I would recommend against the shared RAID array on the App servers. You
won't be hosting data on your app servers. Instead, each machine can have a
simple RAID 0 configuration (mirrored drives) for reliability, and good
backup. It is far more important that you be able to publish changes to
your application server than it is to keep data on them, so good management
and publishing software is essential. Make sure to mirror your system
partition as well as your application partition (some folks forget to
mirror the system partition, and end up with costly downtime). Keep your
applications off of your system partition (configure IIS to run from the D:
drive). Make sure that both machines can be backed up with a nice fast
backup unit.
Practice restoring them.
Don't plan to run Biztalk on this config. If you want Biztalk, it will need
its own servers.
Also, you are going to want to put in load balancing so that the same apps
will run on both machines. Normally, for an app server farm, I *start* at
three servers and go up from there, depending on load. Sounds like your
load is light, but if you wait until you have a heavy load, it may be
difficult to fix this. Get three servers in a load balancing configuration.
As for load balancing, Windows Server 2003 supports an excellent software
solution. If you are looking for a hardware solution, I've been happy with
products from F5 in the past, and I recommend them.
Make sure that you have a staging/testing configuration that mirrors the
load balancing, so that you don't wait until your app gets to production to
discover that the developer is storing the application state locally. (This
is much less of a problem in .Net than it was in ASP). Get a server to host
your Application Center Test and a couple of smaller machines to act as
agents, so that you can stress the test configuration and (after hours) the
production configuration as well.
The switches need to be fast and configurable. You will need to either
configure VLANs or have isolated switches. (Configuring the VLAN requires a
little more expertise, but it is considerably more flexible). Get machines
with dual Gigabit Ethernet cards and configure the cards to run as failover
for one another.
I agree with William about the racks and UPS and keyboard drawer. I would
add that you want a digital KVM solution over IP, so that you don't need a
keyboard and screen in every rack, and so that you can control the servers
from a remote location if you need to. This will save real headaches and
can make it much less expensive to live up to that uptime promise.
(especially if you want to get within any real chance of hitting four
nines).
Hope this helps,
--
--- Nick Malik [Microsoft]
MCSD, CFPS, Certified Scrummaster
http://blogs.msdn.com/nickmalik
Disclaimer: Opinions expressed in this forum are my own, and not
representative of my employer.
I do not answer questions on behalf of my employer. I'm just a
programmer helping programmers.
--
"Chris Gatto" <cg****@nbnet.nb.ca@removeme> wrote in message
news:ut**************@TK2MSFTNGP14.phx.gbl...
Hello,
My organization is currently considering the purchase of a new intranet
application server and we are looking for opinions from those who have
been down this road before. In brief the server must support the following:
- 600+ users using various ASP.Net based applications, only 2 currently
but growing over time as old legacy apps are converted;
- Some applications are mission critical, others not so much;
- DB back-end is a seperate SQL-Server 2000 failover cluster (2 machines)
with a shared RAIDed disk array;
- Service must be available 24/7/365;
- App server will act as combo web server (IIS) and application server
(.Net Framework);
We are considering the following specs:
- A 2-machine fail-over cluster with a shared RAIDed disk array (same as
our db design);
- Duel processor machines (2+ GHz cpu's);
- 4 to 8 Gb RAM;
I'd be interested in hearing what others have done for thier setups (pros,
cons, experiences).
Thanks in advance,
C Gatto