Where I work, I built our VM Infrastructure literally from the ground up. As a part of another project, we had purchased an EMC SAN that had a ton of unused storage, even after factoring long term growth. I had been dabbling in VMware ‘on the side’ before then, and put all of that extra storage to good use. We had about 50 physical servers, in most cases each were existing just for handling a single application. I purchased 4 physical boxes, loaded to the max with processing power and RAM, building a ESX Cluster(three for production + one for fault tolerance).
Where I could, I ran a ‘P2V’ conversion on the existing servers, and was able to bring nearly all of them into the ESX Cluster. In nearly every case, the migrations went so well, there was NO downtime. We’re talking only a single dropped packet, while disabling the NIC on the physical server I was migrating from while connecting the NIC on the newly virtualized server. After I was done, my director walked into our datacenter, and it was nearly quiet. He saw all of the servers that had been shut down and panicked, thinking there was a power failure. I had to explain to him that our once sprawling datacenter was now contained in a single cabinet, only using 4Us. Not counting the SAN, which was in another cabinet.
The whole thing ran for about a year before he and my co-workers finally accepted and trusted it.
We’re a Windows shop, but have most of our stuff virtualized on ESX hosts right now. I’m still fighting a battle to wean people off the GUI, and use server core.
Good work! Didn’t they ever look at the electric savings?
You should do a report on the electrical and HVAC savings over the year to justify an increase in funding. A raise or more equipment is justifiable. Stats for a server’s electrical and cooling requirements are provided by most manufacturers websites. We have to include and tally up these figures before doing any install or upgrade as our server facilities’ space, power and HVAC are managed and budgeted resources.
If you have >50% available capacity in your SAN the first thing I’d do would be to put in a request for an additional chassis for VMware High Availabilty (HA). If the current chassis or even the interconnect switch fails the datacenter is SOL. HA is a parallel chassis and interconnect which VMware manages as an active-passive chassis array. Besides failover it’s also helpful for maintenance and upgrading of equipment without downtime.
It’s pretty good you had the freedom to do your own upgrade without briefing mgmt but it’s also sad they were ignorant of the advantages for a year. These advances save big $$ and you deserve credit for implementing them.