Microsoft's ongoing crusade to establish a firm foothold in the enterprise got a little more traction recently from a benchmark test conducted by Accenture that show Windows .Net servers compared favourably with Unix-based servers running mission-critical applications.
The benchmark, conducted in Accenture's High Performance Scalability Lab, compared actual performance data from a manufacturing company now running SAP AG's R/3 data base on Unix with those results from running R/3 on Microsoft's Windows .Net server.
According to the results, the Windows .Net-based application had the same, and sometimes higher, response times compared to the Unix-based system, specifically 508 milliseconds to 700 milliseconds, respectively. They study also showed that the .Net server could be deployed and maintained at about 30 percent of the cost of the Unix solution.
"The results surprised us in a couple of ways. First in terms of price, but more importantly, we were surprised by the actual online performance with a heavy batch component running [on the .Net server] at the same time," said Tim Donahue, an associate partner with Accenture in Columbus, Ohio.
According to Accenture officials, the specific testing environment included an SAP R/3 logistics, sales, financial/controlling, supply modules, 1,700 active users, and a 60 percent batch load.
For the Unix solution Accenture used SAP R/3 4.0B running on a Sun Microsystems Inc. 34-way 400MHz E10000 server with Solaris 2.6 and Oracle Corp.'s 8.04.4. For Windows .Net Accenture used SAP R/3 4.0B running on an eight-way 900MHz Pentium III-based Compaq Computer Corp. ProLiant 8500 server with Windows Advanced Server and SQL Server 2000.
While the benchmark may give Microsoft fuel for its marketing fires against Unix vendors such as Sun, some observers believe the company will need more than positive benchmarks in the labs to gain real-world market share.
"If Microsoft is serious about getting into the upper reaches of the data center, they need to be successful with large-scale projects like SAP. But you need more than a fast box. They need proven processes and tools to be able to surround that box and turn it into a complete business platform, and that has not happened yet," said Derek Prior, a senior analyst at Gartner Inc., specializing in server-based application such as SAP's.
While the information that comes from a benchmark such as Accenture's should not be taken as the final word, Prior said it does provide users trying to make buying decisions with "some useful data" to consider.
"I won't say this situation is ideal, but it gives you some reference points. Some ERP [enterprise resource planning] vendors are lousy at setting up the professional sizing processes that characterize the performance of software for a user's project. But SAP does a better job than most," Prior said.
Donahue said Accenture has been talking to Microsoft about doing some early testing soon on the follow-up to SQL Server, code-named Yukon, a strategically important XML-based, 64-bit database not scheduled for release until 2003.
"There is a lot of interest at Microsoft and among the major OEMs we work with to test these environments in similar way. When you get the ability to have a 64-bit environment that has a much larger memory footprint, it will be interesting for applications like those for supply chain and other large optimization routines," Donahue said.