From owner-freebsd-current Tue May 7 06:33:59 1996 Return-Path: owner-current Received: (from root@localhost) by freefall.freebsd.org (8.7.3/8.7.3) id GAA06794 for current-outgoing; Tue, 7 May 1996 06:33:59 -0700 (PDT) Received: from paloalto.access.hp.com (daemon@paloalto.access.hp.com [15.254.56.2]) by freefall.freebsd.org (8.7.3/8.7.3) with ESMTP id GAA06775 Tue, 7 May 1996 06:33:55 -0700 (PDT) Received: from fakir.india.hp.com by paloalto.access.hp.com with ESMTP (1.37.109.16/15.5+ECS 3.3) id AA186956012; Tue, 7 May 1996 06:33:38 -0700 Received: from localhost by fakir.india.hp.com with SMTP (1.37.109.16/15.5+ECS 3.3) id AA138846231; Tue, 7 May 1996 19:07:12 +0530 Message-Id: <199605071337.AA138846231@fakir.india.hp.com> To: "Andrew V. Stesin" Cc: hackers@freebsd.org, current@freebsd.org Subject: Re: lmbench IDE anomaly In-Reply-To: Your message of "Sat, 04 May 1996 13:55:44 +0300." <199605041055.NAA23197@office.elvisti.kiev.ua> Date: Tue, 07 May 1996 19:07:11 +0530 From: A JOSEPH KOSHY Sender: owner-current@freebsd.org X-Loop: FreeBSD.org Precedence: bulk Following Andrew Stesin's suggestion I enabled flags 0x80ff80ff on the onboard IDE controller. IDE Disk transfer figures went up dramatically; but the slowdown on simultaneous reads was still around. Here are the figures: (Machine configuration at end) (The test cases involved running "lmdd" from LMBENCH on the various disk devices and timing reads of 16MB of data. i.e# ./lmdd if=/dev/DEVICE bs=BLOCKSIZE count=16MEG/BLOCKSIZE of=internal Throughput for one lmdd reader process and two simultaneous lmdd readers are given below). Per process Per process Device blocksize KB/s blocksize KB/s ~~~~~~ ~~~~~~~~~ ~~~~ ~~~~~~~~~ ~~~~ -- SCSI DISK -- --single reader-- rsd0a bs=1024 653.74 bs=8192 1312.31 682.66 1268.36 677.52 1361.95 --two readers-- rsd0a bs=1024 424.27 bs=8192 805.69 424.24 807.64 -- 812.82 Looks like changing the block size for the read can double throughput. Also, two readers yield better thoughput than a single reader process. So far so good. -- IDE DISK -- --single reader-- rwd0a bs=1024 839.05 bs=8192 2392.08 (!!) 841.53 2402.42 (!!) 841.85 2402.45 (!!) --two readers-- rwd0a bs=1024 199.38 bs=8192 251.83 218.38 237.95 220.68 238.50 The read rates for the single reader case are fantastic, however disaster seems to strike when two reader access the same device So I looked at the block device. --single reader-- wd0a bs=1024 199.80 bs=8192 796.07 200.04 795.06 --two readers-- wd0a bs=1024 200.04 bs=8192 795.60 200.33 795.20 Hmm, block size makes a huge difference still. Is this to be expected? Also the two reader case and the single reader case are around the same performance -- i.e. the buffer cache seems to be working well. Also note the 3x-4x slowdown when enabling the buffer cache compared to the raw device read. Machine config ~~~~~~~~~~~~~~ FreeBSD 2.2-CURRENT #0: Mon May 6 12:16:33 IST 1996 root@krill.india.hp.com:/usr/src/sys/compile/KRILL ... CPU: Pentium (89.99-MHz 586-class CPU) Origin = "GenuineIntel" Id = 0x525 Stepping=5 Features=0x1bf real memory = 16777216 (16384K bytes) avail memory = 14737408 (14392K bytes) ... wdc0 at 0x1f0-0x1f7 irq 14 flags 0x80ff80ff on isa wdc0: unit 0 (wd0): , multi-block-8 wd0: 516MB (1057392 sectors), 1049 cyls, 16 heads, 63 S/T, 512 B/S aha0: Rev 41 (AHA-154x[AB]) V0.5, enabling residuals, target ops aha0: reading board settings, dma=5 int=11 id=7 (bus speed defaulted) aha0 at 0x330-0x333 irq 11 drq 5 on isa (aha0:5:0): "QUANTUM LPS1080S 1220" type 0 fixed SCSI 2 sd0(aha0:5:0): Direct-Access 1001MB (2051460 512 byte sectors) sd0(aha0:5:0): with 2874 cyls, 8 heads, and an average 89 sectors/track ... Koshy