Separate names with a comma.
Discussion in 'macOS' started by jhu, Sep 2, 2005.
. . . over at anandtech
in theory is an intel/adm vs g5 comparison but it appears to show that MacOSX (and not the G5 processor) idoesn't perform well as a DB server.
This was posted at slashdot, and if you wade through the normal crappy posts and jokes there, there is some good insight. One poster made a point that the tests were really geared towards being a large, single purpose server, and not a general use computer like most macs are used. They contended that the tests were specifically geared towards linux's strengths and Mac OS's weaknesses. I'm not that knowledgeable about these sorts of tests, but I find that idea interesting. I think the important thing is that the tests aren't comprehensive and should be taken with a grain of salt.
That requires a LOT of wading anymore.
The tests may be skewed, but still if YDL runs 8x faster than OS X there must be something broken, especially when we know that the G5 is truly competitive with the best x86 CPUs. Hopefully Apple will tweak OS-X server in the future to try to close the gap.
in general this is true. however, apple does sell the xserve for server purposes. thus they need to address these aspects of macosx.
Not only was YDL faster, it completely mopped the floor with OS X, while running on slower hardware! That is a clear sign that something is terribly broken in OS X.
While some say that these tests tested server-performance, fact remains that Apple does sell Powermac/Xserves for server-tasks. Testing them at server-tasks is easy since both can run exact same software (Samba, MySQL etc.), whereas testing them as workstations would be more difficult since they use different software.
Looks like this discussion didn't go very far - but the Anandtech articles were mentioned in another thread: http://forums.macrumors.com/showthread.php?p=1836969
I agree with the review (and with some comments here), the whole pthread abstraction issue is huge for highly concurrent applications like web and db servers and is laid bare in this article.
I'm not sure how they can really "fix" it either without ditching the Mach kernel - which would be anything but trivial. But then again, I'm an app developer, not one of those über geek OS engineers, so maybe there's some super elegant way they can "fix" it and we'll all be happy!