From owner-freebsd-current Mon Mar 6 09:04:48 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id JAA11537 for current-outgoing; Mon, 6 Mar 1995 09:04:48 -0800 Received: from shell1.best.com (shell1.best.com [204.156.128.10]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id JAA11529 for ; Mon, 6 Mar 1995 09:04:47 -0800 Received: from geli.clusternet (rcarter.vip.best.com [204.156.137.2]) by shell1.best.com (8.6.10/8.6.5) with ESMTP id JAA10559; Mon, 6 Mar 1995 09:03:49 -0801 Received: (from rcarter@localhost) by geli.clusternet (8.6.10/8.6.9) id JAA21908; Mon, 6 Mar 1995 09:02:49 -0800 Date: Mon, 6 Mar 1995 09:02:49 -0800 From: "Russell L. Carter" Message-Id: <199503061702.JAA21908@geli.clusternet> To: se@MI.Uni-Koeln.DE Subject: More on ncr scsi performance Cc: current@FreeBSD.org Sender: current-owner@FreeBSD.org Precedence: bulk Hi Stefan, I am still having 50% performance slowdowns on my ncr connected drive. Besides the well known chatter problem, there also appears to be a throughput problem with the driver or scsi system. But others, notably David Greenman, are not seeing this. I think he is using the buslogic card. I haven't been able to run more benchmarks because trying to build the world Sunday failed after 11 hours! This used to complete in five. Last night's build failed (after 2 hours, thank god) with: cc -fpic -DPIC -O2 -m486 -pipe -DLIBC_RCS -DSYSLIBC_RCS -D__DBINTERFACE_PRIVATE -DPOSIX_MISTAKE -I/usr/src/lib/libc/locale -DYP -c /usr/src/lib/libc/i386/sys/i386_set_ldt.c -o i386_set_ldt.so building shared c library (version 2.0) swab.so: Definition of symbol `_swab' (multiply defined) swab.so: Definition of symbol `_swab' (multiply defined) *** Error code 1 Stop. *** Error code 1 Here's /dev/sd0e on a 030495 system, and the bonnie: 33001 files, 551347 used, 339568 free (28000 frags, 38946 blocks, 3.1% fragmentation) Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 030495 100 2275 81.1 2294 24.9 1450 25.7 2552 82.3 3807 43.8 65.6 7.2 Here's /dev/sd0e on a 021095 system, and the bonnie: 27098 files, 480655 used, 413680 free (18032 frags, 49456 blocks, 2.0% fragmentation) Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU 021095 100 3145 96.0 3832 30.9 1630 21.0 3230 95.2 4231 36.8 81.0 7.0 The 030495 system is quite a bit better than a few days earlier. The disk fragmentation then was about 2.5%. Really, it was. System: ASUS P54SP4, ncr, 32MB, IBM 0662. As an aside, I am also sure that the slight delays in interactive response are real, I can tell by comparing the two systems (I have two P5-90s configured identically, running these two systems). Doesn't seem to cause any problems. As another aside, I haven't had any more of the ominous looking messages: Mar 6 02:00:04 zoyd /kernel: sd0(ncr0:1:0): COMMAND FAILED (4 28) @f0992a00. Mar 6 02:00:04 zoyd /kernel: sd0(ncr0:1:0): NOT READY asc:04 ascq:01, retries:4 Mar 6 02:00:04 zoyd last message repeated 2 times in recent systems. Cheers, Russell