<?xml version="1.0" encoding="UTF-8"?>
<rss xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:taxo="http://purl.org/rss/1.0/modules/taxonomy/" version="2.0">
  <channel>
    <title>topic Re: Effective throughput of SCSI Ultra2 card? in Disk</title>
    <link>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498649#M5361</link>
    <description>Actually Ultra2SCSI is 80MByte/s.&lt;BR /&gt;Protocol overhead is something like five to ten percent.&lt;BR /&gt;Negative factors scsi-wise would be a device that doesn't support disconnection (You won't find any today) or a device that is using single-ended SCSI on the same scsi chain(this will make the whole bus drop to 20 or 40MB/s)&lt;BR /&gt;&lt;BR /&gt;See the following example:&lt;BR /&gt;ext_bus     0  0/0/1/0         c720      CLAIMED     INTERFACE    SCSI C896 Ultr&lt;BR /&gt;a Wide LVD&lt;BR /&gt;target      0  0/0/1/0.3       tgt       CLAIMED     DEVICE&lt;BR /&gt;tape        0  0/0/1/0.3.0     stape     CLAIMED     DEVICE       HP      C5683A&lt;BR /&gt;                              /dev/rmt/0m            /dev/rmt/c0t3d0BESTn&lt;BR /&gt;                              /dev/rmt/0mb           /dev/rmt/c0t3d0BESTnb&lt;BR /&gt;                              /dev/rmt/0mn           /dev/rmt/c0t3d0DDS&lt;BR /&gt;                              /dev/rmt/0mnb          /dev/rmt/c0t3d0DDSb&lt;BR /&gt;                              /dev/rmt/c0t3d0BEST    /dev/rmt/c0t3d0DDSn&lt;BR /&gt;                              /dev/rmt/c0t3d0BESTb   /dev/rmt/c0t3d0DDSnb&lt;BR /&gt;target      1  0/0/1/0.7       tgt       CLAIMED     DEVICE&lt;BR /&gt;ctl         0  0/0/1/0.7.0     sctl      CLAIMED     DEVICE       Initiator&lt;BR /&gt;                              /dev/rscsi/c0t7d0&lt;BR /&gt;ext_bus     1  0/0/1/1         c720      CLAIMED     INTERFACE    SCSI C896 Ultr&lt;BR /&gt;a Wide Single-Ended&lt;BR /&gt;target      2  0/0/1/1.0       tgt       CLAIMED     DEVICE&lt;BR /&gt;disk        0  0/0/1/1.0.0     sdisk     CLAIMED     DEVICE       HP 73.4GMAS373&lt;BR /&gt;5NC&lt;BR /&gt;                              /dev/dsk/c1t0d0   /dev/rdsk/c1t0d0&lt;BR /&gt;target      3  0/0/1/1.7       tgt       CLAIMED     DEVICE&lt;BR /&gt;ctl         1  0/0/1/1.7.0     sctl      CLAIMED     DEVICE       Initiator&lt;BR /&gt;                              /dev/rscsi/c1t7d0&lt;BR /&gt;ext_bus     2  0/0/2/0         c720      CLAIMED     INTERFACE    SCSI C87x Ultr&lt;BR /&gt;a Wide Single-Ended&lt;BR /&gt;target      4  0/0/2/0.0       tgt       CLAIMED     DEVICE&lt;BR /&gt;disk        1  0/0/2/0.0.0     sdisk     CLAIMED     DEVICE       HP 73.4GMAS373&lt;BR /&gt;5NC&lt;BR /&gt;                              /dev/dsk/c2t0d0   /dev/rdsk/c2t0d0&lt;BR /&gt;target      5  0/0/2/0.7       tgt       CLAIMED     DEVICE&lt;BR /&gt;ctl         2  0/0/2/0.7.0     sctl      CLAIMED     DEVICE       Initiator&lt;BR /&gt;                              /dev/rscsi/c2t7d0&lt;BR /&gt;&lt;BR /&gt;here it's completely screwed up - the U160SCSI hard drive is connected to ultrawide SE, and the tape plus another disk are on the LVD channel. in the end, both disks are slowed to UltraWide SE speed, which is at maximum 40MB/s.&lt;BR /&gt;&lt;BR /&gt;root@iwonttell:/&amp;gt;time dd if=/dev/rdsk/c2t0d0 of=/dev/null bs=1024k count=100&lt;BR /&gt;100+0 records in&lt;BR /&gt;100+0 records out&lt;BR /&gt;&lt;BR /&gt;real        2.8&lt;BR /&gt;user        0.0&lt;BR /&gt;sys         0.0&lt;BR /&gt;&lt;BR /&gt;this is approximately 35MB/s, and theoretically noone can tell if the adapter or the disk is the bottleneck here.&lt;BR /&gt;&lt;BR /&gt;Most drives don't really manage to move over 35MB/s, the exceptions are 73GB 15K or 144GB drives which will get You up to 70MB/s and of course two drives will sum up quite a bit in raw throughput.&lt;BR /&gt;&lt;BR /&gt;In the end, I'd say the maximum on U2W would be about 70MB/s with two or three drives concurrently reading, this might also be limited by other devices on the same pci bus or the controller itself.&lt;BR /&gt;&lt;BR /&gt;after You ensured a high throughput on the scsi side You'll have to find how little there is left for the applications. :)</description>
