Disk access/MPT under ESX3.5

[Adam McDougall]

On Mon, May 19, 2008 at 11:02:31AM +0200, Daniel Ponticello wrote:

  Hello,
  
  
  monitor# camcontrol negotiate 0:0 -W 16
  Current Parameters:
  (pass0:mpt0:0:0:0): sync parameter: 0
  (pass0:mpt0:0:0:0): offset: 0
  (pass0:mpt0:0:0:0): bus width: 8 bits
  (pass0:mpt0:0:0:0): disconnection is enabled
  (pass0:mpt0:0:0:0): tagged queueing is enabled
  monitor# dd if=/dev/zero of=/var/tmp/dead.file bs=1024k count=1000
  1000+0 records in
  1000+0 records out
  1048576000 bytes transferred in 32.421679 secs (32341817 bytes/sec)
  monitor# dd if=/dev/zero of=/var/tmp/dead.file bs=1024k count=1000
  1000+0 records in
  1000+0 records out
  1048576000 bytes transferred in 20.355797 secs (51512402 bytes/sec)
  
  No improvements. But it looks like it did not renegotiated the transfers 
  data rate that for
  some odd reasons are setted as 3.3MB/S instead of 320mb/s.
  
  I made some tests using linux 2.18 (debian):
  
  debiantest:/home/daniel# uname -a
  Linux debiantest 2.6.18-6-686 #1 SMP Sun Feb 10 22:11:31 UTC 2008 i686 
  GNU/Linux
  
  scsi0 : ioc0: LSI53C1030, FwRev=00000000h, Ports=1, MaxQ=128, IRQ=169
   Vendor: VMware    Model: Virtual disk      Rev: 1.0
   Type:   Direct-Access                      ANSI SCSI revision: 02
  target0:0:0: Beginning Domain Validation
  target0:0:0: Domain Validation skipping write tests
  target0:0:0: Ending Domain Validation
  target0:0:0: FAST-160 WIDE SCSI 320.0 MB/s DT IU RDSTRM RTI WRFLOW PCOMP 
  (6.25
  ns, offset 127)
  
  debiantest:/home/daniel# dd if=/dev/zero of=dead.file bs=1024k count=1000
  1000+0 records in
  1000+0 records out
  1048576000 bytes (1.0 GB) copied, 5.01316 seconds, 209 MB/s
  
For the Linux test, are you sure it didn't cache part of the write 
before returning?  You may need to add some syncs and make it part
of the elapsed time.  Just checking because this seems to be my experience.