home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.arch
- Path: sparky!uunet!rde!ksmith!keith
- From: keith@ksmith.uucp (Keith Smith)
- Subject: Re: IDE faster than SCSI-2
- Organization: Keith's Computer, Hope Mills, NC
- Date: Sat, 07 Nov 92 18:36:35 GMT
- Message-ID: <1992Nov07.183635.14973@ksmith.uucp>
- References: <1992Nov5.094712.19056@ntuix.ntu.ac.sg>
- Lines: 80
-
- In article <1992Nov5.094712.19056@ntuix.ntu.ac.sg> eoahmad@ntuix.ntu.ac.sg (Othman Ahmad) writes:
- >[ Article crossposted from comp.unix.bsd,comp.os.linux ]
- >[ Author was Othman Ahmad (eoahmad@ntuix.ntu.ac.sg) ]
- >[ Posted on Thu, 5 Nov 1992 08:07:16 GMT ]
- >
- >I posted an iozone benchmark using my 386/25Mhz machine which shows that
- >it is slightly slower than a SCSI-2 hard-disk. Now with a 486/33, IDE is
- >definitely faster than a EISA SCSI-2 486/50MHZ hard-disk.
- > Iozone uses normal unix system calls for writing to disks, after all,
- >that is how we write C programs.
- > I may try to run a more sophisticated file system tester called
- >bi? later on.
- >
- >
- > JUlian's machine
- > is a 50MHz EISA 486 with a 1.3GB drive attached via a Bustek 742a SCSI2
- > adapter. It has 16MB of ram. Part of the disk (about 200MB is taken up
- > by mach2.6 and is unavailable to 386bsd.
- >
- >I checked the load on this machine, it is very light. There is only another
- >user apart from I.
- >
- >iozone 1
- >
- >Writing the 1 Megabyte file, 'iozone.tmp'...1.766667 seconds
- >Reading the file...1.983333 seconds
- >
- >IOZONE performance measurements:
- > 593534 bytes/second for writing the file
- > 528694 bytes/second for reading the file
- >
-
- Too small a file. Try 8MB file. These numbers are pretty *BAD*. I do
- much better than that with SCO 3.2.2 on SCSI-I disks (~ 1M/sec). SCO is
- not noted for filesystem speed either. What are the sizes of your
- buffer caches?
-
- Also you are using 2 different machines. I would imagine that the 50Mhz
- 486 EISA is tuned QUITE A BIT differently than the smaller 486/33 System
- too.
-
- What model 1.3G drive? How fragmented the filesystem? If the
- Filesystem is fragmented badly you could be jumping around all over it
- to write that 1M file.
-
- >Running on 486/33 with 256Kbyte RAM cache, maxtor 200Mbyte IDE hard-disk,
- >8Mbyte main RAM.
-
- This is a very new and very fast & very small (comparatively) drive.
- *REMEMBER*, A LARGER filesystem has a logrithmic (not expotential)
- increase in overhead as the size increases, and if fragmented, you could
- easily be bouncing all over the disk to find the blocks you need.
-
- >
- > Whereas the IDE runs XFree86 and open look window manager.
- >I only use 1 megabyte for test file size.
- >
- > IOZONE performance measurements:
- > 873813 bytes/second for writing the file
- > 452623 bytes/second for reading the file
-
- And then ... You WROTE faster, but *READ* slower. That means you wrote
- to a buffer cache more than likely. Again. Use an 8MB or 16MB file for
- a REAL test. Also look at the WIDE difference here. By all rights you
- should READ faster than you write.
-
- Then do a *SECOND* test. Run IOZONE simultaneously in two windows on
- each machines and record the results. This MAY suprise you a bit.
-
- > Both machine uses 386bsd 0.1 heavily patched, so the timing measurements
- >are consistent with each other.
-
- Also, it looks like your SCSI driver sucks, Also with only 1 disk and
- everything else** equal SCSI won't be appreciably faster than IDE.
-
- ** everthing else = Seek time, size, bus master or pio, etc ad nauseum.
- --
- Keith Smith uunet!ksmith!keith 5719 Archer Rd.
- Digital Designs BBS 1-919-423-4216 Hope Mills, NC 28348-2201
- Somewhere in the Styx of North Carolina ...
-