Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Hello guys I hope that you are well with everything that matters to you including health, family, finance and everything else. We are having a big project that initially includes

• Active Directory migration to new hardware and upgrades from Windows 2003 to 2008 • Exchange server, migrating to new hardware upgrade from 2003 to 2010 and implementation redundant system. • Migration of ISA 2006 firewall and upgrade to Microsoft Forefront Threat Management Gateway (TMG).

Client currently has these servers

  • 2x Active Directory including these services WSUS, DHCP, DNS
  • 1x Exchange Server 2003
  • 2x SQL Servers
  • 2x File Servers
  • 1x ISA Server 2006
  • 1x Web Server

The current platform is in Windows Server 2003. On this project is planned to migrate and upgrade

  • 2x Active Directory including these services WSUS, DHCP, DNS from windows 2003 - 2008
  • 2x Exchange Server 2003 migrate to 2010
  • 2x SQL Servers
  • 2x File Servers extend capacity
  • 1x TMG
  • 1x Web Server Operating systems must be upgraded to Windows Server 2008.

The new hardware that client requests the project to be implemented in is

Four 4 x Dell PowerEdge R710 with this configuration

Processor - 2 x Processor Intel Xeon E5620 (2.40GHz, 12Mb Cache, 5.86 GT/s QPI, 80W TDP, Turbo, HT, 4 Cores) Memory - 32GB 1333MHz DDR3 Memory (8x4GB Dual Ranked LV RDIMMs) RAID Connectivity C4 - RAID 5 for PERC H700, Min. 3 Max. 8 Drives Hard Drive 6 x 300GB SAS 6Gbps 10k 2.5" HD Hot Plug Network Card Two Dual-Port Embedded Broadcom NetXtreme II 5709c Gigabit Ethernet Controller with 4P TOE

I know that this information is not enough to make the decisions but the client does not give further information for the current platform

My questions are: Is this hardware enough? Can you propose a system diagram or something like that how system infrastructure should look? Also if you can describe the best practices for the migration and update of the current server to new servers. Any additional information is appriciated. Sorry if I bothered you in any form. Best Regards.

share|improve this question
RAID 5...? I wouldn't personally. And those four machines are running hyper-v I take it. You don't mention anything about current performance. Or what they're doing for backups right now. – Bart Silverstrim Oct 26 '11 at 13:01
up vote 2 down vote accepted

There is no easy answer to this question, and one cannot answer "is the hardware enough" without having distinct metrics for your environment.

This is a quite large environment you're proposing (for reference, I've got a 2 host Hyper-V cluster for a 350 user company with one each of those roles you mention), so going by best practices or vendor recommendations is only going to end in poor performance or wasteful spending.

What needs to be done is a performance analysis of existing infrastructure, for all facets of hardware (CPU, disk, Memory), ESPECIALLY disk I/O. Most vendors will do this for you if you're already in contact for quotes, and if they won't, they don't need your money.

You don't mention virtualization in your post, but do tag it as Hyper-V so I'm assuming you're thinking of going that route. To be at least a little bit helpful:

  • I suggest 1U servers like the Dell R410 or R610 as your Hyper-V hosts, specced according to your performance analysis results.
  • Get at least 6 NIC ports per server. You may not use them all immediately, but will be wishing you had when you consider SAN performance, multi-path I/O, and live migration networks.
  • Have at least 3 Hyper-V hosts (or more if warranted), so that failover of one host only requires 33% (or less) of available capacity on the remaining.
  • Bring in a SAN sized for your I/O. If you size for I/O, you will inevitably have more than enough capacity, unless you have a really large data set.
  • Make use of Hyper-V Cluster Storage Volumes for high availability
  • If you're wary of keeping your SQL I/O mixed with everything else, add a shelf (or at least a disk group) of disks in your SAN just for SQL, and configure it as pass-through rather than VHD storage.
  • Buy Software Assurance on your Windows Server licenses, because Windows 8 is coming next year and offers MANY Hyper-V improvements.
  • Don't skimp on networking for your SAN. A bad switch will bring a lot of well designed infrastructure to its knees.
share|improve this answer
Well, technically the Softwar Assurance is not needed - you can run the company safely on Hyper-V server (for Hyper-V)... the other updates are a separate issue, but the Hyper-V update wont cost anything anyway. – TomTom Oct 26 '11 at 14:53
That is true, but with an environment where the VMs are mostly Windows, where Hyper-V clustering is going to be used, it may be better to use Datacenter licenses for the Hyper-V Hosts, especially if this post by Aidin Finn is accurate ( - Guest OS Availability section) – Jeff Miles Oct 26 '11 at 16:02
Yes, BUT... unless you want to upgrade the clients that still does not mean SA makes sense. YOu can update Hyper-V without updating the datacenter license ;) Hyper-V 3.0 will fully support to my knowledge 2008R2 clients. – TomTom Oct 26 '11 at 16:20
Well said; I've been thinking "inside the box" too much today, and am now experiencing Microsoft tunnel vision. – Jeff Miles Oct 26 '11 at 17:06

Consult a virtualization specialist for a tailor-made solution; the subject is too broad, and the variables are too many, to provide one definitive answer.

Points you need to think about include:

  • while RAID-5 could work on low-I/O VMs, it is entirely unsuited to databases or Exchange; you aren't going to get away with it on applications that are always I/O-bound
  • 32GB per host is way too little, especially when you intend to combine SQL, Exchange, and other special-purpose functionality on one host machine.
  • since it doesn't look like you are pooling these hosts for their resources, you're pretty much stuck on setting up N virtual machines on each of 4 hosts.

This essentially means you're getting a pretty suboptimal resource spread from your high-powered boxen.

share|improve this answer
After Exchange is migrated to 2010 the IO thing is a moot point. – Chris N Oct 26 '11 at 13:46
How so ? Is storage in 2010 so radically different that it no longer uses hard disks ? – adaptr Oct 26 '11 at 13:48
The IO issue is less, but exchange and sql server LIVE and die by IO budgets. THe server is bad. Seriously. I run 8 velociraptors in a Raid 10 for Hyper-V and I have IO problems at time - anmd that is with my SQL Server using another 8 disc RAID 10 for data. And both (host, sql) have another 2 mirrored discs (each) for hot files. STILL IO issues. Unless your Exchange doesn othing and SQL is "ok, I have some data" it wont be enough. (my sql database has around 800gb, I deal with... – TomTom Oct 26 '11 at 13:55
...500 million row queries regularly). – TomTom Oct 26 '11 at 13:56
In that case, Disk I/O certainly matters for you. As the other responder said, buy a SAN, or dedicate extra disks to SQL server, in RAID-10. – adaptr Oct 26 '11 at 15:33

Your Answer


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

Not the answer you're looking for? Browse other questions tagged or ask your own question.