I don’t know if anyone saw the reports this week that Samba 3 out-performed Windows 2003 Server by two-and-a-half times…? Read on for an “explanation” why by Microsoft 
Did old clients mar Samba test?
Microsoft says Windows 2003 can keep pace with Samba 3 in file and print performance - given XP clients.
IT Week recently reported that version 3 of the open-source Samba file and print server software is now two-and-a-half times faster than the commercial alternative of using Microsoft Windows - see the first web link below.
While the web is alive with rumours that Microsoft forbids any benchmarking of its products, IT Week has never had problems testing Microsoft kit. In fact, we have full access to Microsoft products and much of its beta software via the Microsoft Developer Network (MSDN), which has a facility to download CD images of pretty much all its current offerings.
Even so, our report apparently caused heads to turn at Microsoft, and eventually prompted a telephone call from the file and print gurus who work at the vendor’s labs at its Redmond headquarters.
Clearly, Microsoft takes the issue of file and print server performance extremely seriously. This is not surprising, even though hype about web services, dot-Net and Longhorn might lead one to believe that file and print is old hat. After all, while some firms use Windows for other enterprise applications, if you have only one Windows server, it’s almost certainly running file and print services. Indeed, in the early days, this was how Windows NT competed against old rival Novell NetWare.
More recently, with NetWare in effect buried, Microsoft set up a file and print business unit, which has released a version of Windows specifically for network attached storage (NAS) file devices. Various NAS vendors offer low-end and mid-range systems based on this, and the idea that Samba delivers the same performance with half the hardware required by Windows might be a cause for concern for some of them - because rivals could make products that cost less but deliver the same performance as the Windows-based systems.
But rather than dispute our results, Microsoft seems content to point out that our test results would have showed less difference between Windows and Samba if we had used PCs running Windows XP rather than NT 4.0 as clients on our test network.
Actually, it seems even Windows XP is not sufficient, and the Redmond gurus also suggested configuring our clients with specific patches to improve performance for our tests. While I agree with Microsoft’s implication that few firms use NT 4 clients, I’m surprised it suggested upgrading to XP. After all, there may not be many NT 4 clients out there, but I’d guess there are even fewer XP ones. Of course, Microsoft suggested XP because it’s the only version of Windows to contain these file and print enhancements.
Unsurprisingly, we cannot upgrade our test network clients to XP without replacing hardware, so even though the cost of the software is not an issue, we can’t simply upgrade to XP.
More to the point, it seems strange that one must upgrade desktops to improve the performance of the Windows server operating at the back-end.
But I guess it’s because Samba delivers excellent performance regardless of the client system that it’s so popular.
[i]original article: http://www.vnunet.com/Comment/1147194[/i]