Last 2 weeks, I run MVC training kit to IIS 7.5 and was trying to figure out the baseline for the performance tuning for my project. I tested 3 Scenarios,
S1) Hosting MVC 1.0 Training Kit in Application under Default Website
S2) Hosting MVC 1.0 Training Kit in New Website with port 8081
S3) Hosting my company’s MVC Project in New Website with port 8082
And their result of max throughput on Default Page is around,
R1 Result Screen Shot
R2 Result Screen Shot
This time, I tested in more scenarios,
S4) Moving MVC 1.0 Training Kit to root of Default Website
S5) Hosting MVC 1.0 Training Kit in Application under New Website with port 8081
S6) Hosting my company’s MVC Project in Root of Default Website
And their result now are,
R4 Result Screen Shot
R5 Result Screen Shot
R6 Result Screen Shot
OK, let me group the result in a table and so we could view all results more clear.
|MVC 1.0 Kit in Default Website
|MVC 1.0 Kit in New Website with port 8081
|Company MVC Project in Default Website
|Company MVC Project in New Website with port 8082
Now the results are showing that
1) The performance will be gained about 10% if you move the MVC project from Application under website into root of website.
2) The performance will be gained in a range from 240% – 1325% if you move the MVC project from Default Webste to New Website.
Recent, I am focusing on MVC/.NET/IIS performance because of my job duty. I have to try and running load test in order to get the number of max thoughput, so that I could figure out a baseline for the web projects in my working company. I then setup a computer to host IIS with MVC projects. Here is what I get,
CPU: Core 2 Due E8400
Ram: 4Gb DDR2-800
OS: 64-bit Windows 2008 R2 Enterprise Edition
Data: SQL 2008 Express Edition (for training kit data) + XML files (for my project data)
CPU: Dual-Core E5300
Ram: 4Gb DDR2-800
OS: 32-bit Windows 2008 Enterprise Edition SP2
Apache JMeter v2.3.4 with Java Runtime v188.8.131.52
Hosting MVC 1.0 Training Kit
Hosting as Application in Virtual Directory in Default Website
Separate Application Pool from default website
Testing with 100 threads, 500 threads, 1000 threads and 1500 threads
Default Page: ~200/sec
About Page: ~190/sec
Customers Page: ~73/sec
Interesting, this is the first time I try to do some benchmark. And I found out the performance between “Virtual Directory under Default Website“ and “New Website“ is hugh. It is 15 times difference. I tried to look around and still cannot figure out the reason. I will come back here later when I have some updates.
P.S., the above number of the max thoughput are just for reference. They might be up or down with difference hardware and applications installed. In my testing, Server CPU is running up to 100% already. So the bottleneck may be SQL service in same machine, network loading, Machine is not a “REAL” server hardware….etc.