    <pubDate>Tue, 08 Mar 2005 10:01:41 GMT</pubDate>
    <dc:creator>Florian Heigl (new acc)</dc:creator>
    <dc:date>2005-03-08T10:01:41Z</dc:date>
    <item>
      <title>Effective throughput of SCSI Ultra2 card?</title>
      <link>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498646#M5358</link>
      <description>Hello--&lt;BR /&gt;&lt;BR /&gt;I know the maximum theoretical throughput of an Ultra2 SCSI card is 40 MBytes/sec. But what is the effective throughput of such a controller?&lt;BR /&gt;&lt;BR /&gt;Regards,&lt;BR /&gt;cjw&lt;BR /&gt;</description>
      <pubDate>Sun, 06 Mar 2005 16:14:24 GMT</pubDate>
      <guid>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498646#M5358</guid>
      <dc:creator>Rudy Williams</dc:creator>
      <dc:date>2005-03-06T16:14:24Z</dc:date>
    </item>
    <item>
      <title>Re: Effective throughput of SCSI Ultra2 card?</title>
      <link>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498647#M5359</link>
      <description>&lt;BR /&gt;Hah... the old 'it depends' question.&lt;BR /&gt;&lt;BR /&gt;If you are looking at a single process, then with large IOs you can maybe get to 30MB/sec and not be held back by the bus.&lt;BR /&gt;&lt;BR /&gt;With highly concurrent activities, I expect to see the bus being responsible for measurable throughput degradation when it hit 20MB/sec or so.&lt;BR /&gt;&lt;BR /&gt;Simplistically speaking, at that utilization you created 50% odds that an IO is busy when an other IO tries to get on the same bus.&lt;BR /&gt;Now if nothing is directly waiting for the IO (say an async buffer flush, or a read ahead) then no harm is done. But if both users are reading blocks and waiting for the data to be able in order to continue processing then they will be hurting each other too often.&lt;BR /&gt;&lt;BR /&gt;For a single app with concurrent threads it may well be acceptable to just push it to 35MB/sec.&lt;BR /&gt;&lt;BR /&gt;So it all depends, but personally I do not like seeing more than 50% utilization of the theoretical throughput.&lt;BR /&gt;&lt;BR /&gt;Hope this helps some,&lt;BR /&gt;Hein.&lt;BR /&gt;</description>
      <pubDate>Mon, 07 Mar 2005 00:11:50 GMT</pubDate>
      <guid>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498647#M5359</guid>
      <dc:creator>Hein van den Heuvel</dc:creator>
      <dc:date>2005-03-07T00:11:50Z</dc:date>
    </item>
    <item>
      <title>Re: Effective throughput of SCSI Ultra2 card?</title>
      <link>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498648#M5360</link>
      <description>This also depends on how many devices are attached, their addresses, etc. Remember that one "slow" device can effect all the others on the bus.&lt;BR /&gt;&lt;BR /&gt;The best way to maximise the throughput on a SCSI bus is to place the busiest disks toward the end of the chain, with progressively quieter devices closer to the adapter.&lt;BR /&gt;&lt;BR /&gt;If you are using wide addresses remember also the interrupt priority goes from 7-0, then 15 to 8, which is why your adapter will nearly always be device id=7.</description>
      <pubDate>Mon, 07 Mar 2005 02:36:54 GMT</pubDate>
      <guid>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498648#M5360</guid>
      <dc:creator>Andrew Cowan</dc:creator>
      <dc:date>2005-03-07T02:36:54Z</dc:date>
    </item>
    <item>
      <title>Re: Effective throughput of SCSI Ultra2 card?</title>
      <link>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498649#M5361</link>
      <description>Actually Ultra2SCSI is 80MByte/s.&lt;BR /&gt;Protocol overhead is something like five to ten percent.