Just a quick question on good practices: I have an existing Test/Staging environment that consists of 2 dell servers. Each server has 4 cores and one is rated at 3.6 Ghtz with 8 gigs of memory and the other is at 2.0 Ghtz with 8 gigs of memory. Between the two servers they have about 20 active VMservers. At this time because of limited disc space and resources I may need to add another server to my farm. Is the general consensus that I should go with a state of the art Dell 2950 with 2 quad core processors and 32 gigs to replace my existing Test/Staging environment? This would match two new production servers that are being introduced and allow me then benefit by adding it to the farm and using any spare resources for production or HA needs.
This is a build it wide, build it high argument. You can either spend your money on RAM kits as anything above 16GB is screamingly expensive or buy the extra CPU licenses from VMware. Those seem to be around the same cost of the RAM kits after you add in maintenance. IMHO, the bottom line is whether your test/dev environment SLA/SLOs have enough leeway to permit unplanned downtime. A single host leaves you hanging a bit... Also remember that test/dev is production to someone.
Sorry, didn't catch that part about leaving the other hosts in the farm. This should be OK. If you plan on using VMotion, then all of the hosts must have the same CPU family. Speed isn't important...
I'm new and taking over this VMfarm. What do you mean by the fact that it will leave me hanging? I'm just looking for additional clarification, do you believe it a better practice to have different boxes for Test/Staging? Just to clarify - I currently have 3 in production and 2 for Test/Staging. I'm replacing a production server and adding an addition server. It occured that it might be worth it to replace the Test/Staging with a new unit that would function for both and then could add any additional resources to the production or assist with a HA situation. In the end I would end up with two seperate clusters , 3 Dell 2950's and 2 Dell 2900's.
Thanks for the info.
I thought you were rolling up the two existing hosts into the new Dell. Once complete, you'd decomm the old ESX hosts. After re-reading your post, it seems that you're going to leave one or both of the existing hosts in the farm/cluster. If this is the case, you should be OK.
I guess the bottom line is that dev/test environments; while having way different RPO/RTOs than PRD, end up being production systems to developers. If you have a single host, then you have to weigh the financial impact of downtime to your dev/test users.
I think I was more or less looking for agreement that it would be an equivalent server replacement of the older servers with the newer system I was wondering if anyone has had any experience with this scenario before. Did they consolidate and find out that the performance was worst even thought it is the same available processors and more memory.
Personally, most of my workloads fit well within the dual core envelope, so quad core was gravy. Memory on the other hand was constrained with both. Disk and network I/O have never been a factor for me.
One other thing, you can segregate your environments. It's a matter of data sensitivity and getting the most out of your investment. Some would argue to to mix the environments to get the max out of your investment. The rationale is that dev/test platforms sit idle for long stretches of time. Others would never consider this due to proprietary and sensitive data being on the same hosts and disks as dev/test. In some cases such as SOX compliance, you'd have to treat the host with many extra compliance controls since SOX systems are on on the same host as dev/test.