Server Resources

[cfk]

Has anyone else found CFServer to be highly inefficiant?
I need to know as many tips and tricks to spead up
the server.
Currently I'm putting as much code onto the MSSQLServer
in the form of UserDefinedFunctions,StoredProcedures AND Triggers.
The site I'm working on will potentialy have 200 hits/second.
Yes I know this is a high figure, but a competitior is
currently having a much higher hits per second.

 

[nsampson1]

I'm running an application that receives around 2000 concurrent hits at busy times. Despite all possible optimisation in the CF administrator, coldfusion still ground to a halt every couple of hours and the logs would become full of 242 errors.

Although my app is still written in CF, I had to re-write all of my core scripts that handle incoming requests in asp. Works like a treat now.

Maybe not what you wanted to hear but it saved me lots of valuable time in the long run.

 

[cfk]

Traditionaly I'm an ASP programmer but through PHBs the criteria ended up with cold fusion. Not a major problem for most sites. So I'm not surprised but the developent has been 11 months now. Thats a large amount of code to adjust.
ITSHAPPENS.
Do you know of any accuate, independant documentation
confirming what we know to be true. That is, ASP
is far more efficiant than coldfusion.

 

[cfk]

By the way. My boss is bald, but wishes he had pointy hair.

 

[nsampson1]

there are many similar postings to this on the allaire (now macromedia) forums (go to support, then coldfusion on the macromedia site) but allaire/macromedia have been loathe to admit to any performance issues at all.

many of these threads suspect that the performance issues are caused by the 242 errors. However, 242 errors are simply notifications that a user has kicked of a script which has taken so long to run that they've ended the process and gone elsewhere so all they prove is that CF is running slowly.

When I had this problem, by running cfstat from a command line it was apparent that cf requests were simply building up in queues and not being processed quickly enough. Once the queues reached a terminal level, the service would hang. This happened despite our load-balanced 4-server architecture, each with dual P3 1GB CPUs and countless re-writing/ optimising of scripts.

Make your own conclusions

I'll let you know if I come accross any documentary evidence

 

[cfk]

Wait till the client finds out??