&lt;BR /&gt;Negative factors scsi-wise would be a device that doesn't support disconnection (You won't find any today) or a device that is using single-ended SCSI on the same scsi chain(this will make the whole bus drop to 20 or 40MB/s)&lt;BR /&gt;&lt;BR /&gt;See the following example:&lt;BR /&gt;ext_bus     0  0/0/1/0         c720      CLAIMED     INTERFACE    SCSI C896 Ultr&lt;BR /&gt;a Wide LVD&lt;BR /&gt;target      0  0/0/1/0.3       tgt       CLAIMED     DEVICE&lt;BR /&gt;tape        0  0/0/1/0.3.0     stape     CLAIMED     DEVICE       HP      C5683A&lt;BR /&gt;                              /dev/rmt/0m            /dev/rmt/c0t3d0BESTn&lt;BR /&gt;                              /dev/rmt/0mb           /dev/rmt/c0t3d0BESTnb&lt;BR /&gt;                              /dev/rmt/0mn           /dev/rmt/c0t3d0DDS&lt;BR /&gt;                              /dev/rmt/0mnb          /dev/rmt/c0t3d0DDSb&lt;BR /&gt;                              /dev/rmt/c0t3d0BEST    /dev/rmt/c0t3d0DDSn&lt;BR /&gt;                              /dev/rmt/c0t3d0BESTb   /dev/rmt/c0t3d0DDSnb&lt;BR /&gt;target      1  0/0/1/0.7       tgt       CLAIMED     DEVICE&lt;BR /&gt;ctl         0  0/0/1/0.7.0     sctl      CLAIMED     DEVICE       Initiator&lt;BR /&gt;                              /dev/rscsi/c0t7d0&lt;BR /&gt;ext_bus     1  0/0/1/1         c720      CLAIMED     INTERFACE    SCSI C896 Ultr&lt;BR /&gt;a Wide Single-Ended&lt;BR /&gt;target      2  0/0/1/1.0       tgt       CLAIMED     DEVICE&lt;BR /&gt;disk        0  0/0/1/1.0.0     sdisk     CLAIMED     DEVICE       HP 73.4GMAS373&lt;BR /&gt;5NC&lt;BR /&gt;                              /dev/dsk/c1t0d0   /dev/rdsk/c1t0d0&lt;BR /&gt;target      3  0/0/1/1.7       tgt       CLAIMED     DEVICE&lt;BR /&gt;ctl         1  0/0/1/1.7.0     sctl      CLAIMED     DEVICE       Initiator&lt;BR /&gt;                              /dev/rscsi/c1t7d0&lt;BR /&gt;ext_bus     2  0/0/2/0         c720      CLAIMED     INTERFACE    SCSI C87x Ultr&lt;BR /&gt;a Wide Single-Ended&lt;BR /&gt;target      4  0/0/2/0.0       tgt       CLAIMED     DEVICE&lt;BR /&gt;disk        1  0/0/2/0.0.0     sdisk     CLAIMED     DEVICE       HP 73.4GMAS373&lt;BR /&gt;5NC&lt;BR /&gt;                              /dev/dsk/c2t0d0   /dev/rdsk/c2t0d0&lt;BR /&gt;target      5  0/0/2/0.7       tgt       CLAIMED     DEVICE&lt;BR /&gt;ctl         2  0/0/2/0.7.0     sctl      CLAIMED     DEVICE       Initiator&lt;BR /&gt;                              /dev/rscsi/c2t7d0&lt;BR /&gt;&lt;BR /&gt;here it's completely screwed up - the U160SCSI hard drive is connected to ultrawide SE, and the tape plus another disk are on the LVD channel. in the end, both disks are slowed to UltraWide SE speed, which is at maximum 40MB/s.&lt;BR /&gt;&lt;BR /&gt;root@iwonttell:/&amp;gt;time dd if=/dev/rdsk/c2t0d0 of=/dev/null bs=1024k count=100&lt;BR /&gt;100+0 records in&lt;BR /&gt;100+0 records out&lt;BR /&gt;&lt;BR /&gt;real        2.8&lt;BR /&gt;user        0.0&lt;BR /&gt;sys         0.0&lt;BR /&gt;&lt;BR /&gt;this is approximately 35MB/s, and theoretically noone can tell if the adapter or the disk is the bottleneck here.&lt;BR /&gt;&lt;BR /&gt;Most drives don't really manage to move over 35MB/s, the exceptions are 73GB 15K or 144GB drives which will get You up to 70MB/s and of course two drives will sum up quite a bit in raw throughput.&lt;BR /&gt;&lt;BR /&gt;In the end, I'd say the maximum on U2W would be about 70MB/s with two or three drives concurrently reading, this might also be limited by other devices on the same pci bus or the controller itself.&lt;BR /&gt;&lt;BR /&gt;after You ensured a high throughput on the scsi side You'll have to find how little there is left for the applications. :)</description>
      <pubDate>Tue, 08 Mar 2005 10:01:41 GMT</pubDate>
      <guid>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498649#M5361</guid>
      <dc:creator>Florian Heigl (new acc)</dc:creator>
      <dc:date>2005-03-08T10:01:41Z</dc:date>
    </item>
    <item>
      <title>Re: Effective throughput of SCSI Ultra2 card?</title>
      <link>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498650#M5362</link>
      <description>According to HP, the spec the A5149A card at 40 MB/s. I do not believe that this is an Ultra2 wide card (which runs at 80 MB/s).&lt;BR /&gt;&lt;BR /&gt;Thank you to everyone. I think that I have the information that I need.</description>
      <pubDate>Tue, 08 Mar 2005 10:23:11 GMT</pubDate>
      <guid>https://community.hpe.com/t5/disk/effective-throughput-of-scsi-ultra2-card/m-p/3498650#M5362</guid>
      <dc:creator>Rudy Williams</dc:creator>
      <dc:date>2005-03-08T10:23:11Z</dc:date>
    </item>
  </channel>
</rss>

