On Thu, 2003-06-19 at 03:48, Andreas Tille wrote: > On 18 Jun 2003, Steve Pacenka wrote: > > > > 2.4.21: > > > ehec2,496M,3316,98,13866,27,5968,13,2856,92,12546,13,123.9,1,16,261,97, > ^^^^^ > > +++++,+++,13110,100,270,97,+++++,+++,1658,96 > > > Some comparative bonnie++ results ... > > > > 2.4.21 from Debian, Ultra 10 333 CPU, 512M RAM, mdma2 IDE 7200 RPM, > > 2M buffer > > > > 1G,3363,98,20275,38,5936,14,3148,95,14885,16,174.8,1,16,279,98, > ^^^^^ > > +++++,+++,13568,100,289,99,+++++,+++,1666,96 > Well, the "Sequential Output --Block--" (marked with ^^^^) seems to be the > only noticeable difference by about a factor two. I don't know whether my > feeling is right that it is connected to the factor of two regarding RAM.
Data density difference on a track? From its name, this test would be writing a block much larger than the on-drive buffer and should be limited by the rate at which the drive can write to whole tracks on a platter, and seek between nearby tracks. More data per track = fewer seeks. Another thing that could be limiting might be file fragmentation on the drive. The partition I used has had little written to it ever. > but I hoped that we could find some use for the Ultra 10 ... My U10 does file, IMAP, apache, and print server duty most of the time. It provides a decently responsive 1600x1200 15 bit GUI when needed. These things must have been ~$5K engineering or financial workstations originally. -- SP