如何解决ASP.NET 4.0 Web应用程序中包含ashx处理程序的内存泄漏？

Question

My team has built a web corporate ERP application using ASP.NET 4.0 with Entity Framework,ASHX handlers and web pages(>20). The app has been built to accomodate more than 15000 users. But when the application is hosted on Test Enviroment, it throws Bad Gateway error when logged in users exceeds 600. The app contains 3 layer - Front End(ASPX pages), Entity Layer, DataAccess Layer(Classes).

Backend server is SQL Server 2008 R2.

Testing report says - Connection TimeOuts, Large amount of Memory Leaks and Unmanaged Codes. We have tried our best to fix memory leaks by disposing the objects and classes,etc. Application doesn't contain minimal images.

But still the problem remains same. Even the server configuration of both Web and DB server is at very high end. 16 CPU and 4 GB RAM.

Can anyone help me out on this?

Answer 1

This is going to take you some time and 'digging in' to get a handle on what is causing the issue. Without being there and with access to the code, all we can offer is things to investigate, such as:

• Use performance monitor to get a better picture of what exactly is going on. Do this on both your web and database servers. There are different counters that you want to use for ASP.NET and for SQL Server. A search for perfmon.exe, counters, ASP.NET and SQL Server should get you on the right path as to which counters to start with. Run through various scenarios and examine the numbers in detail: a single user on the site performing various actions, 10 users, 100, etc.

• Use SQL Server Profiler to get some insight into what is happening on the database side of things.

• Use WinDbg - Internet search will get you on the right path - will take some ramp-up if you're not familiar with it, but it's powerful.

• I've had great luck in the past using tools like Red Gate ANTS Performance Profiler and Memory Profiler ( Red Gate ). The tools may cost a bit, but they have saved me way more than they cost in being able to quickly identify the 'low hanging fruit' issues. Disclosure - I'm not affiliated with Red Gate in any way.

• Do the problems in the test environment start exactly when the users exceed 600, or just generally at some number that varies but is over 600? If the problems always starts at the same number of users, that is a clue to look at what exactly happens when a new user logs in - are you loading a ton of data that isn't necessarily needed, etc. Actually, you should look into this even if the number of users where the problem starts varies, it just has a 'smell' if it always starts at exactly 600...

• Are you making heavy use of Session? Where is it stored?

• What is the required number of simultaneous users you need to support? You mention 15,000, are you load balancing across web servers? (This isn't necessarily related to fixing your problem, but something you may need to consider for your architecture)

• 4 GB of RAM is not very high end, especially for a database server. You likely need more, but this is not to say 'just throw RAM' at your problem, the above items should still be investigated.