Hi Vasily, It would great if you could point us to your sample code which uses high-res timers. We hope to add delays in the range of 20 - 500 u sec in dm-cache path. Thanks, /Girish BK On 1/22/15, 8:42 PM, "Vasily Tarasov" <tarasov@xxxxxxxxxxx> wrote: >Hi Sourav, > >Delay target provides uniform latency with 1ms granularity. You can >specify different latency for reads or writes. No matter what is the >size of the request or the number of requests submitted - the delay >will remain the same. > >To change granularity to microseconds the code need to be modified to >use high-res timers (instead of regular Linux timers). It is >relatively straight forward modification, I did it once and have a >patch somewhere. Let me see if I can find it. > >HTH, >Vasily > >On Wed, Jan 21, 2015 at 12:15 PM, Sourav Basu <sourav.juit@xxxxxxxxx> >wrote: >> Hello, >> I have few questions regarding dm-delay component. >> 1. Why type of emulations is targeted while implementation of dm-delay? >> 2. Is the delay uniform across different i/o going to the device? >> 3. Can the granularity of be reduced from mili-sec to tens of >>micro-sec? If >> so how? >> >> My aim is to emulate ssd devices latency varying from 10us to 100ms on >> ramdisk using dm-delay. >> >> Regards, >> Sourav >> >> -- >> dm-devel mailing list >> dm-devel@xxxxxxxxxx >> https://www.redhat.com/mailman/listinfo/dm-devel > >-- >dm-devel mailing list >dm-devel@xxxxxxxxxx >https://www.redhat.com/mailman/listinfo/dm-devel -- dm-devel mailing list dm-devel@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/dm-devel
