Virtualization

yes this is what i am planning to do do you have idea the recommended specification for  each node i mean thee server and the SAN storage then the san switch what type or models, like HP , or jujitsu , then in order to backup the environment i can use acronis let us say or?

You will need as much memory in each node to support ALL your current serves, and one physical processor core can support 5 VMs, bottleneck is often memory. As for storage you will need enough storage for all your current servers, once virtualised.

What budget have you got?

situation:
we are 160 employees
we have an exchange server 2008 (i think it's 2008, willl not log to my job just to check hehe)
we have an ERP server with a separate database server in sql server 2008 sp3

what we did 2 years ago:
we virtualized a few servers on a vmware vmx san

what we had to do after:
revert exchange and sql back to a real hardware server

reasons:
exchange was too slow
sql was too slow on big reports / query

now, the real reasons:
exchange and sql database use a defragmentation based on the hard disk clusters
also they do a lot of network traffic
the controller cache for harddisk read and write was saturated by others vm (probable)

now, hardware have evolved since then
(that was like 3-4 years ago)

so this is just an advice, from an old migration wich have turned bad for us (lost too much speed)

today i would probably put exchange on a vm, because the local cache can compensate for slow access to the database

but not sql, since our ERP is a little too hungry in big query

I'm a fan of HP servers.  I'd also use VMWare's VSphere instead of Hyper-V.  I just find it to be more solid and its added features work better, although I admit I haven't used Hyper-V in a while so the latest version may resolve some of the things I've found lacking.

I'd probably go with 3 HP DL380 servers with at least 1 Quad port NIC and I'd also use a Qlogic card for iSCSI SAN storage.  Then you'll need a SAN.  On the low end, I've worked with a number of different windows storage servers and have found them to be quite decent.  My first choice is a Dell Equallogic SAN - price is not cheap though.

With 3 physical servers, you could still run if a host failed and you were down to 2.  I'd put at least 32GB of RAM in each host, possible more depending on how much RAM you want to assign to each VM.  I'd make sure the RAM you assign to the VM's is at least the amount of RAM combined in 2 hosts.  This will help prevent disk caching and allow some room for expansion.  

I also like HP Switches.  A GB switch is what you'd want for both LAN and iSCSI and ideally two switches minimum for redundancy.  You could use VLAN's so at a minimum you could get away with 2 physical switches depending on how many connections you'll have.

Exchange will run fine in a virtual environment, but you do need to make sure there is enough I/O.  This depends on a lot of things including the size of your info stores.  The quality of the SAN makes a huge difference as well.  Never had an issue with an Equallogic SAN myself, but have seen some performance issues with other lower end SAN's.  The key is the max amount of disks in a single array for the max I/O.  

And with all this, you can't forget about backups.  They are still necessary.

Hi,

First of all Ive recently set-up similar environments. When it comes to your questions:

-When it comes to virtualizing systems you have 2 options. You'll either go to a solution of a full firtualizing embedded in a general Purpose OS (Such as Windows Hyper-V) or you'll go bare-machine virtualization (Such as VMWare or Xen Server). I personally prefer Bare Machine installations since it is always simpler and faster and easier on remote on such systems. Also more powerful.
- Since you're virtualiszing lots of servers you'd not like to place all your eggs on the same basket and you'll use at leaset 2 sergvers and go for some DRM setup. (Owhen one system dies it will restart on another server.
- Recently a client of mine wanted to deploy Exchange 2008 and What I had suggested was to place create 2 Microsoft clustered Mailbox Servers (because Exchange mailbox architecture is not that robust). So DRM does not reboot or move mailbox servers to another physical server and all HA clustering done over Microsift Clustering.
- We've placed other components like Hub + CS on a server and edge on another and it works very well..
- As hancocka suggested this will require a SAN and may be some server with 10Gb ethernet too to implement.

Cheers,
K.

Here's a config worth considering, this may be overkill but you can reduce the hardware requirements based on your budget:

3 HyperV Host servers:  maybe and IBM x3650 M3 6 core proc x 2 processors for a total of 12 cores
3 Copies of Windows Server 2008 R2 SP1 Data Center Edition (About $4100 each)
Quad NICs in each server and, say, an Intel dual port NIC for iSCSI
SAN:  maybe an EMC VNXe3100
System Center Essentials about $100 per seat, can be run as a VM
Run MPIO on the iSCSI NICs and purchase 2 good switches, like HP

Can provide more details later

You can indeed cluster VMs, services like File and Print, Exchange, etc.
Clustering VMs in particular with System Center and a SAN will allow live migration of running VMs from one host to another.  With R2 Clustering has been improved.
This can be a very good combination:  Hyper-V+SAN+Clustering (and ideally System Center).  And with Windows Server Data Center Edition you'll get all the OS licenses included that you can fit on any particular Hyper-V host.

One thing about Data Center Edition:  you have to license it for two processor sockets minimum so you might as well purchase another processor per host the get the best bang for your buck.  These days 6 core processors are reasonably priced so you can get a 12 core host for maybe $1000 more than a 6 core solution.

If you put all of that together you'll have an amazing environment.

If you go with clustering the Hyper-V hosts, you may want to shoot for 3 hyper-V host servers.  That way you can still take one system offline for maintenance, disasters, etc. and still have full fault tolerance and redundancy on everything.

good suggestion