![]() | |||||||||||||
Messages in this thread |
On Monday 29 March 2004 06:29, Wim Coekaerts wrote: > > In 2.4 reaching 512k DMA units that helped a lot, but going past 512k > > didn't help in my measurements. 1M maybe these days is needed (as Jens > > suggested) but >1M still sounds overkill and I completely agree with > > Jens about that. > > at least 1Mb... more than 1mb (I doubt 32mb is really necessarily > useful) is nice for flushing contiguous journal data to disk. between > 1-8mb in one io. > > at least 1mb is good when you have to process massive amounts of data, > just read huge chunks of files, we tend to do 1mb. anyway, > as you said, at some point it's a bit overkill . I thinkg 1-8mb makes > sense. It makes sense *today*. Do you want to return to this discussion ad infinitum? Andrea is right, block layer can be coded to detect such a (per-device) request size increasing which does not give you any more throughput, or increasing which hurts your latency too much, and stick with it. More detailed: 1. size = 1 sector 2 measure throughput and latency 3. size *= 2 4. measure new_throughput and new_latency 5. if(new_throughput<=throughput*1.01 || new_latency>50ms) break 6. throughput = new_throughput; latency = new_latency; goto 3 Real code will have admin overrides (start from N sectors, never go higher than M sectors etc), but you got the idea. -- vda - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ | ||||||||||||
| Last update: 2005-03-22 13:02 [from the cache] ©2003-2008 | |||||||||||||