Two LTO3 drives, different performance

[Dave Chamberlin]

Greetings. I have a tale of two servers and two tape drives. Servers are PA-RISC, an rp8400 (test) and an rp7420 (production), both running running 11.23. I bought an LTO3 drive. I tested it on the rp8400. Worked great. Bought 2nd drive with different enclosure from 2nd vendor (same firmware). Installed on rp7420. Performance worse than LTO2 drives on same server. Thought it was hardware issue, tried several things. No change. Decided to move it to the test server for easier testing, and take the test server drive and move it to production. Did not move the cable. Driver performance on the prod server still sucks, and the former prod drive works great on the test server. Cables are identical (C2362). Both have identical dual ultra320 cards, and firmware is very close. Using LTT utility shows negotiated burst rate on the prod system is only 20MB/s - but on the test system is 320MB/s. Any ideas would be appreciated.

[Alzhy]

Do you have the same FCMS versions and patches between the rp8400 and rp8420?

[Bulent ILIMAN]

are the HBAs same?

may be your SCSI adapter on the 7420 does not support U320.

If possible replace the SCSI cards and test again, there may be a HW issue about the cables also, you tested the drives and both are OK, now time to test the cables and HBA's. If nothing changes you can also check the slot where HBA's installed, because all PCI slots on rp servers does not have same specs,

[Alzhy]

HOW are you testing/comparing the throughputs? And can you post the "device" name used by your tests -- also those of LTT?

20MB/s versus 320MB/s Burst looks like on one server you have an UW-320 SCSI whilst the other is just a FW-Diff 20 (very old).

IF you say your HBAs are practically the same -- then I suspect the Enclosure's BUS is probably NOT optimal and is only able to burst at FW-SCSI speeds.... Just a guess though. You can try opening up the innards of the fast and slow enclosures and see if the ribbon cables, chips used on the IC board are the same -- it will surely give yo a clue.

Also Was your 2nd Enclosured LTO3 CHEAPER than the first? If it was - then the may explain likely the issue. The ENclosure is likely using slow (for DLT era) SCSI internals...

[Torsten.]

Possible reasons are a bad cable (bent pin) or a broken/wrong terminator. Both can force a bus speed downshift ...

[Dave Chamberlin]

The HBAs are identical (A7173A) on both servers, and Ultra320 is supported on the rp7420. I am using the health tests in the hp_ltt utility. The LTO2 drives have burst rates of 160MB/s on this server, and out perform the LTO3 by 2-3x. I will try the terminator and cable next. I was trying to determine if there were any config settings that determined the burst speed. Does anyone know when/how the burst rate is determined (bootup? on tape activity?) Thanks for the ideas.

[Marino Meloni_1]

negoziation at 20 MBs is usually due to a bad cable or terminator,
running LTT and collecting a support ticket usually can confirm this
can you attach it here for a review?
you should see the scsi negoziation running at 16 bit LVD and not 8 bit SE