Am Sonntag, 16. Oktober 2011 schrieb Sthu Deus: > Thank You for Your time and answer, Martin: > >Huh? Well compared to SSD they are dog slow with small size random > >read/write operations - typical refered to as IOPS - but I cannot > >confirm that they are slower than PATA drives. They might not be that > >much faster, but thats IMHO more a question of the physical attributes > >of their mechanics. > > I agree. That's what I want to solve. So for farther discussion let's > just have as initial data: SATA HDD lacks performance on a system and > we are looking for solutions. - Just to avoid the comparison of the > standards. > > >As far as I have seen and measured, SATA drivers in the kernel operate > >almost at hardware speeds. libata is a quite mature and performant > >driver framewark as far as I have seen. > > How do measure? W/ hdparm?
hdparm -tT might give an initial figure. But I suggest it to run it at least three times and look at how the values deviate, as the test time of hdparm -tT is to short. You might want to try dd with conv=fsync or oflag=direct a try as well. Be careful not to overwrite your harddisk ;). For more sophisticated test you can try to use fio with a simple workload. For just copying a big file that would be a mixed read/write workload, but you can start by testing read and write separately. > >That said there might be exceptions with certain boards or > >controllers. So if SATA is measurably slower on your machine than PATA > >it might be an idea to replace the SATA controller or the board with > >something sensible. > > It is not the case for now - what I said was long ago and I have no > more those machines. Why I said that is just to share my experience w/ > S/P -ATA HDDs performance in Linux. And that's it. Ok, then we disregard it unless proved otherwise. > >Exactly. Please start with supplying the information I requested in > >the other mail. > > OK. Please check it. Did so. See there. Ciao, -- Martin 'Helios' Steigerwald - http://www.Lichtvoll.de GPG: 03B0 0D6C 0040 0710 4AFA B82F 991B EAAC A599 84C7 -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/201110171658.27687.mar...@lichtvoll.de
