What is proper architecture for my website (1000+ concurrent users)

I have a social networking site with 100000 members and around 1000 concurrent users.

I am looking for architecture recomendations as my servers keep crashing.

The site is build in Coldfusion 7 and MS SQL (all sql is in stored procedures).

4 Windows 2003 servers.
One is for the db and has 2 procs with 4gigs ram.
Two web servers with 2 procs and 2 gigs ram
One server that stores and serves the members photos (this in particular gets really slow) Same specs as web servers.

Would this be an issue of IIS 6? Can it be the application pool? Web garden?

Please let me know what do you think is the proper architecture for this site. Keep in mind the traffic is increasing by 3-5% daily, so i need room for growth.

Thanks
TsoukiasAsked:
Who is Participating?
 
Dave_DietzCommented:
There have been documented memory leaks in jrun in the past - what version of jrun are you using?

Dave Dietz
0
 
Dave_DietzCommented:
How many requests per second is the image server handling and are the image files local to that server or is it pulling them from a different location?

Not likely to be an issue with IIS - more likely a bandwidth problem.

Dave Dietz
0
 
TsoukiasAuthor Commented:
I have been surprised by the traffic increase and I'm saying this because I will have to ask some novice questions.

How do I check the per second requests of the image server?

PS. The image files are local to the server yes.

Thanks
0
Network Scalability - Handle Complex Environments

Monitor your entire network from a single platform. Free 30 Day Trial Now!

 
Dave_DietzCommented:
One way is to look at the w3svc logs and count the number of requests that are handled in one second, although this wouldn't account for spikes and drops in the requests if you picked a slow/busy time.

Best bet would be to use PerfMon - under the Web Service object is a counter "Get Requests/sec" that would probably do the trick.  If you go this route track the "Bytes Sent/sec" as well so we can tell how much data you're pushing out.

Dave Dietz

 
0
 
TsoukiasAuthor Commented:
Dave, as I am doing what you mentioned last, i just want to also note that the effect i get is the X box when the image cannot be found. After a long wait they load.

Another note here is that as the images are not loaded the progress bar of the browser does not complete.

Thanks
0
 
TsoukiasAuthor Commented:
First finding is that something called Avg. Disk Queue Length is always 100% (its scale is 100.000 - dont know what all this means just giving you more info)

Get Requests/sec:
Maximum: 142.636
Avg: 95.512

Bytes Sent/sec
Maximum: 3595918.812
Avg: 324645.648

Third finding is that
0
 
Dave_DietzCommented:
How fast is the network connection of the web server and how fast is your pipe to the internet?

Dave Dietz
0
 
TsoukiasAuthor Commented:
Sorry for this, but how can i check these two things? My host is aplus.net.
0
 
Dave_DietzCommented:
If you have remote access to the servers you can simply look at the network adapter to see how fast it's local area connection is.

You might need to contact your hoster to see what your available bandwidth is.  It is possible that they may be capping your bandwidth at a level that isn't high enough for the traffic you are serving.

Dave Dietz
0
 
rbbrownleeCommented:
What server is crashing?  
Is it freezing up or is is restarting automatically?
Is it the Web Server or is it the SQL server?
0
 
TsoukiasAuthor Commented:
Dave, I found out that I am on 100mb, and that there are no restrictions from them (although it might be a lie, its what they told me).

Its doing it again atm http://www.fatsa.gr/new_members.cfm while the rest of the site is loading very fast.

its killing my traffic, what can i do? help!! :(
0
 
TsoukiasAuthor Commented:
this is my hosts findings:

I have checked into your server and as the previous technician said I have confirmed, your server is essentially running out of memory. The process jrun.exe which is associated with coldfusion is using up a large amount of memory forcing your server to start using the windows pagefile. Now the pagefile is being used and the hardrive is being used so much that the hardrive write queue gets filled up to 100% and ccauses a really severe bottleneck on the server.

Do you agree it can be this? The server has 2gb ram currenctly
0
 
rbbrownleeCommented:
Hi Dave,
I have found our server to occasionally do the same thing.  

Do you use this server for any PDF generation?  

Is the photo server behind a firewall?

The combination of these two things can cause the jrun.exe file to run out of control if you are not careful.

0
 
TsoukiasAuthor Commented:
I am using 1.4.2_11 which is the latest one officially supported by Coldfusion. Adobe (macromedia)

leaks such as?
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.