Sekhon's R Statistical Software Benchmarks
Another recent example of published benchmarks comparing Linux and Mac OS X were performed by Jasjeet Sekhon at UC Berkeley. Yes, that's the same Berkeley that built BSD, but Sekhon isn't a computer scientist; he's an associate professor of political science.
In a scathing article, Sekhon made references to Steve Job's "reality distortion field" and threw around comments such as "OS X is incredibly slow by design," and described the XNU kernel as "very inefficient and less stable" than Linux or BSD.
Sekhon posted some numbers and graphs showing how Mac OS X was as much as two times slower at running his statistical software than Linux or Windows XP, and attributed the problem to flaws in Apple's kernel.
As it turned out, comments Sekhon made about how Mac OS X works were incorrect; his application was performing excessive and unnecessary memory allocations; and finally, he had used an optimized allocator routine under Linux and Windows, but not on Mac OS X.
Sekhon posted a correction, and admitted that his benchmarks weren't an indication of problems with the XNU kernel, but rather with his software and the choice of allocator used on Mac OS X. Still, the impression left with anyone who read the article was that something must be wrong with the XNU kernel, when nothing of the sort was supported by Sekhon's problematic statistics.
Sekhon continues to publish his flawed benchmarks "for the curious," and he cites other proof that, while not reflected in his own findings, Apple's "Darwin is slower than the 2.6 Linux kernel in these benchmarks sometimes by several times." The proof? Links to Anantech!
Up to Five times More Ridiculous than other Benchmarks
How is it that two rather inept attempts to benchmark Mac OS X performance could morph into incontrovertible evidence that the Mac OS X kernel is so inherently flawed that the only solution is to replace it?
Doesn't it seem reasonable that engineers at Apple probably have a better grasp on how to optimize Mac OS X than some Linux affectionados at an adclick driven website, or an upset political scientist, or for that matter, a tech pundit who bases his writings on such dubious benchmarks?
Mac OS X isn't the right tool for every job, but considering that Apple uses Xserves running Mac OS X Server and WebObjects to run the tenth largest Internet website, handle millions of downloads for the world-leading iTunes Music Store, as well as host an array of high volume Apple Stores for public, government, and education sales, it's a bit hard to swallow the idea that Mac OS X is anemic by design and can only deal with a couple concurrent users before collapsing.
But what about Linux having so much more support as a industry standard? This will be considered in:
Unraveling The Mac OS X Linux Kernel Myth Part 3