Seeking Advice: DataCenter Consolidation w/VMWare
darkerosxx
Banned Posts: 1,343
Looking for advice about taking on consolidating about 150 servers and around 20TB storage in a datacenter.
I think I'm finally going to give in and take a real look at consolidating our datacenter with VMWare. I've always wanted to use open source, but I've always had issues that turned me away, so I'm giving in to VMWare because I have a training course I can go take soon in anything I want, so I figured VCP4 would be a good choice.
Initially, the business case is going to have to support saving the company money, although productivity increase is a close second. So what I'm looking for is a way to be able to show concrete evidence as to how consolidating will save the company money, either through power savings, reduced hardware costs in the long run (we have a lot of old servers we could replace with just a few servers instead of replacing them all in the near future), etc..
I'm thinking that replacing 150 servers with about 10-12 servers for virtualization will do the trick, leaving the extra 2 (of 12) available for high resource usage applications. I believe most of our storage needs to be replaced entirely... we have some old storage systems that *could* be used, but I don't know if I'd trust them in a high usage system like this.
Is there any literature or advice anyone can offer for a plan to go about doing this?
Should I buy Platespin Recon and use that for the assessment? Will that give me an idea of power savings?
I think I'm finally going to give in and take a real look at consolidating our datacenter with VMWare. I've always wanted to use open source, but I've always had issues that turned me away, so I'm giving in to VMWare because I have a training course I can go take soon in anything I want, so I figured VCP4 would be a good choice.
Initially, the business case is going to have to support saving the company money, although productivity increase is a close second. So what I'm looking for is a way to be able to show concrete evidence as to how consolidating will save the company money, either through power savings, reduced hardware costs in the long run (we have a lot of old servers we could replace with just a few servers instead of replacing them all in the near future), etc..
I'm thinking that replacing 150 servers with about 10-12 servers for virtualization will do the trick, leaving the extra 2 (of 12) available for high resource usage applications. I believe most of our storage needs to be replaced entirely... we have some old storage systems that *could* be used, but I don't know if I'd trust them in a high usage system like this.
Is there any literature or advice anyone can offer for a plan to go about doing this?
Should I buy Platespin Recon and use that for the assessment? Will that give me an idea of power savings?
Comments
-
chrisone
Member Posts: 2,278 ■■■■■■■■■□
At the company i am working for in our data center we run many servers on VMWARE. I am not a server guy nor apart of our server infrastructure team, but i know they run a lot of servers with VMWARE and it seems to be a life saver for the company. With VMWARE not only is it a management improvement, but there is not that many moving parts as far as server components that can break down and cause services outage. However if your vmware server goes down that whole farm will be out. Which is why you have backup VMWARE servers that can take the load off 2 servers if they go down. I am speaking based off logic here lol
Another good benefit is that you save your company power and cooling bills! This is a huge and i mean HUGE! bonus/points/the man/kobe/jordan look for you in the company. Whenever you can save the company money it will benefit you for a big payday and raise!
Certs: CISSP, EnCE, OSCP, CRTP, eCTHPv2, eCPPT, eCIR, LFCS, CEH, SPLK-1002, SC-200, SC-300, AZ-900, AZ-500, VHL:Advanced+
2023 Cert Goals: SC-100, eCPTX -
slinuxuzer
Member Posts: 665 ■■■■□□□□□□
10-12 sounds really high, the first thing you might want to do is to size the amounts of memory currently being utilized and find a server model you want to work off of. We are currently using DL380 G6's with 48 GB of memory no where near maxed out, but you should have no trouble getting all these on six ESX servers or less.
Also, note about memory, if you seperate your Vm's by OS and put all windows boxes etc on the same host running the same patch level, the your machine should only consume the amount of memory required to run the OS one time and then + the amount of additional memory required for each machines app's and differences.
There are some Vmware best practices that say x amount of Vm's per server, but I know people that have run up to 50 Vm's on one host no issues.
I would definatley use your class for Vmware, you can't obtain the VCP cert without the class, so make sure you get the cert.
It would also be best to tinker around with Vmware in a lab before the class, that way you have a foundation to work from during the class, Vmware has real nice EVAL's that are fully functional for six months.
Check out Mastering Vmware Vsphere 4 book.
When trying to sell this don't forget the decreased cost of server maintenance contracts, these add up quickly.
If you really want to get detailed in sizing and specing the project, check out Powerrecon made by Platespin or formally platespin, I think novell bought them.
This will install a client piece on each one of your servers and monitor them for a while and take a baseline, you input all the spec's of the server you plan to run ESX on and it will tell you how many servers to buy, and what Vm's to group toghether on which host's for optimization reasons.
It also actually has an option to input your cost per kilowatt hour and it has reports you can run to show the cost savings, with 150 servers this may be worth your time, with my environement we have 50 so its not as worthwhile, also remember what ever you power you have to cool, so there will be some extra electricity saved there.