Date: Wed, 6 Dec 1995 16:38:23 -0600 From: "Charles F. Randall" <crandall@dmacc.cc.ia.us> To: julian@TFS.COM (Julian Elischer) Cc: questions@freebsd.org Subject: Re: Adaptec 2940W vs Seagate 32550W Message-ID: <9512061638.ZM20932@vislab19.dmacc.cc.ia.us> In-Reply-To: julian@TFS.COM (Julian Elischer) "Re: Adaptec 2940W vs Seagate 32550W" (Dec 4, 5:10pm) References: <m0tMltp-0003w9C@TFS.COM>
next in thread | previous in thread | raw e-mail | index | archive | help
Julian,
Note: I've carbon-copied this to questions@freebsd.org
System: Gateway 133MHz Pentium
64 MB RAM
Adaptec 2940W SCSI Controller
Seagate "Barracuda" 2.1 GB Drive (ST32550W)
I've re-partitioned and installed 2.1 on another Barracuda and am seeing
similar (poor) filesystem performance (~3 MB/s compared to the ~5 MB/s
sequential block read rate I saw on the Seagate Hawk drive).
Figuring that I'd messed something up before, I used a "standard" 2047/64/32
geometry and took the default partition sizes. That didn't change performance
much (if any).
Is there some "optimal" geometry and partition size combination that I could
derive from some set of formulas or is this a "black art"? I've appended
Seagate's tech sheet for the 'Cuda drive below.
I can only note that this drive is listed with a speed of 3600 RPM in the
disklabel but it is actually a 7200 RPM drive.
Here are the performance numbers I'm seeing with this configuration.
For the raw device,
flash:~> time dd if=/dev/rsd0 bs=64k of=/dev/null count=1000
1000+0 records in
1000+0 records out
65536000 bytes transferred in 9 secs (7281777 bytes/sec)
0.000u 0.155s 0:09.68 1.5% 63+331k 0+0io 0pf+0w
flash:~> time dd if=/dev/rsd0 bs=64k of=/dev/null count=1000
1000+0 records in
1000+0 records out
65536000 bytes transferred in 9 secs (7281777 bytes/sec)
0.010u 0.148s 0:09.50 1.5% 67+301k 0+0io 0pf+0w
Using Bonnie on /usr/cuda,
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
cuda 128 3217 62.9 3072 13.8 2016 17.3 5067 97.2 5542 23.5 146.3 6.5
The disklabel is, (output of "disklabel -r sd0")
# /dev/rsd0c:
type: SCSI
disk: sd0s1
label:
flags:
bytes/sector: 512
sectors/track: 32
tracks/cylinder: 64
sectors/cylinder: 2048
cylinders: 2047
sectors/unit: 4194058
rpm: 3600
interleave: 1
trackskew: 0
cylinderskew: 0
headswitch: 0 # milliseconds
track-to-track seek: 0 # milliseconds
drivedata: 0
8 partitions:
# size offset fstype [fsize bsize bps/cpg]
a: 65536 0 4.2BSD 0 0 0 # (Cyl. 0 - 31)
b: 284864 65536 swap # (Cyl. 32 - 171*)
c: 4194058 0 unused 0 0 # (Cyl. 0 - 2047*)
e: 61440 350400 4.2BSD 0 0 0 # (Cyl. 171*- 201)
f: 3782218 411840 4.2BSD 0 0 0 # (Cyl. 201*- 2047*)
During boot, I noticed the following:
...
ahc0 <Adaptec 2940 SCSI host adapter> rev 0 int a irq 9 on pci0:13
^^^^^^
ahc0: 2940 Wide Channel, SCSI Id=7, aic7870, 16 SCBs
ahc0 waiting for scsi devices to settle
(ahc0:0:0): "SEAGATE ST32550W 0016" type 0 fixed SCSI 2
sd0(ahc0:0:0): Direct-Access 2047MB (4194058 512 byte sectors)
de0 <Digital DC21041 Ethernet> rev 17 int a irq 9 on pci0:14
^^^^^^
de0: DC21041 [10Mb/s] pass 1.1 Ethernet address 00:00:c0:01:7f:c7
...
Is it normal that the controller and ethernet card would both use IRQ 9? Could
this interfere with performance?
If I'm bothering you or there's someone else I should be contacting about this,
just let me know.
Thanks for your help,
-Randy
http://www.seagate.com/techsuppt/drivespecs/scsi/st32550w.html
--- snip ---
ST-32550W
Barracuda 2LP
UNFORMATTED CAPACITY (MB) ________________2541.3
FORMATTED CAPACITY (xx SECTORS) (MB) _____2147.4
AVERAGE SECTORS PER TRACK ________________108 rounded down
ACTUATOR TYPE ____________________________ROTARY VOICE COIL
TRACKS ___________________________________38,610
CYLINDERS ________________________________3,510 user
HEADS ______PHYSICAL______________________11
DISCS (3.5 in) ___________________________6
MEDIA TYPE _______________________________THIN FILM
RECORDING METHOD _________________________ZBR RLL (1,7)
INTERNAL TRANSFER RATE (mbits/sec)________49.4 to 72.0
EXTERNAL TRANSFER RATE (mbyte/sec) _______5 Async
EXTERNAL TRANSFER RATE (mbyte/sec) _______5.0 to 20 Sync
SPINDLE SPEED (RPM) ______________________7,200
AVERAGE LATENCY (mSEC) ___________________4.17
BUFFER ___________________________________512 KByte
Read Look-Ahead, Adaptive,
Multi-Segmented Cache
INTERFACE ________________________________SCSI-2 FAST Wide
BYTES PER TRACK __________________________64,160 average
SECTORS PER DRIVE ________________________
TPI (TRACKS PER INCH) ____________________4,048
BPI (BITS PER INCH) ______________________72,680
AVERAGE ACCESS (ms) (read/write)__________8.0/9.0
Drive level without controller overhead
SINGLE TRACK SEEK (ms) ___________________0.6/0.9
MAX FULL SEEK (ms) _______________________17/19
MTBF (power-on hours) ____________________800,000
SHOCK (G's):
operating (Read/Write) __________2
nonoperating ____________________50
ACOUSTICS (typ/max dBA) __________________4.7/
POWER REQUIREMENTS: +12V START-UP (amps) _1.26
+12V TYPICAL (amps) __0.88 idle
+5V START-UP (amps) __0.91
+5V TYPICAL (amps) ___0.33 idle
IDLE (watts) _________
LANDING ZONE (cyl) _______________________AUTO PARK
IBM AT DRIVE TYPE ________________________0 or NONE
Physical:
Height (inches/mm): 1.00/25.4
Width (inches/mm): 4.00/101.6
Depth (inches/mm): 5.75/146.05
Weight (lbs/g): 1.2/545
Already low-level formatted at the factory with 9 spare sectors
per cylinder and 1 spare cylinder per unit.
ZBR = Zone Bit Recording = Variable sectors per track
--- snip ---
--
Charles F. Randall E-mail: crandall@dmacc.cc.ia.us
UNIX Systems Programmer Voice: (515) 965-7057 FAX: (515) 965-7305
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9512061638.ZM20932>