Subject: News Server - Increasing disk performance?
To: None <netbsd-help@NetBSD.ORG>
From: John D. Smerdon <jds@smerdon.livonia.mi.us>
List: netbsd-help
Date: 09/13/1996 19:17:03
Are there configuration options that would increase disk performance? I
have a news server that can not sustain two articles per second and it is
starting to fall behind. It usually takes from 200-800ms after receiving
an article before it is written to disk and an acknowledgement (235?) is
sent.
I am getting more memory, a second SCSI controller, and disk, but it just
seems that the current configuration should be faster.
Nothing else runs on the system, and there are between 5 and 15 news
readers at any one time.
Thanks for any advice!
Config: NetBSD 1.2_BETA from a couple weeks ago, Pentium 90, 32MB memory,
single 2940, two 4GB disks with root on sd0a, swap on sd0b and sd1b, /usr
ccd'ed over both disks, and /news ccd'ed over both disks.
The news history file is on /usr/home/news, overview data is on
/news/over.view. 14 days of news in the big eight and a couple small
hierarchies and no alt.binaries or alt.sex.
# tcpdump port nntp (with my comments)
12:17:30.374507 netbsd.nntp > feed.2656: P 140:145(5) ack 23871 win 24576
235\r\n
12:17:30.440567 feed.2656 > netbsd.nntp: P 23871:23922(51) ack 145 win 16384
ihave <message-id>\r\n
12:17:30.440790 netbsd.nntp > feed.2656: . ack 23922 win 24525
12:17:30.443796 netbsd.nntp > feed.2656: P 145:150(5) ack 23922 win 24576
335\r\n
12:17:30.501732 feed.2656 > netbsd.nntp: . 23922:24434(512) ack 150 win 16384
12:17:30.502205 feed.2656 > netbsd.nntp: P 24434:24490(56) ack 150 win 16384
12:17:30.504225 netbsd.nntp > feed.2656: . ack 24490 win 24576
12:17:30.524423 feed.2656 > netbsd.nntp: P 24490:24493(3) ack 150 win 16384
.\r\n
12:17:30.524894 netbsd.nntp > feed.2656: . ack 24493 win 24576
12:17:30.792764 netbsd.nntp > feed.2656: P 150:155(5) ack 24493 win 24576
235\r\n
# pstat -s
Device 1K-blocks Used Avail Capacity Type
/dev/sd0b 133760 30072 103688 22% Interleaved
/dev/sd1b 133760 31216 102544 23% Interleaved
Total 267520 61288 206232 23%
# df -i
Filesystem 1K-blocks Used Avail Capacity iused ifree %iused
Mounted on
/dev/sd0a 32399 14039 16740 46% 712 7222 9% /
/dev/sd1a 32399 14041 16738 46% 714 7220 9% /altroot
/dev/ccd0d 1556494 826469 652200 56% 49760 326558 13% /usr
/dev/ccd1d 6209801 5175084 724226 88% 1302045 199393 87% /news
mfs:21 123887 3 117689 0% 3 30715 0% /tmp
# cat /etc/ccd.conf
ccd0 110 0 /dev/sd0e /dev/sd1e
ccd1 110 0 /dev/sd0f /dev/sd1f
# vmstat -w 5 sd0 sd1 ccd0 ccd1
procs memory page disks faults cpu
r b w avm fre flt re pi po fr sr s0 s1 c0 c1 in sy cs us sy id
0 1 0 70608 1664 3 0 0 0 0 0 40 42 11 64 0 102 77 3 5 92
0 2 0 80000 1664 3 0 0 0 0 0 36 36 6 58 0 170 89 4 7 89
0 2 0 84868 1300 82 3 3 0 0 0 41 37 18 51 0 135 108 7 6 87
0 1 0 84868 1300 3 0 0 0 0 0 43 37 12 60 0 99 98 4 4 92
0 1 0 84804 1300 4 0 0 0 0 0 38 35 10 54 0 81 108 3 3 94
0 2 0 84804 1300 1 0 0 0 0 0 45 42 6 73 0 85 111 2 6 91
1 0 0 89948 1300 15 3 0 0 0 0 39 41 11 59 0 140 128 4 5 92
0 2 0 99340 1300 13 3 0 0 0 0 38 41 10 57 0 151 110 4 6 90
0 0 0 94472 1664 4 0 0 0 0 0 35 32 15 45 0 116 103 4 4 92
0 0 0 94472 1664 3 0 0 0 0 0 35 40 13 54 0 109 77 4 5 90
# disklabel sd0
# /dev/rsd0d:
type: SCSI
disk: mysd
label:
flags:
bytes/sector: 512
sectors/track: 110
tracks/cylinder: 19
sectors/cylinder: 2090
cylinders: 3992
total sectors: 8343280
rpm: 3600
interleave: 1
trackskew: 0
cylinderskew: 0
headswitch: 0 # milliseconds
track-to-track seek: 0 # milliseconds
drivedata: 0
6 partitions:
# size offset fstype [fsize bsize cpg]
a: 66880 0 4.2BSD 1024 8192 16 # (Cyl. 0 - 31)
b: 267520 66880 swap # (Cyl. 32 - 159)
c: 8343280 0 unused 0 0 # (Cyl. 0 - 3991)
d: 8343280 0 unused 0 0 # (Cyl. 0 - 3991)
e: 1605120 334400 4.2BSD 1024 8192 16 # (Cyl. 160 - 927)
f: 6403760 1939520 4.2BSD 1024 8192 16 # (Cyl. 928 - 3991)
--
John D. Smerdon; Livonia, Michigan, USA; Contents are my opinion.
Home: jds@smerdon.livonia.mi.us