Date: Wed, 5 Aug 2015 14:59:55 +0100 From: Steven Hartland <steven.hartland@multiplay.co.uk> To: Alexander Motin <mav@FreeBSD.org>, src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r286320 - head/sys/cam/ctl Message-ID: <55C216DB.10703@multiplay.co.uk> In-Reply-To: <201508051346.t75DkFrY042506@repo.freebsd.org> References: <201508051346.t75DkFrY042506@repo.freebsd.org>
next in thread | previous in thread | raw e-mail | index | archive | help
Nice!! On 05/08/2015 14:46, Alexander Motin wrote: > Author: mav > Date: Wed Aug 5 13:46:15 2015 > New Revision: 286320 > URL: https://svnweb.freebsd.org/changeset/base/286320 > > Log: > Issue all reads of single XCOPY segment simultaneously. > > During vMotion and Clone VMware by default runs multiple sequential 4MB > XCOPY requests same time. If CTL issues reads sequentially in 1MB chunks > for each XCOPY command, reads from different commands are not detected > as sequential by serseq option code and allowed to execute simultaneously. > Such read pattern confused ZFS prefetcher, causing suboptimal disk access. > Issuing all reads same time make serseq code work properly, serializing > reads both within each XCOPY command and between them. > > My tests with ZFS pool of 14 disks in RAID10 shows prefetcher efficiency > improved from 37% to 99.7%, copying speed improved by 10-60%, average > read latency reduced twice on HDD layer and by five times on zvol layer. > > MFC after: 2 weeks > Sponsored by: iXsystems, Inc. > > Modified: > head/sys/cam/ctl/ctl_tpc.c > > Modified: head/sys/cam/ctl/ctl_tpc.c > ============================================================================== > --- head/sys/cam/ctl/ctl_tpc.c Wed Aug 5 13:10:13 2015 (r286319) > +++ head/sys/cam/ctl/ctl_tpc.c Wed Aug 5 13:46:15 2015 (r286320) > @@ -817,7 +817,7 @@ tpc_process_b2b(struct tpc_list *list) > struct scsi_ec_segment_b2b *seg; > struct scsi_ec_cscd_dtsp *sdstp, *ddstp; > struct tpc_io *tior, *tiow; > - struct runl run, *prun; > + struct runl run; > uint64_t sl, dl; > off_t srclba, dstlba, numbytes, donebytes, roundbytes; > int numlba; > @@ -889,8 +889,7 @@ tpc_process_b2b(struct tpc_list *list) > list->segsectors = numbytes / dstblock; > donebytes = 0; > TAILQ_INIT(&run); > - prun = &run; > - list->tbdio = 1; > + list->tbdio = 0; > while (donebytes < numbytes) { > roundbytes = numbytes - donebytes; > if (roundbytes > TPC_MAX_IO_SIZE) { > @@ -942,8 +941,8 @@ tpc_process_b2b(struct tpc_list *list) > tiow->io->io_hdr.ctl_private[CTL_PRIV_FRONTEND].ptr = tiow; > > TAILQ_INSERT_TAIL(&tior->run, tiow, rlinks); > - TAILQ_INSERT_TAIL(prun, tior, rlinks); > - prun = &tior->run; > + TAILQ_INSERT_TAIL(&run, tior, rlinks); > + list->tbdio++; > donebytes += roundbytes; > srclba += roundbytes / srcblock; > dstlba += roundbytes / dstblock; >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?55C216DB.10703>