 3 Dell PowerEdge 2950’s that each have a RAID 5 array, but also utilize something called Network Raid. Performance and application availability has not been an issue because we can allocate all the resources we need to a particular server. Each ESX server has 2 quad core processors and 16GB of RAM. These resources are shared amoung all the servers. Hope that answered most of your questions.

One more thing, we did physical to virtual conversions on many of our physical servers so we had some spare servers left over. We put them in our alternate location with esx loaded on them and they are pointed to the SAN so that they are ready to start up the servers in the event of a disaster.