From owner-freebsd-smp Sat Nov 6 12:39:29 1999 Delivered-To: freebsd-smp@freebsd.org Received: from verdi.nethelp.no (verdi.nethelp.no [158.36.41.162]) by hub.freebsd.org (Postfix) with SMTP id D1F9314CE7 for ; Sat, 6 Nov 1999 12:39:15 -0800 (PST) (envelope-from sthaug@nethelp.no) Received: (qmail 98449 invoked by uid 1001); 6 Nov 1999 20:39:13 +0000 (GMT) To: darrylo@sr.hp.com Cc: mwm@phone.net, freebsd-smp@FreeBSD.ORG Subject: Re: Dual Celeron + FreeBSD? From: sthaug@nethelp.no In-Reply-To: Your message of "Fri, 05 Nov 1999 17:26:31 -0800" References: <199911060126.RAA14109@mina.sr.hp.com> X-Mailer: Mew version 1.05+ on Emacs 19.34.2 Mime-Version: 1.0 Content-Type: Text/Plain; charset=us-ascii Date: Sat, 06 Nov 1999 21:39:13 +0100 Message-ID: <98447.941920753@verdi.nethelp.no> Sender: owner-freebsd-smp@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org > > ;->With your typical > > ;->1- or 2-drive system, IDE drives shouldn't be significantly slower than > > ;->SCSI. > > > > My tests show different - see > http://www.phone.net/home/mwm/disktest/disktest.html >. If there's > > something I can do to improve the performance of the IDE drive, I'd > > love to hear it. > > Did you explicitly configure your kernel to enable DMA? I couldn't > find any mention of that on your page. I'd say that the picture is not at all clearcut any more. My tests indicate that SCSI drives still have an edge on EIDE drives for random access, but it's not huge. The CPU usage is lower for SCSI in some cases, but not all. Below are some bonnie *and* rawio measurements, with comments. After the first few tests of bonnie with 100 MB files I stopped doing this test, since it was obvious that the memory size of the machine had a great effect - even on the seek transactions. I stuck to 1000 MB test afterwards - this is greater than the memory on all the machines I tried. Steinar Haug, Nethelp consulting, sthaug@nethelp.no ---------------------------------------------------------------------- 1. Dual Celeron-366, 66 MHz system bus, 256 MB memory, IBM DPTA-372730 (EIDE), 3.3-STABLE. Typical wall clock time: 33 s for 100 MB, 423 s for 1000 MB On the 100 MB measurements you can clearly see the unrealistic figures for number of seeks per second, and also the effect of the buffer cache on sequential input. The 1000 MB case shows much more realistic figures. The DPTA-372730 is one darn fast disk! -------Sequential Output-------- ---Sequential Input-- --Random-- -Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks--- MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 100 17419 99.2 15785 37.6 16355 44.8 19008 100.0 128068 100.0 1691.1 33.3 100 17482 99.0 15952 38.5 16474 46.5 19025 100.0 127537 100.0 1914.6 38.0 100 17481 99.3 16687 40.7 16396 46.2 19030 100.0 127944 100.0 1866.8 36.7 100 17564 98.5 15868 37.3 16373 45.8 19053 100.0 126797 100.0 1952.8 38.7 