Hi, over a year and a half ago, our company developed a web site for a local news radio statio using DNN. The site was performing well, until a few months later, when the performance started to decrease.

The site has it's personalized set of modules, so the admins and the journalists can post news every day (aprox. 50 every day), list the news in categories and read a single news; and it was this set of modules that caused (partially) the problems, so after a few fixes to the news modules, the site increased it's performance.

However, the station's QA departament made some stress tests to the site, and the results didn't satisfied them. We at our company desingned a set of test using jMeter.

(In case someone has used it here's how we designed the test)

Created a Thread group for 100 users, with a 1 second Ramp-Up Period, the results are:

	Samples	Average (miliseconds)	Median	90% line	Min	Max	Error %	Throughput	KB/sec
Page	100	21578	22309	33012	6672	34854	0.0	2.78621381404809	513.0633158382325
TOTAL	100	21578	22309	33012	6672	34854	0.0	2.78621381404809	513.0633158382325

As you can see, the average response time is 21.5 seconds with 100 concurrent users, and even though before the fixes, it was over 50 seconds, QA says it's still unacceptable, and say we need to bring that number down to 10 or below.

The site is hosted on Rackspace, the specifications are:

Memory / Size	Storage / Size	CPU Allocation (Theoretical)	Network Throughput (pub / priv)
8 GB	320 GB	4 vCPUs	150 Mbps / 300 Mbps

Software Specs are:

Windows Server 2008 Enterprise, 64 bits.

SQL server 2008 R2, (it's on the same server as IIS, separate them it's not an opition).

DNN's version is 5.5.0

Module Cache Provider: File

Performance Settings: Heavy Cache

Compression: Gzip

Given those specs, are there any recommendations we can implement in order to achieve our 10 seconds goal? In case you need to know any other information, just ask and i'll try to post it.

Thanks in advance for any help anyone can provide!

Is there a way to implement Page Level Output Caching in Dot Net Nuke?

Sebastian is nudging you in the right direction. You MUST measure before you start fiddling.

The purpose of measuring is to find out where the time goes. Then you have clues about where to spend effort.

Before even that I quesition if the stress test is reasonable.

If the journalists post 50 articles PER DAY, then a workload simulating 100 CONCURRENT journalists tells you nothing useful. The site will never see that workload in real life.

What is the customers real problem? Do the journalists think the response is slow? Do the radio station listeners think the response is slow?

---

Best wishes,
- Richard
Agile Development Consultant, Practitioner, and Trainer
www.dynamisys.co.uk