Virtualisation of Webservers

[hairy_potter]

Hi everyone,

I would like to ask you for your professional input:

Attached you find a scenario of my planned new webserver-virtualisation....

• first a load balancer & reverse proxy (and additional another in standby) .. reverse proxy (which is clear), but regarding "Load Balancer": can one and the same VM be booted from SAN at the same time on several machines and be clustered or load balanced or does the ESX do those clustering/balancing things itself?

• 2 redundant switches

• 3 ESX Server as cluster with HBA

• LAN (Heartbeat, VMotion, HA) for ESX with redundancy

• finally 2 redundant 4GBit FC Switches to the SAN storage

Do I have an error in reasoning, or does it really work like this…..(aside from dimensioning, load level,…..)

Thanks for your input & greets from Austria

Cheers Tom

[Dave_Mishchenko]

Hi Tom, for the below a VM will only be able to run on a single ESX host at a time. You'd be able to configure it with up to 4 virtual CPUs and 16 GB of memory. If the case of the failure of an ESX host, the high availability option would restart the VM on another host, but it would be rebooting from scratch. You can use DRS/vmotion to load balance your ESX hosts, but even in that case a VM will only be running on a single host at a time and would be move to another host automatically if the load rules were setup in such a manner.

> reverse proxy (which is clear), but regarding "Load Balancer": can one and the same VM be booted from SAN at the same time on several machines and be clustered or load balanced or does the ESX do those clustering/balancing things itself?

Are you looking at 2 physical switches or 4? 2 is fine, but in that case you'll want to use VLANs to isolate your vmotion / VC traffic. Would there be any other servers on the management LAN? Also what kind of switches are you planning to use?

[hairy_potter]

Hi Dave,

thx for your quick reply!!

I`m looking for two physical switches in "WAN" area (two redundant connections between the caching reverse proxy {+behind the uplink+} and the ESX hosts - red&blue zone in my graphic), and two physical switches in the "LAN" (green&brown colored - shown above in the attached scheme) for VLANs to isolate the vmotion / VC traffic. Just the 3 (or maybe 4) ESX Hosts and the "Management Center" host are on the management LAN (maybe the "Management Center" will not run on a dedicated physical host but inside a virtual machine instead).

so I´m planning to buy 4 ethernet switches and 2 FC switches:

2 x FC switch <- redundancy to SAN

DELL FC4 McDATA 4400 switch :: 16 port

4 x ethernet switch <- 2xVLAN 2xWAN

HP ProCurve switch 2824

does my scheme make any sense this way?

cheers tom

[DFN]

1. I would recommend not hosting your Management Center on a virtual machine. As the Management Center is going to manage the virtual environment I like the concept of keeping it "outside".

2. Two switches in the WAN sounds fine with two redundant drops.

3. Internally I like three VLANs minimum. 1 for Vmotion, 1 for Console and 1 for production. But these can all reside on two physical devices.

4. I load balance my web servers using Cisco Content switches between my WAN switches and my LAN switches so this is where our configurations are going to differ.

5. I have my web server farm across 3 ESX hosts configured in a HA cluster. I configure no load balancing at the guest level.

DFN