100 17580 99.7 15995 38.9 15517 58.3 19044 100.0 127725 100.0 1897.8 36.5 1000 17359 96.4 15486 34.9 6589 16.3 16694 97.9 22465 30.9 130.3 2.9 1000 16709 94.6 15619 34.4 6929 17.0 16694 97.8 22429 30.8 126.9 2.9 1000 16641 94.5 15631 34.9 6707 16.6 16676 97.7 22470 30.9 126.7 2.8 1000 17397 96.3 15513 35.2 6612 16.1 16694 97.9 22446 30.9 129.3 2.9 1000 17378 96.4 15558 35.0 6565 16.2 16680 97.8 22458 30.9 126.9 2.9 2. Same except single Celeron-366 (kernel config without SMP). Notice that the CPU figures are significantly lower for the single CPU case. 1000 17288 88.8 15509 27.8 6526 11.4 17951 96.6 22389 20.6 121.6 1.8 1000 17428 90.2 15642 27.4 6496 11.2 17949 96.9 22402 20.6 123.3 1.9 1000 16872 86.2 15666 27.4 6961 12.1 17938 96.7 22424 20.7 124.2 1.8 1000 17480 89.8 15691 28.1 6487 11.2 17975 96.8 22418 20.8 124.4 1.8 1000 17404 89.0 15580 27.6 6505 11.2 17954 96.8 22411 20.7 124.9 1.9 3. Dual Celeron-366, 66 MHz system bus, 256 MB memory, IBM DPTA-372730 (EIDE), 4.0-CURRENT as of 29. October, new ATA driver. Typical wall clock time 453 s for 1000 MB (30 s longer than 3.3 with old wd driver). But the block output figures are better :-) 1000 16430 96.1 17374 45.0 5547 16.1 15941 95.3 22465 35.6 123.8 3.1 1000 16040 94.1 17447 44.7 5557 15.7 15796 95.2 22452 35.5 122.9 3.0 1000 16374 95.4 17508 45.6 5566 15.9 15804 95.0 22434 35.6 121.8 3.0 1000 15724 95.4 17753 43.5 5558 15.7 15849 95.6 22408 35.2 121.3 3.1 1000 16415 96.0 17428 44.3 5561 15.7 15855 95.5 22451 35.5 122.1 3.0 4. Same except single Celeron-366. 1000 17496 93.0 17420 34.0 5603 11.7 17290 94.0 22465 22.7 120.1 2.0 1000 17330 91.7 17466 33.4 5572 11.8 17302 94.1 22467 22.9 121.5 2.0 1000 17388 91.3 17554 32.0 5564 11.4 17145 93.4 22375 22.4 122.5 2.1 1000 17239 92.1 17486 31.8 5565 11.3 17300 94.4 22442 22.5 121.6 2.1 1000 17492 90.7 17627 31.5 5575 10.9 17251 93.9 22462 22.5 121.4 2.0 1000 17756 91.8 17544 33.5 5557 11.7 17219 93.7 22443 23.1 122.3 1.9 5. Single PIII-450, 100 MHz system bus, 256 MB memory, IBM DNES-318350W (SCSI), 3.3-STABLE. Typical wall clock time: 600 s for 1000 MB. 1000 17172 75.6 12696 19.3 3240 5.6 18585 84.9 18632 16.5 136.7 1.7 1000 17584 76.8 12718 19.2 3241 5.6 18610 85.0 18655 16.6 137.2 1.7 1000 17527 76.9 12709 19.4 3242 5.6 18560 84.6 18635 16.5 137.1 1.8 1000 17369 76.1 12805 19.6 3242 5.6 18602 84.8 18657 16.5 138.2 1.8 1000 17583 77.0 12723 19.4 3244 5.6 18577 85.1 18621 16.5 135.4 1.7 6. Dual PII-400, 100 MHz system bus, 512 MB memory, Seagate ST39102LC (SCSI), FreeBSD 3.2. System lightly loaded. Typical wall clock time: 510 s for 1000 MB. 1000 16668 76.0 10888 14.6 5277 11.0 15049 75.5 17213 20.9 155.8 2.6 1000 15030 68.1 10816 14.5 5260 11.1 14911 74.7 16832 20.6 156.7 2.5 1000 14663 66.6 10966 14.8 5281 11.0 14959 75.0 17213 21.0 155.5 2.6 1000 14361 65.0 10779 14.4 5250 11.0 15455 77.4 17070 20.8 155.5 2.4 1000 16052 73.0 10492 14.0 5498 11.5 15721 78.5 17481 21.2 157.1 2.6 Same configurations (numbers 1 to 6 above), but rawio instead of bonnie. 