Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

I would like to get comments on a proposed configuration for a new Hyper-V server. The hardware is existing that will be re-deployed.

Hardware: Dell PowerEdge 2900

Processors: 2x Xenon 5160 3 GHz dual-core

Memory: 16 GB

Storage: 6x 68GB 15k SAS HDD (HDD0-HDD5) 4x 136 GB 15k SAS HDD (HDD6-HDD9) 2x 64 GB SSS (SSD0-SSD1)

OS: Server 2008 R2 Enterprise

VMs: DC1, EXCHANGE1, SQL1, RD1 (Remote Desktop)

Physical OS: on 68 GB RAID 1 volume consisting of HDD0 and HDD1

DC1.vhd: on HDD2

EXCHANGE1.vhd: on HDD3

SQL1.vhd: on HDD4

RD1-SYSTEM.vhd: on HDD5

SQL1-DATA.vhd (90GB) and RD1-DATA.vhd (182 GB) on 272 GB RAID 10 volume consisting of HDD6-HDD9

Notes:

  1. SSD0-1 could be used instead of HDD4 and HDD5, on the idea that they contain the most active vhds.

  2. Backup for all vhds is 1TB USB HDD.

  3. Would it be preferable to create a 136 GB RAID 10 volume using HDD2-HDD and put DC1.vhd, EXCHANGE1.vhd, SQL1.vhd, and RD1.vhd on it, instead of giving each their own spindle?

Thanks

share|improve this question
    
What are you going to be running on it? The proposed layout looks great! As long as you don't need more CPU, RAM, or drive space :) –  warren Oct 11 '09 at 14:43
    
What is the "RD1" VM for? Is it a terminal server for many simultaneous users? –  Massimo Oct 11 '09 at 15:05
    
@warren, @Massimo: DC1 - about 20 computers and 25 users; EXCHANGE1 - about 25 users; SQL1 - one database, currently about 20 GB, a back end for a propriary LOB application; RD1 - terminal server - about 25 users, with front end for LOB application, Office, and serveral misc apps. –  sejong Oct 11 '09 at 15:13

3 Answers 3

I totally agree with splitting the VMs on as much physical disks as possible in order to reduce I/O contention, but I'd suggest using some sort of RAID, because I don't think you want to lose a VM in case of a disk crash.

I think creating two-disks RAID 1 arrays would be a better solution.

Don't worry too much about the DC1 VM: a standard DC/DNS doesn't do really much disk activity, so you can (usually) safely put it on the same array as other VMs.

share|improve this answer

Thanks. I think I can implement that as follows.

Physical OS on HDD0-1

DC1.vhd (20GB) and EXCHANGE1.vhd (48GB) on HDD2-3

SQL1.vhd on HDD4-5

RD1.vhd on SDD0-1

SQL1-DATA.vhd (90GB) and RD1-DATA.vhd (182 GB) on HDD6-9

This gets all four VMs on RAID. The only tradeoff is the the SQL1.vhd (the SQL1 VM system drive) will be on a HDD instead of a SSD. This is acceptable because its that RD1 VM that has the heaviest use, and will benefit the most from the fast access and read of the SSDs.

share|improve this answer

Don't forget about your network interface needs. You might want to have at least two interfaces (our four for load balance/redundancy) so that you can have the Host communicating through a different interface than the VM's. You could also consider dedicated interfaces for certain VM's to avoid network bottlenecks for things like SQL and Exchange...

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.