I'm not sure what they expect to happen. We have been running our dev/test systems in a virtual cluster for about 18 months and our operational systems for just over a year. If anything, the virtual systems benchmark slightly faster than the physical systems (depending on the applications, this ranged from 5 to 25% faster on the virtual server compared to a physical server running on the same hardware as the virtual system).
Our dev/test cluster is comprised of 4 DL380's, fairly hefty with 32GB of RAM. We utilize all of about 2GB of drive space on each server, just for VMware ESX Server 4.0 OS. All of the virtual images are on an enterprise storage solution connected through fibre channel. We're running 23 virtual systems on four physical boxes, including a development Remedy app and web server, a test Remedy app and web server, an MSSQL server utilized for about a half dozen different application in addition to the Remedy dev/test applications. All of the physical boxes are connected to the various network segments, allowing the virtual systems to run on any physical host. Even with this load, the virtual db and Remedy servers benchmark faster than they did on the physical systems. From what I've gathered in digging through everything, the reason for this is that the virtual server only need to load a handful of drivers rather than the hundreds that are loaded for the physical server. The hyper visor virtualizes all of the hardware and this frees up the guest OS to run only what is required to function rather than having to deal with the myriad various pieces of the hardware. Use of the virtual cluster has a lot of other advantages besides the cost savings for hardware. All of our server have high availability, without any of the complexities and foibles requires for clustering. Fault tolerance and resource balancing happen automatically. The virtual manager automatically restarts failed servers, moves servers between the physical hosts as load shifts during the course of the day, automatically generates snapshot images on a scheduled basis than can be restored in a matter of minutes, etc. These benefits are what drove us to virtualize our operational systems, the faster benchmarks were just a nice side effect. After running with virtualized and non-virtualized servers, I don't see any benefit to running a physical system unless you have to for some reason specific to the software you are running. Even then, I would probably try it in a virtual environment just to make sure the information that I had on why it couldn't run there is accurate. (Case in point, I was told repetitively that I could not run Win 2k8 in a VMware environment, after I had been running a pair of boxes for over six months.) Thanks, Roy //SIGNED// ROY ASHCRAFT, Contractor, 2 SOS/SYOE Remedy ARS Support, SAIC (402) 294-8225, DSN 271-8225 roy.ashcraft....@offutt.af.mil -----Original Message----- From: Action Request System discussion list(ARSList) [mailto:arsl...@arslist.org] On Behalf Of Drew Shuller Sent: Friday, July 30, 2010 11:59 AM To: arslist@ARSLIST.ORG Subject: MSSQL Virtualization? ---------------------- Information from the mail header ----------------------- Sender: "Action Request System discussion list(ARSList)" _______________________________________________________________________________ UNSUBSCRIBE or access ARSlist Archives at www.arslist.org attend wwrug10 www.wwrug.com ARSlist: "Where the Answers Are"
