![]() | |||||||||||||
Messages in this thread |
On Mon, Mar 29 at 13:36, Pavel Machek wrote: >32MB is one second if everything goes okay... That will be horrible for latency, right? It'll be somewhere between half a second and 1.5 seconds, depending on how old your hard drive is, all other things being removed from the equation. The media rate of modern IDE drives is about 60MB/s right now, and going up slightly soon. The top SCSI drives are around 85MB/s if I remember correctly. On average, 1s is about right. It becomes a tradeoff between a big hit now and a lot of little hits over time. If you don't expect to have any idle time soon, you're probably better off flushing as big a chunk as possible to free RAM for the upcoming/continuous workload. If you think you might have idle time, the lower MB/s of smaller requests will keep you more granular and improve latency, even though overall work completed per unit time will be slightly lower. For a bursty access pattern, you're probably best off with small requests that fit within the drive's cache, but I don't have any sort of measurable data on this. Most of the OS testing that we're subjected to doesn't attempt large requests. --eric -- Eric D. Mudama edmudama@mail.bounceswoosh.org - 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 [W:0.447 / U:0.040 seconds] ©2003-2008 Jasper Spaans | |||||||||||||