Tape - Performance Problems, Jobs Taking Too Long to Run

Asset ID:	1-75-1019816.1
Update Date:	2012-04-30
Keywords:

Solution Type Troubleshooting Sure

Solution 1019816.1 : Tape - Performance Problems, Jobs Taking Too Long to Run

Applies to:

Sun StorageTek 9840D Tape Drive - Version Not Applicable and later
Sun StorageTek 9940 Tape Drive - Version Not Applicable and later
Sun StorageTek 9940B Tape Drive - Version Not Applicable and later
Sun StorageTek 9840 Tape Drive - Version Not Applicable and later
Sun StorageTek T10000A Tape Drive - Version Not Applicable and later
All Platforms
Checked for relevance on 10-Jan-2011.

Purpose

Diagnosing Read/Write Issues on Tape.
Diagnosing Performance Issues on Tape.

Troubleshooting Steps

Steps to Follow
How to work with Performance Problems
Performance problems in the tape world usually mean that the job is taking too long to run.

Run time may exceed some timer either in the job istself or in the system.

    What indicators are available?
       - Failing job step. What was this step doing?
       - Are there any messages such as SIM errors that would indicate a time out condition is occurring?

    What command or command sequence is being sent to the drive?
       - This can be obtained from system error messages, from a FC trace, or a drive dump.
       - Locate or Space command failing - the tape may have an invalide or partially valid MIR.

Slow Writes
    Writes are taking too long:
       Look at the block size being used.
          - If small block sizes are being used on 9x40 drives, turn on the performance feature.
       Verify how much idle time there is between commands.
       Verify how long it takes to transfer the data from the host to the buffer.
       Verify how long it takes for the host to write to the drive.
       It is possible that the drive is not being kept busy enough and has to stop then start back up if the data stream is not sefficient to keep up.
          Note: T10000 has two speeds to help compensate for slower host transfer rates. Later versions of LTO drives also have this feature.
       Bottleneck in the Interface/SAN.
       All of the above can be identified from a drive dump.

Slow Reads
    Reading a tape takes too long:
       Verify how much idle time there is between read commands.
       Verify how long it takes to transfer the data from the drive to the host.
          - Possible bottleneck in the Interface/SAN.
          - Could indicate an overloaded server.
       It is possible that the drive cannot stream data fast enough and the buffer fills. If this occurs, the drive has to stop then start back up when buffer space is available.
          - T10000 has two speeds to help compensate for slower host transfer rates.
       All of the above can be identified from a drive dump.

==========================
Information on pulling drive dumps:
1013674.1 for T10000 drives
1014147.1 for 9X40 drives
1013260.1 for LTO drives
1013576.1 for SDLT drives
==========================

If problems still persist after reviewing the above information, a case should be opened up through MOS to provide further assistance.@<document:><document:><document:><document:><document:>

Attachments

This solution has no attachment