NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/arch/10532

Internet Message Format | 1992-11-08 | 3.7 KB

Xref: sparky comp.arch:10532 comp.unix.bsd:8616 comp.os.linux:16066
Newsgroups: comp.arch,comp.unix.bsd,comp.os.linux
Path: sparky!uunet!zaphod.mps.ohio-state.edu!sol.ctr.columbia.edu!ira.uka.de!math.fu-berlin.de!informatik.tu-muenchen.de!LRZnews!k2
From: [e-mail] (Klaus Steinberger)
Subject: Re: IDE faster than Mips SCSI disk
Message-ID: <k2.721260853@woodstock>
Sender: [e-mail] (Mr. News)
Organization: Leibniz-Rechenzentrum, Muenchen (Germany)
References: <[e-mail]> <[e-mail]>
Date: Sun, 8 Nov 1992 22:14:13 GMT
Lines: 90

[e-mail] (MAEDA Atusi) writes:

>In article <[e-mail]> [e-mail] (Othman Ahmad) writes:
>I think larger buffer size (e.g. 8192 bytes) would be better for
>testing disk drive/interface performance, while smaller buffer size
>test can measure total I/O performance including OS/library overhead.
Sure.

>I'm very interested in the result of EISA SCSI-2 with large file size
>& large buffer size, if it's available.

I didn't have a EISA box, so I could not post numbers for such a beast,
but with good OS, controller, disk, they should perform quite good.

Instead here are some numbers from my systems:

1. A CD4330 with EP/IX 1.4.3, 64 Mbyte RAM (upto 32 Mbyte Diskcache)
   (CD4330 == Mips RC3230). The machine has an integrated SCSI Controller
   with up to 5 Mbyte/sec. The tested disk is a DEC DSP3105 (1Gbyte)
   Note that the speed is very high even with 512 byte per record!

    131072 records which are each 512 bytes in length.
    It then reads the file.  It prints the bytes-per-second
    rate at which the computer can read and write files.


Writing the 64 Megabyte file, 'iozone.tmp'...44.140000 seconds
Reading the file...38.640000 seconds

IOZONE performance measurements:
    1520363 bytes/second for writing the file
    1736771 bytes/second for reading the file

Same box with 8192 byte/record:


    8192 records which are each 8192 bytes in length.
    It then reads the file.  It prints the bytes-per-second
    rate at which the computer can read and write files.


Writing the 64 Megabyte file, 'iozone.tmp'...43.690000 seconds
Reading the file...33.580000 seconds

IOZONE performance measurements:
    1536023 bytes/second for writing the file
    1998477 bytes/second for reading the file

2. A CD4680 (== Mips RC6280) with 3 SCSI Controller (Interphase Jaguar),
   128 Mbyte RAM (upto 64Mbyte cache). The tested disk is a striped
   virtual disk containing two Seagate Wren VII. The disks are on different
   SCSI controllers:


    IOZONE writes a 128 Megabyte sequential file consisting of
    262144 records which are each 512 bytes in length.
    It then reads the file.  It prints the bytes-per-second
    rate at which the computer can read and write files.


Writing the 128 Megabyte file, 'iozone.tmp'...51.460000 seconds
Reading the file...47.680000 seconds

IOZONE performance measurements:
    2608195 bytes/second for writing the file
    2814969 bytes/second for reading the file

And the same with 8192 bytes/record:

    IOZONE writes a 128 Megabyte sequential file consisting of
    16384 records which are each 8192 bytes in length.
    It then reads the file.  It prints the bytes-per-second
    rate at which the computer can read and write files.


Writing the 128 Megabyte file, 'iozone.tmp'...41.820000 seconds
Reading the file...47.120000 seconds

IOZONE performance measurements:
    3209414 bytes/second for writing the file
    2848423 bytes/second for reading the file


Sincerely,
Klaus
--
Klaus Steinberger               Beschleunigerlabor der TU und LMU Muenchen
Phone: [phone]        Hochschulgelaende
FAX:   [phone]        D-8046 Garching, Germany
Internet: [e-mail]