Got it.

With compression turned on, the 3590's were getting 28 MB/sec using
FICON.

With compression turned on, the 3590's were getting 13 MB/sec using
ESCON.

Like Tom said, "You've gotta feed the beast".

____________________
Jim Hughes
603-271-5586
"It is fun to do the impossible."

==>-----Original Message-----
==>From: The IBM z/VM Operating System [mailto:[email protected]]
On
==>Behalf Of Hughes, Jim
==>Sent: Monday, June 01, 2009 2:55 PM
==>To: [email protected]
==>Subject: Re: Z/VM and 3590 tape speed
==>
==>I am looking at table 3 on page 10 of the 3590 Introduction and
Planning
==>Guide.
==>
==>It mentions two things of interest:
==>Device Data Rate(native) and Data Transfer Rate(maximum
instantaneous).
==>
==>Device Data Rate for the E11 is 14 MB/Sec.
==>Data Transfer Rate depends on the channel.
==>
==>I guess what I am reading says the tape will move around at 14
meg/sec
==>no matter how fast I get the data to the device. FICON for tapes is a
==>waste of time.
==>
==>When would FICON for tapes be a good idea?
==>
==>
==>____________________
==>Jim Hughes
==>603-271-5586
==>"It is fun to do the impossible."
==>
==>==>-----Original Message-----
==>==>From: The IBM z/VM Operating System
[mailto:[email protected]]
==>On
==>==>Behalf Of Tom Duerbusch
==>==>Sent: Monday, June 01, 2009 2:46 PM
==>==>To: [email protected]
==>==>Subject: Re: Z/VM and 3590 tape speed
==>==>
==>==>You're getting confused between tape drive speed and channel
speed.
==>==>
==>==>The IBM 3590 E and H drives will read/write to the tape at a rate
of
==>14
==>==>MBs.  Multiple that rate by your compression rate, if you have
==>==>compression turned on.
==>==>
==>==>The FICON channel, can operate at 200 MB or 270 MBs (on a z/890).
==>==>
==>==>So if you had 2:1 compression, you would see a transfer of 28 MBs.
==>==>Uncompressed it will be 14 MBs at best.
==>==>
==>==>And that is what you are seeing.
==>==>
==>==>These are data streaming drives.  You have to feed the beast.  If
==>your
==>==>source (disk) can't keep up with the tape drive and it's
compression,
==>you
==>==>will see much, much lower transfer numbers.
==>==>
==>==>Tom Duerbusch
==>==>THD Consulting
==>==>
==>==>>>> "Hughes, Jim" <[email protected]> 6/1/2009 1:08 PM >>>
==>==>We have Z/VM 5.4, with 4 3590's attached to the z890 using FICON
==>Express
==>==>with the 2-Gbps feature. The tape controller is the A60.
==>==>
==>==>
==>==>
==>==>I am doing some tape speed tests and my results are not very good.
I
==>==>have a pipeline(look below) that writes 65,500 byte blocks to the
==>3590.
==>==>I used the cms command TAPE MODESET ( NOCOMP/COMP  to turn
==>compression
==>==>off/on.
==>==>
==>==>
==>==>
==>==>I see around 14 MB/sec with compression off and 28 MB/sec with
==>==>compression on.
==>==>
==>==>
==>==>
==>==>The 3590 Planning manual says the 3590's should operate at
200MB/sec
==>==>when using FICON with the 2-Gbps feature.
==>==>
==>==>
==>==>
==>==>There maybe a problem with my pipeline or my expectations.
==>==>
==>==>
==>==>
==>==>Here is the pipeline.  Any insight or suggestions are welcome.
==>==>
==>==>
==>==>
==>==>/*  */
==>==>
==>==>arg blkcnt .
==>==>
==>==>if blkcnt  = "" then exit 99
==>==>
==>==>
==>==>
==>==>numeric digits 12
==>==>
==>==>
==>==>
==>==>"TAPE REW"
==>==>
==>==>"TAPE FSF"
==>==>
==>==>say time()
==>==>
==>==>x = time("E")
==>==>
==>==>
==>==>
==>==>"PIPE literal  <------------>  | >> tapetest 3590 a"
==>==>
==>==>
==>==>
==>==>"PIPE (endchar ?)        ",
==>==>
==>==>  "? xrange 00-ff       |",    /* 256 characters*/
==>==>
==>==>  "  dup 256            |",
==>==>
==>==>  "  specs todclock 1 1-* n |",
==>==>
==>==>  "  join  256          |",    /* 64k  */
==>==>
==>==>  "  dup  *             |",
==>==>
==>==>  "  specs todclock 1 1-* n |",
==>==>
==>==>  "  fblock 65500 00        |",
==>==>
==>==>  "  take " blkcnt "        |",
==>==>
==>==>  "  cnt: count lines       |",
==>==>
==>==>  "   tape 181              |",
==>==>
==>==>    " specs 1-1 1           |",
==>==>
==>==>    " join 4999             |",
==>==>
==>==>    " specs todclock 1      |",
==>==>
==>==>    " dateconv 1.8 todabs     ISODATE timeout        |",
==>==>
==>==>    " specs recno 1                                   ",
==>==>
==>==>    "       substring    fs . f 1 of word 2  nw       ",
==>==>
==>==>    "    /5,000 Blocks Interval Check/  nw           |",
==>==>
==>==>    " console                                        |",
==>==>
==>==>    " >> tapetest 3590 a                              ",
==>==>
==>==>  "? cnt: | var xblkcnt | specs / Blocks written =/ 1 w 1 nw |",
==>==>
==>==>  " console"
==>==>
==>==>
==>==>
==>==>____________________
==>==>
==>==>Jim Hughes
==>==>
==>==>603-271-5586
==>==>
==>==>"It is fun to do the impossible."
==>==>
==>==>

Reply via email to