Web Performance: Sum of the Pieces

By David Strom

May 1, 2000 —

It seems odd, but one of the biggest risks you face with your Web-application development projects is the price of popularity. The mistakes of Victoria's Secret and other sites that buckled because they didn't plan for scalable success should be familiar warnings to us all. But building your applications for speedy response (or even merely acceptable response) under heavy loads is more of an art than an exact science because finding the potential performance bottlenecks can be difficult. In fact, there are three different general approaches to performance measurement; which one or ones you use can determine how you need to adjust your applications accordingly.

THREE APPROACHES

First is the brute force approach, where you ask the IT staff, or maybe your company's employees, to hammer away at a particular URL a week or so before you are ready to unveil it to the public at large. Your hope is that if you tell enough people and they hammer away diligently, you might expose some weakness or potential choke points. A variation is to use a test lab with hundreds of machines capable of creating thousands of simultaneous sessions, such as the KeyLabs facility in Utah, and let them do the stress testing under more scientifically controlled conditions.

Brute force is fine for simple things, but the biggest test lab can only do so much stress testing.

A second method is to use application modeling tools as a substitute. A good example of this kind of approach is with Ganymede Software's Chariot and Pegasus tools. You can record typical Web application scenarios and play them back through your network to determine if these applications will bog down anywhere. These are useful if you don't have too many machines or users available for real-time testing (you can play back streams of multiple users' actions on a single machine), but getting the tool set up can be time-consuming, and the results are complex to analyze. You need to make sure that you indeed tested the right thing and found the actual bottlenecks in your applications and not something inherently wrong with your network-or with your test configuration.

A third approach involves using a third-party company to monitor your applications from afar, over the actual Internet and in real time. Such a "management service provider" is still a relatively new concept, although several companies have been involved in doing relatively crude monitoring for years. Most of these efforts have tested only whether or not a Web server is actually up and running. Ipswitch Software's What's Up, Freshwater Software's SiteSeer and Vertical Software's WebPartner will all send you e-mail or pager alerts when your Web server connection from the Internet is lost. And NetResolve's SiteWatch takes these a step further with more details, such as the number of e-commerce transactions and page load times.

But none of these services truly provides any insights into what your applications are actually doing when they are under the load of multiple hundreds or thousands of users-insights that your team can take stock of, so that the developers can get inside the code to improve the performance of the applications.

There are two companies that are in this new area of management service providers: aptly named Manage.com and Luminate Software. Manage.com monitors mostly e-commerce applications, providing details on transaction throughput and end-to-end metrics that will help you avoid problems if your storefront becomes popular and the virtual aisles become clogged with throngs of cyber- shoppers.

Luminate's Mango offering got its start monitoring SAP applications on local-area networks and is just now branching out into generalized remote Web application monitoring, including covering popular Windows NT-based Web, Exchange and SQL Server applications. You can track inter-relationships among Web, database and Internet applications using a series of reports, all available via a Web browser.

Both of these companies have the beginnings of a great service. Manage.com is an expensive proposition, befitting more high-end Web storefront providers. Mango, on the other hand, charges only a few dollars per monitored server per month.

CHOOSE AT LEAST ONE
No matter which approach you choose, you still have to spend some time understanding the different pieces of your Web applications and how they all fit together if you are going to plan for popularity. But do spend some time with at least one approach-to do nothing and just assume that because a site's software works correctly under test conditions it's ready for prime time is foolish.

Share this link: http://www.sdtimes.com/link/24944

Open Web Foundation delivers legal framework for creating standards Industry consortium comes up with legal definitions and licensing terms for work done to advance Web services technologies

ActiveBatch 7 puts Web services into workflows For Web developers using ActiveBatch, the inclusion of a library of WSDLs allows them to build workflows across applications by using templates. Integration with enterprise applications fills out the latest version.

Companies link up to test their services against Web standards The Web Services Test Forum is an alliance of Web service infrastructure software makers and their clients. The members create scenarios and try to detect any problems that their services have when implementing different Web standards.

Add comment

Name*
Email*
Country

Comment
Preview

News on Monday
more>>
SharePoint Tech Report
more>>

Download Current Issue
ISSUE 3/15/2010 PDF

Need Back Issues?
DOWNLOAD HERE

Receive the print Edition?

Google Code turns 5
Google Code Turns 5, and adds a Paxos Algorithm to make the system more stable and reliable.
03/17/2010 11:16 AM EST

Test your Visual Studio 2010 know-how
Microsoft is offering free beta certification exams for Visual Studio 2010.
03/17/2010 11:08 AM EST

Microsoft lifts the hood on IE9
Microsoft is previewing IE9.
03/16/2010 01:10 PM EST