Mike Brunt just posted an article describing basic ColdFusion clustering over on the Adobe site. It’s amazing that it took this long for someone to provide information like this. I know the info has been available in other places, but this article is clearly written and very understandable.

Luckily at work we have employed this exact method of clustering ColdFusion instances already. The only thing we didn’t do “correctly” was use the built-in CF web-server to power our Master Instance. Not a big deal really, but why not lighten the load on IIS another fraction of a percent if you possible can.

Mike really needs to follow up with an advanced article that addresses questions such as:

1) How many instances can you reliably run on a CF Server?

We can help answer this as we have effectively hit the cap on our two production CF servers. I don’t have the specs of our servers on hand, but they are 5-year old moderately powerful servers and they are each running about 12-14 instances of CF.

2) Horizontal Clustering

Mike’s article deals exclusively with Vertical Clustering, which is creating and running CF clusters all on the same machine. I know that CF also supports Horizontal Clustering, which is maintaining CF clusters across multiple physical machines.

I have only briefly played with this and I don’t believe I could get it to work acceptably.

3) Resource Management and Optimization

How much resources (memeory, etc) should be allocated to a clustered instance? How does one go about tweaking memory allocations? How can you test whether or not a chance is an improvement or not?

4) What Happens When Things Go Wrong?

How do you know when a cluster member has failed? Are they any common signs that a failure is imminent? What information can you examine in order to determine the susceptibility of your infrastructure to failure? What does (and should) a failure look like from the user’s perspective?

Anyway, it’s nice to know that there is someone out there actively working on answering these questions.