Messages in this thread | | | Subject | Re: Cheetah vs. UDMA: Bonnie says UDMA is faster! why? | Date | Sat, 28 Mar 1998 08:55:39 -0800 | From | Vaughan Pratt <> |
| |
From: Dave Cinege <dcinege@psychosis.com> >I don't trust bonnie, or any other benchmark at all. The drasitic changes you >saw don't make sense. Could you TELL it was that much of a change? NO? Then it >wasn't. >... >Next the numbers below are just plain wrong and inconsistent >... >Compare those to yours. Bonnie lies! All disk benchmarks are the spawn of >Satan!
Why do you blame all these variations on bonnie? Given the number of factors that influence disk performance, the a priori probability that the variations you're seeing are all bonnie's fault is 1/m^n where m is the number of factors and n is the number of observed discrepancies. The more discrepancies you cite as further evidence for bonnie's dishonesty, the less believable you become, by a factor of m for every additional discrepancy.
Given that bonnie is the same program every time and that you haven't told us what you think bonnie's contribution to the variance is, I'd guess it was even smaller than that.
>first thing to notice is it's make a >difference where on the disks the program is benching because of the change in >rotational speed.
And somehow this makes bonnie a liar? This is one of the many factors that are well known to influence drive performance. Bonnie does not magically compensate for these factors, other than for the artifact of cache hit speedups which it compensates for by writing large amounts. Bonnie reports what it observes. Unless you have some explanation of the variance, what it observes is indicative of performance of the measured system as configured, well known to be highly variable even if bonnie had never been written.
Vaughan Pratt
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |