Hi Brent, thanks for your comments! An unbiased outside pov always helps!
On Thu, 6 Jan 2005 14:32:16 -0500, Brent Baisley <[EMAIL PROTECTED]> wrote: > While part of the problem may be OS X, Apple is still optimizing parts > of the OS, I would say the "problem" is that you are comparing it to an > aging PIII. Some people have gotten better performance from a PIII than > a P4. The reason is cache. The PIII has a larger cache, MySQL loves > cache. The G5 XServe have a comparatively small cache (512k L2, 1MB L3 > I think). If you have a G4 XServe around, I would try that. You may get > better performance. The G4 XServe came with 2MB L3 cache. Your best > hardware may actually be Intel's Extreme Edition CPU (4MB cache) > designed for gaming. Although I haven't seen any benchmarks, just my > speculation. Actually the G5 has no L3 cache and 512k of L2. The Coppermines PIII I was comparing to has 256k of L2 cache and also no L3. I don't have any G4s available to compare to unfortunately. > > Another possibility is your disks. Did the PIII have SCSI? SCSI uses a > technique called command queueing to optimize reads and writes. Command > queueing has only recently become available on some ATA drives and > Apple does not ship SATA drives with command queueing in their XServes. > If your old system had SCSI disks, you could try moving them over to > the XServe if you have a SCSI card in the XServe. Or replace the SATA > drives in the XServe with SATA drives with command queueing. True, the PIII does use SCSI disks, but it doesn't seem that the XServe is disk bound. > The G5 is excellent for compute intensive tasks, databases are > typically throughput intensive. If you can't take advantage of the > vector processor in the G5 (and G4), you're missing a big part of where > the G5 gets it's performance. You may try compiling a version of MySQL > yourself, optimized for the G5 chip. > http://developer.apple.com/technotes/tn/tn2086.html#G5options > This thread from the archive may help in compiling: > http://archives.neohapsis.com/archives/mysql/2004-q2/0759.html Thanks for that pointer. I've compiled a G5 optimized copy of 4.1.8a and it definitely makes a subtle improvement to performance. > But besides all that, you should first determine what's causing the > bottleneck. It's either disk (I/O), CPU, RAM (not enough allocated), or > Network. Still trying to get a clear sense of this but I'm thinking it's CPU. Thanks again! scott > On Jan 6, 2005, at 1:58 PM, Scott Wilson wrote: > > > Hello, > > > > I'm interested to hear peoples' experiences running mysql on OS X. > > I've moved the database for a fairly heaily used website (~ 2M queries > > a day) over to a new dual 2GHz XServe running OS X Server 10.3.7. > > This database has run smoothly on an aging dual PIII machine running > > freebsd for the past several years. > > > > My initial impression is that the performance gains aren't nearly what > > I would have expected. For the most part the new machine is less > > loaded, but at peak times it's arguably doing worse that it did the > > old freebsd machine. > > > > The number a variables involved has hindered my creating comprehensive > > benchmarks but some initial impressions from running stock mysql > > benchmarks are that 4.0.23a on OS X performs around 10% faster than > > 4.1.8a and that my old freebsd machine running 4.0.18 is less than a > > factor of two slower. These are all using similar my.cnf settings > > tuned along the lines of the my-huge.cnf sample config. > > > > Does anyone have any tips to offer for tuning OS X and mysql to play > > well together? Is anyone running a heavily loaded mysql server in > > production under OS X? > > > > Thanks for you help! > > > > scott > > > > -- > > MySQL General Mailing List > > For list archives: http://lists.mysql.com/mysql > > To unsubscribe: > > http://lists.mysql.com/[EMAIL PROTECTED] > > > > > -- > Brent Baisley > Systems Architect > Landover Associates, Inc. > Search & Advisory Services for Advanced Technology Environments > p: 212.759.6400/800.759.0577 > > -- MySQL General Mailing List For list archives: http://lists.mysql.com/mysql To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
