SPXTAPE - the Next Generation!

Table of Contents

SPXTAPE - the Next Generation!

  • What is SPXTAPE?
  • SPXTAPE Performance
  • SPXTAPE Tape Usage
  • Multiple Tape Drives
  • SPXTAPE Command Syntax
  • Error Handling
  • Console Output
  • Other Improvements
  • Summary

    • SPXTAPE - the Next Generation!

    What is SPXTAPE?

    Have you always wanted to back up the spool files on your system? Did you have to do it but dreaded the thoughts of using the SPTAPE command? Never fear, the next generation of VM spool backup has arrived! It is called SPXTAPE and became available with VM/ESA Release 1.2.2.

    With a first glance at the documentation for the new SPXTAPE command you might wonder "What's new?" The command syntax appears to be very much like that of SPTAPE. On further reading, you'll find that there is a world of difference between the two. Many outstanding SPTAPE requirements have been satisfied by SPXTAPE. Many improvements have been made in the areas of performance, tape usage, and usability.

    The new SPXTAPE CP command will be discussed in this article. Also the answer to the question that probably popped into your mind when you first saw the title of this article, "What does the 'X' stand for in SPXTAPE?" Read on if you've ever considered backing up your spool files, if you currently back up your spool files and want a better way, or if you simply are curious about the 'X' in SPXTAPE.

    SPXTAPE Performance

    If you've been dumping large numbers of spool files to tape using SPTAPE, you know that it can take quite some time. SPTAPE takes one spool file at a time and processes it serially. That is, a data page is read in from the spool DASD first. After that read is completed, the data page is written to the tape. After that I/O is complete, another data page is read from DASD. And so on, until all of the data pages for all of the selected files have been dumped. This can take hours if you are dumping large numbers of files.

    By contrast, SPXTAPE uses a series of asynchronous tasks to overlap the DASD I/O. During an SPXTAPE DUMP operation, tasks are created to read ahead; i.e. many data pages of a file are read in from DASD before the tape tasks are ready to write them out to the tape volume. In this way, records are prepared in advance and the next record to be written is generally ready to go as soon as the tape task is ready for it. This keeps the tape as busy as possible, thus writing the files to tape as quickly as possible. A similar technique is used by SPXTAPE to improve the performance of SPXTAPE LOAD over that of SPTAPE LOAD.

    Requesting many spool DASD reads or writes at one time leads to other advantages. Since SPXTAPE (and spool code in general) uses CP's paging system to read and write spool data pages, advantage can be taken of the paging system's built in performance algorithms. These algorithms sort the read or write requests in a way that minimizes DASD head movement and generally leads to the best possible performance.

    The result of these improvements is that SPXTAPE is much faster than SPTAPE. In performance runs designed to simulate the spool file size distribution observed on a local CMS-intensive production system, the VM performance people obtained the following results.

    The system contained 34,000 spool files on four 3390 spool volumes. The reader files ranged in size from 100 to 600,000 80-byte records and occupied, on average, 10.4 pages of spool DASD. SPXTAPE was 9.0 to 10.8 times faster than SPTAPE for dumping these spool files. SPTAPE took 3 hours and 24 minutes to dump these files but SPXTAPE took approximately 20 minutes to dump the same files. SPXTAPE loaded these files 1.9 times faster than SPTAPE; 1 hour 8 minutes versus 2 hours 10 minutes.

    Your exact results will depend on many factors such as number of spool volumes, number of tape drives used, other work running on the system, spool data page distribution, etc.. However, it is clear that SPXTAPE is much faster than SPTAPE.

    For more detailed information on the performance study done, refer to VM/ESA Release 2.2 Performance Report, GC24-5673-01.

    SPXTAPE Tape Usage

    The format of the records written to tape by SPXTAPE is completely different from the format of the records written by SPTAPE. The disadvantage is that SPTAPE and SPXTAPE cannot be used to read tapes dumped by each other. The advantage is that there were no compatibility restrictions on the design, so the tape format was designed for maximum advantage. SPXTAPE was designed to write records to the tape in blocks of slightly more than 32K bytes of data. Any combination of file description information (SPFBKs), data pages (SPDBKs), and XAB data can be written into a block of approximately 32K bytes in length. A header is used to describe exactly what data exists in each record.

    Using this format significantly increases the number of spool files and the amount of data that will fit on a single tape volume. An additional advantage is that less I/O operations are required to the tape device to actually dump or load the data to or from the tape. In the performance study described in the previous section, the 34,000 files required 11 tape volumes when dumped with SPTAPE. The same files dumped with SPXTAPE required 7 tape volumes.

    Multiple Tape Drives

    There are two major differences in the way SPXTAPE and SPTAPE handle tape drives.

    • With SPXTAPE, more than one tape drive may be used for the same SPXTAPE operation, so a range of tape drives may be specified on each SPXTAPE command. SPTAPE only uses one tape drive per operation.
    • Tape drives to be used by SPXTAPE are attached to the virtual machine that issues the SPXTAPE command. SPTAPE does not use tape drives that are attached to a virtual machine; a drive specified on an SPTAPE command must be "FREE". The SPXTAPE command uses virtual addresses (either a single address or a range of addresses is accepted) for its tape drives, while SPTAPE uses a single real address for its tape drive.

