Re: Performance difference between 2 machines

Noons wrote:
>If so do you agree?

Of course I agree, since anything I would have said would have been based on your previous posts :-)

The bit about I/O including cpu (or other) waits often seems to get lost in online wait analysis postings, even though I'm sure it is explicit in most books that explain the method. I didn't really want to say Sybrand was wrong in immediately blaming the blocksize, since I consider him up there with the intuitive tuners Milsap mentions he used for his method, and we really don't know yet if it is the/an issue. Simply trying a different blocksize may or may show that to be a problem, but your scripts have shown there may be a different problem (which my intuition flagged, as I might immediately have said "throw more memory and more/faster cpu at it" if such pronouncements weren't the staple of laughter these days). There is still an analogue of rotational latency with cached RAID (which your script showed isn't happening with this configuration/test - as you pointed out, likely cache waiting on cpu - but may yet show up under real load), but all that might change with any of the many variables that can affect this stuff. What struck me was the implication that standard DW advice would be wrong.

Besides that, the OP kind of counting test is like asking "is the motorcycle faster than the school bus?" without saying the problem is getting 150 kids to the football game. I certainly wouldn't want to discourage anyone from using the various tools available, I just want people to understand the limits, usage and capabilities of the tools. There was some mention of build time being the real issue...

jg

--
@home.com is bogus.  Dutch keep record!  "What was unsettling was that
the fluid just kept coming.  It was quite a lot of fluid." - Stuart
Claxton, Guiness Book of World Records
http://www.signonsandiego.com/news/nation/20050622-1215-popsicledisaster.html#video