1. Dual Celeron-366, DPTA-372730, 3.3-STABLE. Test ID K/sec /sec %User %Sys %Total RR anon 1942.9 121 0.0 0.7 0.8 16384 SR anon 28884.8 1763 0.1 9.7 9.8 16384 RW anon 1755.8 109 0.1 0.7 0.7 16384 SW anon 14389.8 878 0.3 4.7 4.9 16384 RR anon 1952.5 121 0.0 0.8 0.8 16384 SR anon 28879.5 1763 0.6 9.3 9.9 16384 RW anon 1759.3 109 0.1 0.7 0.7 16384 SW anon 14340.2 875 0.2 4.6 4.7 16384 RR anon 1959.9 121 0.0 0.7 0.8 16384 SR anon 28884.4 1763 0.7 9.1 9.8 16384 RW anon 1760.6 109 0.0 0.7 0.7 16384 SW anon 14383.6 878 0.4 4.6 4.9 16384 RR anon 1959.8 121 0.1 0.7 0.8 16384 SR anon 28877.2 1763 0.5 9.4 9.9 16384 RW anon 1766.6 109 0.0 0.7 0.7 16384 SW anon 14314.6 874 0.1 4.7 4.8 16384 RR anon 1950.1 121 0.0 0.8 0.8 16384 SR anon 28884.1 1763 0.5 9.4 9.9 16384 RW anon 1762.5 109 0.1 0.7 0.7 16384 SW anon 14420.5 880 0.2 4.7 4.9 16384 2. Same except single Celeron-366. Test ID K/sec /sec %User %Sys %Total RR anon 1951.0 121 0.1 0.3 0.4 16384 SR anon 29397.9 1794 0.6 3.6 4.3 16384 RW anon 1756.5 110 0.1 0.3 0.3 16384 SW anon 14385.8 878 0.2 1.9 2.1 16384 RR anon 1974.6 122 0.1 0.3 0.4 16384 SR anon 29383.2 1793 0.3 4.0 4.3 16384 RW anon 1746.3 109 0.1 0.3 0.3 16384 SW anon 14368.7 877 0.2 1.8 2.0 16384 RR anon 1958.3 121 0.1 0.3 0.4 16384 SR anon 29395.5 1794 0.3 4.0 4.4 16384 RW anon 1760.8 110 0.1 0.3 0.3 16384 SW anon 14430.1 881 0.3 1.9 2.2 16384 RR anon 1966.8 122 0.1 0.3 0.4 16384 SR anon 29399.7 1794 0.3 3.9 4.2 16384 RW anon 1748.1 109 0.1 0.3 0.3 16384 SW anon 14354.1 876 0.3 1.8 2.2 16384 RR anon 1960.4 121 0.1 0.3 0.4 16384 SR anon 29376.8 1793 0.3 3.9 4.3 16384 RW anon 1752.0 109 0.1 0.3 0.3 16384 SW anon 14368.7 877 0.2 1.9 2.1 16384 3. Dual Celeron-366, DPTA-372730, 4.0-CURRENT. Test ID K/sec /sec %User %Sys %Total RR anon 1736.4 108 0.0 0.7 0.7 16384 SR anon 29836.4 1821 0.3 10.1 10.3 16384 RW anon 1556.6 97 0.0 0.6 0.7 16384 SW anon 14377.5 878 0.1 4.9 5.0 16384 RR anon 1714.8 107 0.1 0.6 0.7 16384 SR anon 29834.9 1821 0.3 10.0 10.4 16384 RW anon 1552.6 97 0.0 0.6 0.7 16384 SW anon 14388.3 878 0.1 4.8 4.9 16384 RR anon 1724.0 107 0.1 0.7 0.7 16384 SR anon 29828.9 1821 0.6 9.7 10.3 16384 RW anon 1556.4 97 0.0 0.6 0.7 16384 SW anon 14279.2 872 0.1 4.8 4.9 16384 RR anon 1716.7 107 0.1 0.7 0.7 16384 SR anon 29841.8 1821 1.2 9.1 10.3 16384 RW anon 1554.6 97 0.0 0.6 0.7 16384 SW anon 14323.6 874 0.3 4.6 4.9 16384 RR anon 1730.9 107 0.1 0.6 0.7 16384 SR anon 29842.9 1821 0.2 10.2 10.4 16384 RW anon 1554.1 97 0.0 0.6 0.7 16384 SW anon 14366.9 877 0.2 4.7 4.9 16384 4. Same except single Celeron-366. Test ID K/sec /sec %User %Sys %Total RR anon 1732.9 107 0.1 0.3 0.3 16384 SR anon 30428.4 1857 0.4 3.7 4.1 16384 RW anon 1570.2 97 0.1 0.2 0.3 16384 SW anon 14424.8 880 0.2 1.8 2.0 16384 RR anon 1712.8 107 0.1 0.2 0.3 16384 