On 4/3/2013 3:36 PM, Peter Landmann wrote: > I'm very sorry but now i can't confirm anymore that stripe_cache_size helps. > > My test was to short. With every minute the IOPS decrease. So > stripe_cache_size does only help for very short tests. If you're running the tests for multiple minutes and many tens of GBs at a time, then this slowdown is due to garbage collection, not stripe cache sizing. You are not performing proper testing methodologies, and you're jumping to conclusions way too quickly, and likely incorrectly. If you're not familiar with SSD garbage collection then you must learn about it. It will affect everything you do with SSDs, especially when doing these kinds of tests where you're writing huge amounts of data to the flash cells. Wear leveling, part of garbage collection, dramatically slows down SSD throughput. And when you're pushing this much data, TRIM won't help. It'll actually slow the SDDs down even more. -- Stan -- To unsubscribe from this list: send the line "unsubscribe linux-raid" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
