Strangeloop Uses NetForecast’s Response Time Equation to Explain ASP.NET Performance Issues

Richard Campbell and Kent Alstad at Strangeloop Networks use the NetForecast response time equation to explain ASP.NET performance issues in an article entitled "Scaling Strategies for ASP.NET Applications" published on the Microsoft MSDN Magazine web site.

The article discusses user application performance satisfaction (response time) and the throughput needed to support many users (scale).  They propose ways to measure each key equation element in an operational system and improve its value.

They add a factor to the equation reflecting a browsers' ability to operate concurrent flows to load the content.  This concurrency is also called multi-threading or TCP multiplexing.  We have added a similar extension to our equation in many of our engagements (we call it M for multiplexing factor).  The multiplexing value is an interesting phenomenon, which can significantly effect task response time. But the developmental history of the browser has caused the factor to bounce around in the following sequence: 1, 4, 2, and now 6.

Early browsers had none (M=1).  Then came IE3 which raised it to 4.  But by IE5 we were complying with HTTP1.1 so the factor dropped to 2.  But Microsoft's recently introduced IE8 pushes multiplexing to 6.  This is great for response time experienced by a single user, but potentially bad for multi-user throughput since many web servers can't handle three times as many TCP connections to support the same number of users.

But we digress.  Back to the Strangeloop article.  It explains the unique performance and scaling pitfalls of ASP.NET and how to overcome them.  Most of the fixes require a lot of heavy lifting involving application configuration or tuning.  Fortunately most of the performance improvements are automatically implemented by the Strangeloop AS1000 appliance, and many of those scaling improvements also save server resources and/or bandwidth costs.

This is just one example of an application delivery system (ADS) solution that employs a group of control or acceleration techniques.  Our "Field Guide to Application Delivery Systems" describes the techniques used by ADS vendors.  You can find it at the NetForecast web site as report NFR5085, September 2006.

If you need to control or accelerate your application over the WAN, the guide helps you figure out which techniques address your particular set of challenges.  It describes all the techniques and shows where they apply their "magic" and which application class or protocol they improve.  Each technique is referred back to the response time equation to show which term(s) it affects.

Should you want to wear a copy of the NetForecast response time equation, Strangeloop will be giving away T shirts emblazoned with the equation at Interop in booth 657.  But the offer is exclusive to the readers of this blog - you have to ask for the equation T shirt.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.
Related:
Now read: Getting grounded in IoT