SR anon 30433.4 1858 0.3 3.8 4.1 16384 RW anon 1565.1 97 0.1 0.2 0.3 16384 SW anon 14404.8 879 0.2 1.8 2.0 16384 RR anon 1722.8 107 0.1 0.3 0.3 16384 SR anon 30418.9 1857 0.6 3.5 4.1 16384 RW anon 1557.7 96 0.1 0.2 0.3 16384 SW anon 14452.4 882 0.1 1.9 2.0 16384 RR anon 1730.2 107 0.1 0.3 0.3 16384 SR anon 30411.4 1856 0.6 3.5 4.1 16384 RW anon 1566.8 97 0.1 0.2 0.3 16384 SW anon 14432.1 881 0.1 1.9 2.0 16384 RR anon 1724.6 107 0.1 0.3 0.3 16384 SR anon 30403.2 1856 0.6 3.4 4.1 16384 RW anon 1558.4 97 0.1 0.2 0.3 16384 SW anon 14437.8 881 0.2 1.8 2.0 16384 5. Single PIII-450, DNES-318350W, 3.3-STABLE. Notice the sequential write figures look suspiciously low. I tried reducing the number of tags, but never got anywhere near expected figures. Test ID K/sec /sec %User %Sys %Total RR anon 2784.7 173 0.1 0.6 0.7 16384 SR anon 25155.2 1535 0.1 5.3 5.4 16384 RW anon 2325.6 145 0.1 0.5 0.6 16384 SW anon 2601.2 159 0.0 0.5 0.6 16384 RR anon 2796.1 173 0.1 0.6 0.7 16384 SR anon 25152.8 1535 0.2 5.2 5.5 16384 RW anon 2323.5 144 0.0 0.5 0.6 16384 SW anon 2600.7 159 0.0 0.5 0.6 16384 RR anon 2799.0 172 0.1 0.6 0.7 16384 SR anon 25159.5 1536 0.2 5.3 5.5 16384 RW anon 2333.0 144 0.1 0.5 0.6 16384 SW anon 2235.6 136 0.0 0.5 0.5 16384 (same but tags reduced from 64 to 32) RR anon 2796.9 173 0.1 0.6 0.7 16384 SR anon 25158.9 1536 0.3 5.2 5.5 16384 RW anon 2321.0 144 0.1 0.5 0.6 16384 SW anon 3669.7 224 0.1 0.7 0.8 16384 (same but tags reduced from 32 to 16) RR anon 2796.8 173 0.1 0.6 0.7 16384 SR anon 25029.7 1528 0.4 5.0 5.5 16384 RW anon 2345.9 144 0.0 0.5 0.6 16384 SW anon 2599.4 159 0.0 0.5 0.6 16384 (same but tags reduced from 16 to 8) RR anon 2775.0 172 0.1 0.6 0.7 16384 SR anon 25188.7 1537 0.6 4.8 5.4 16384 RW anon 2321.8 144 0.1 0.5 0.6 16384 SW anon 2601.9 159 0.0 0.6 0.6 16384 (same but tags disabled) RR anon 2039.0 127 0.1 0.3 0.4 16384 SR anon 46742.3 2853 0.7 6.3 7.0 16384 RW anon 2139.8 132 0.1 0.3 0.4 16384 SW anon 2235.5 136 0.0 0.3 0.3 16384 6. Dual PII-400, Seagate ST39102LC, FreeBSD 3.2. Test ID K/sec /sec %User %Sys %Total RR anon 3420.8 212 0.1 1.0 1.1 16384 SR anon 9201.2 562 0.1 2.3 2.3 16384 RW anon 3322.9 206 0.1 0.9 0.9 16384 SW anon 4355.2 266 0.1 0.9 1.0 16384 RR anon 3425.8 211 0.2 0.9 1.1 16384 SR anon 9238.3 564 0.1 2.2 2.3 16384 RW anon 3332.9 205 0.1 0.8 0.9 16384 SW anon 4146.8 253 0.1 0.9 1.0 16384 RR anon 3411.1 212 0.1 0.9 1.0 16384 SR anon 9227.5 563 0.2 2.1 2.3 16384 RW anon 3315.8 205 0.1 0.8 0.9 16384 SW anon 4267.2 260 0.0 1.0 1.0 16384 RR anon 3365.2 210 0.2 0.9 1.0 16384 SR anon 8909.0 544 0.1 2.1 2.2 16384 RW anon 3282.8 205 0.1 0.8 0.9 16384 SW anon 4347.1 265 0.1 1.0 1.1 16384 RR anon 3409.7 212 0.2 0.8 1.0 16384 SR anon 9480.2 579 0.1 2.2 2.3 16384 RW anon 3309.2 206 0.1 0.8 0.9 16384 SW anon 4184.4 255 0.1 0.9 1.0 16384 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-smp" in the body of the message