On Sunday 04 March 2007 18:45, Con Kolivas wrote: > On Sunday 04 March 2007 18:00, Con Kolivas wrote: > > This message is to announce the first general public release of the > > "Rotating Staircase DeadLine" cpu scheduler. > > > > Based on previous work from the staircase cpu scheduler I set out to > > design, from scratch, a new scheduling policy design which satisfies > > every requirement for SCHED_NORMAL (otherwise known as SCHED_OTHER) task > > management. > > > > Available for download are: > > > > A full rollup of the patch for 2.6.20: > > http://ck.kolivas.org/patches/staircase-deadline/sched-rsdl-0.26.patch > > It's probably worth mentioning that this scheduler shows a not > insignificant improvement in the mysql test case that recently received a > lot of publicity. Why exactly that's the case I'm not sure but it may help > track down what is actually responsible for the performance drop off as > well. Note that the SMP balancing of this cpu scheduler is essentially > unchanged from the mainline one.
Only preliminary other benchmarking has been done so far on RSDL, and so far the results are equal to or slightly better than mainline on scalability grounds. These are the sysbench graphs just for completion that Nishanth Aravamudan performed. We were actually trying to track down the cause of the mysql performance problem and I suggested trying different cpu schedulers as well. Needless to say I'm not unhappy with the results although they haven't told us exactly where the problem lies but it certainly does add evidence that perhaps it is a scheduling issue. -- -ck
transactions.png
Description: PNG image
idle.png
Description: PNG image
