Samba beats Windows Server 2003
Or does it?I've been reading in IT Week that Samba 3 extends is lead over Windows Server 2003. But before getting too excited, I felt it worthwhile to read the details carefully.
Two comments stood out for me: First: "We selected a low-specification but otherwise modern server for our tests. We used an HP ProLiant BL10 eClass Server fitted with a 900MHz Pentium III chip, a single 40GB ATA hard disk and 512MB of RAM. We did not tune any of the software to improve performance." And later: "Each NetBench client makes a constant stream of file requests to the server under test, whereas in real-world environments many users would remain idle for long periods. Consequently our test environment simulates the workload of some 500 client PCs in a typical production environment." So out of the box, on a low end server, a Linux/Samba box performed better than Win2k3 out of the box and untuned. I guess the first question I have on this is to ask why you'd seriously consider putting an important mission critical file server, serving a large community, on a single ATA disk, using a small, underpowered blade computer with limited memory. The test is meant to simulate 500 users, that equates to around 80mb per user - this is 1/3 the size of the memory card for my digital camera.
Their comment about not tuning the system also does not ring totally true. In my experience, installing Linux is an exercise in tuning at least to a degree. If they chose a very thin Linux kernel, possibly one compiled only for only the PIII chip, and loaded only Samba, then they are doing tuning. One thing that could make a huge difference to Windows is how the file and print service is setup.
What I'd really liked to have seen where the bottle neck was while this test was underway and to have seen what effect adding a decent amount of RAM would have had. I suspect the system was kind of busy paging. I've not studied the NetBench benchmark well enough to know how it works when running in this configuration.
So, I'm not really sure if this test if all that valid. Of course, it looks good but what I'd like to see is this test repeated on a properly specified/configured system.