Martin MC Brown wrote: > Hi, > >>>> This is predicated on having data on which to base this decision >>>> though. >>>> >>> >>> >>> If somebody wants me to do any necessary tests with the various >>> MySQL versions and SPARC/intel combinations with 32bit/64bit let me >>> know. >>> >> Yes, if you have any chance of doing that, it would be great! > > > OK, here are the headline results for x86. Tests were doing using > MySQL 5.0.56 (which I happen to be smoke testing at the moment), on > the same machine, running snv_81.
Thanks! Quite useful. Could you share any details on the build. Which compiler did you use, any optimization flags, etc.? It would be useful to have that info. > > Results are using the included sql-bench, standard tests (InnoDB not > included): > > Using 32-bit packages: > > alter-table: Total time: 18 wallclock secs ( 0.05 usr 0.04 sys + 0.00 > cusr 0.00 csys = 0.09 CPU) > ATIS: Total time: 20 wallclock secs ( 2.94 usr 0.40 sys + 0.00 cusr > 0.00 csys = 3.34 CPU) > big-tables: Total time: 14 wallclock secs ( 3.43 usr 0.25 sys + 0.00 > cusr 0.00 csys = 3.68 CPU) > connect: Total time: 134 wallclock secs (29.68 usr 23.16 sys + 0.00 > cusr 0.00 csys = 52.84 CPU) > create: Total time: 348 wallclock secs ( 2.87 usr 1.88 sys + 0.00 cusr > 0.00 csys = 4.75 CPU) > insert: Total time: 1038 wallclock secs (217.49 usr 77.17 sys + 0.00 > cusr 0.00 csys = 294.66 CPU) > select: Total time: 399 wallclock secs (19.76 usr 4.95 sys + 0.00 cusr > 0.00 csys = 24.71 CPU) > transactions: Test skipped because the database doesn't support > transactions > wisconsin: Total time: 10 wallclock secs ( 1.70 usr 0.95 sys + 0.00 > cusr 0.00 csys = 2.65 CPU) > > Using 64-bit packages: > > alter-table: Total time: 15 wallclock secs ( 0.05 usr 0.04 sys + 0.00 > cusr 0.00 csys = 0.09 CPU) > ATIS: Total time: 17 wallclock secs ( 2.95 usr 0.42 sys + 0.00 cusr > 0.00 csys = 3.37 CPU) > big-tables: Total time: 11 wallclock secs ( 3.43 usr 0.23 sys + 0.00 > cusr 0.00 csys = 3.66 CPU) > connect: Total time: 121 wallclock secs (29.69 usr 24.75 sys + 0.00 > cusr 0.00 csys = 54.44 CPU) > create: Total time: 348 wallclock secs ( 2.87 usr 1.94 sys + 0.00 cusr > 0.00 csys = 4.81 CPU) > insert: Total time: 885 wallclock secs (215.78 usr 74.40 sys + 0.00 > cusr 0.00 csys = 290.18 CPU) > select: Total time: 257 wallclock secs (19.59 usr 5.06 sys + 0.00 cusr > 0.00 csys = 24.65 CPU) > transactions: Test skipped because the database doesn't support > transactions > wisconsin: Total time: 8 wallclock secs ( 1.67 usr 0.93 sys + 0.00 > cusr 0.00 csys = 2.60 CPU) > > This shows a moderate increase in speed for insert (15%) and select > (35%), and smaller increases for some other tests. > > I'll be doing longer testing of this over the weekend to see if I can > repeat this level of increase over a larger number of iterations. > > SPARC is showing similar output, but I had an unrelated problem with > the box I was running the tests on so I've had to restart. > > I'll also be running dbt2 tests, where I expect to get pretty similar > increase numbers. > > MC > > -- > Martin 'MC' Brown, mc at mcslp.com > Everything MCslp: http://planet.mcslp.com > > -- Matt Ingenthron - Web Infrastructure Solutions Architect Sun Microsystems, Inc. - Global Systems Practice http://blogs.sun.com/mingenthron/ email: matt.ingenthron at sun.com Phone: 310-242-6439
