I think the results for Linux sound pretty good. For one thing they compared the bleeding edge release of Windows (Win Server 2003) to the near end of release cycle of Linux (2.4 kernel). And its Linux implementing Windows protocol with Samba. Even with the deck stacked in Windows favor on this test, it only managed a 86% to 95% improvement in the server function that it should do best. Even at this level I'll take the freedom, cost, security, and stability of Linux over Windows. While enterprises are locked into Windows Server 2003 for at least the next 2 years, Linux users will be enjoying the performance of at least 2 more kernel releases.
On the test method: it shows Windows Block Size as 64K, but Linux Block Size as 'default'. What is the default size for Redhat Advanced Server 2.1 and 8.0? Isn't this going to be a key parameter in file server performance? Was this changed as part of their tuning effort?
If this was a valid benchmark then it gives us something to shoot for (and I believe the new kernel has improvements in SMP and 'no-copy' file transfers). If its a heavily biased test then its another example of Microsoft having trouble speaking the truth.
Slashdot Mirror
I think the results for Linux sound pretty good. For one thing they compared the bleeding edge release of Windows (Win Server 2003) to the near end of release cycle of Linux (2.4 kernel). And its Linux implementing Windows protocol with Samba. Even with the deck stacked in Windows favor on this test, it only managed a 86% to 95% improvement in the server function that it should do best. Even at this level I'll take the freedom, cost, security, and stability of Linux over Windows. While enterprises are locked into Windows Server 2003 for at least the next 2 years, Linux users will be enjoying the performance of at least 2 more kernel releases. On the test method: it shows Windows Block Size as 64K, but Linux Block Size as 'default'. What is the default size for Redhat Advanced Server 2.1 and 8.0? Isn't this going to be a key parameter in file server performance? Was this changed as part of their tuning effort? If this was a valid benchmark then it gives us something to shoot for (and I believe the new kernel has improvements in SMP and 'no-copy' file transfers). If its a heavily biased test then its another example of Microsoft having trouble speaking the truth.