I am current on windows server 2008 R2 and using IIS 7.5. There 15+ apps running under the iis.

My question is that : how can we determine when we need a web farm setup? What statistics do we need to use to tell 'ok dude, this is not good. We need to set up a web farm'

  • 1
    When somebody complains!? – mailq Aug 10 '11 at 17:57
  • It is the easy way. When the web sites are starting to shut themselves down, it is easy to decide to do something. The deal is to be able to prevent those kind of situations. – tugberk Aug 10 '11 at 18:00

Performance monitoring.

Simple stuff starts with CPU, RAM, disk IO of your webserver, usually via SNMP or WMI, using something like SCOM or Nagios or MRTG. if your CPU is pegged or you've used all your RAM, you've probably got a problem. There's a lot more you can monitor on the server, but those are the basics.

More advanced performance monitoring is using an external host with a browser agent to run certain browsing or transaction operations against the website, and measure the response times; if it starts regularly peaking above a desired response time, it's time to upgrade something. What it is that needs to be upgraded can sometimes be determined by the above simple monitoring, or it may require a deeper dive into your web application's code.


Besides performance and availability advantages, the web farm framework of IIS also helps you to patch machines individually. So apply your software to the primary machine and the framework will copy this to the other machines without downtime for your users.


This question can't really be answered with what you want - there's no rule of thumb for how much traffic or how many services you should be at to start considering more systems, aside from this:

When the performance or resiliency of the single device is no longer satisfactory.

It's a business decision - point to the single device that's floating everything, express how easy it would be to trip over a power cord/network cord/windows update and bring everything to a screeching halt, then let the decision makers make the call. That's all you can do.


Since your question is quite abstract i'll try to answer it in the same way: either when latency is becoming a problem or throughput is, or a mix of both.

  • either you are missing the point or I am lack of ability to explain myself. see my comment on the question. Maybe it gives you something. – tugberk Aug 10 '11 at 18:02

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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