    Since more than one tape drive can be used for one SPXTAPE operation, enough tape drives can be used so that the maximum capacity of the spool DASD volumes can be utilized. Clearly the number of tape drives to use will be installation dependent but a good general guideline is to use one tape drive for each 4-6 spool volumes with a minimum of two if you expect the data to fill more than one volume. Using at least two drives allows SPXTAPE operations to continue on other drives while rewind is occurring and new volumes are being mounted.

    Using at least two tape drives also allows use of the capability for dynamically adding or removing drives from the running SPXTAPE operation. If more than one drive is being used to dump files and one of them fails, the operation will continue on the drive that is still working. The failing drive can actually be removed from the operation and another can be added to take its place. All of this can occur while the operation continues on the original drive that didn't fail. By contrast, if the drive that SPTAPE was dumping to fails, the operation stops and must be started all over again.

    Below is an example of how this could be done. 

      SPXTAPE DUMP 182-183 RDR ALL REWIND <==> starts the operation
  =============> tape drive 183 fails <=======================
  SPXTAPE DUMP 182-184 <========> includes 184 in the operation
  SPXTAPE CANCEL 183 <==========> removes 183 from the operation
    Notice that the command which started the operation includes the file selection criteria but when tape drives are dynamically added to or removed from the operation, the file selection criteria is not included on the command. Also, notice that the original virtual address range is included in the virtual address range of the command to add the new drive into the dump operation (i.e. the original virtual address range is "expanded" to include the new tape drive).

    SPXTAPE Command Syntax

    As mentioned earlier, the syntax diagram for SPXTAPE looks a lot like the syntax diagram for SPTAPE. However, there are differences. A range of virtual tape drives is specified as opposed to a single real address and there is much more flexibility in the area of file selection criteria.

    There are additional spool queue keywords. For example, 

      SPXTAPE DUMP 182 STD ALL
    will dump all RDR, PRT, and PUN files. With SPTAPE, each of those three sets of files had to be dumped separately. The following command will dump all of the system data files. 
      SPXTAPE DUMP 182 SDF ALL
    And this command 
      SPXTAPE DUMP 182 SPOOL ALL
    will dump all RDR, PRT, PUN, and SDF files.

    The new APPEND option adds even more flexibility to file selection. This option can be used to tie several SPXTAPE commands together so that all of the files selected by all of the commands can be dumped or loaded or scanned as if by a single command. For example, the following command only loads files that are on the reader queue and have a FORM set to "std" and have a FILE TYPE of "assemble". 

      SPXTAPE LOAD 181 RDR FORM std FT assemble
    However, the following set of appended commands (or "logical command") dumps reader files that have FORM set to "std" or have a FILE TYPE of "assemble". Note that files which match both of these criteria may be dumped twice. 
      SPXTAPE DUMP 181 RDR FORM std    APPEND
  SPXTAPE DUMP 181 RDR FT assemble REWIND

    Another example is a way of dumping all printer files except those that are Class A. The following set of appended commands accomplishes this. 

      SPXTAPE DUMP 181 PRT CLASS bcdefghi APPEND
  SPXTAPE DUMP 181 PRT CLASS jklmnopq APPEND
  SPXTAPE DUMP 181 PRT CLASS rstuvwxy APPEND
  SPXTAPE DUMP 181 PRT CLASS z1234567 APPEND
  SPXTAPE DUMP 181 PRT CLASS 890      REWIND

    For more details on the syntax and use of SPXTAPE, see the VM/ESA Release 2.2 CP Command and Utility Reference, SC24-5519-04.

    Error Handling

    The error recovery approach has been totally changed from that of SPTAPE. Rather than stop the whole DUMP or LOAD operation when an error is encountered, SPXTAPE is designed to continue dumping or loading as much data as possible.

    All selected files will be dumped as long as their control information is readable. If any XAB data is not readable, the XAB data is omitted. If any of a file's data is not readable, only the unreadable data is not dumped. The part of the file that is readable will be dumped.

    Similarly, if SPXTAPE LOAD has trouble reading a portion of the tape, it will keep trying and will continue loading if a readable record in SPXTAPE format is found after the error. Even if the reading of one tape fails, the operation will continue on any other drives started with this operation.

    Console Output

    SPXTAPE reduces console output by logging each file rather than writing a message to the screen for each one. Each SPXTAPE operation creates at least two log files. One, called the Command Summary Log (CSL), contains progress information, error messages (many of which are also displayed on the screen), and summary data when the command ends. There is one CSL created for each "logical SPXTAPE command".

    In addition to the CSL, a Volume Log (VL) is created for each tape volume that is written to or read from. The main information in the volume log(s) is the list of files that were dumped to or read from the tape volume and any error indications for each of those files. The list of files dumped, loaded, or scanned contains much of the same information about each file that is returned from the QUERY RDR/PRT/PUN command.

    Since each file is no longer displayed as it is dumped, a progress report (heart beat) is displayed periodically so that the command issuer knows SPXTAPE is progressing. This progress report is displayed approximately every 15 seconds and has the following information: 

      SPXTAPE DUMPing 182:    35 FILES,    1,019 SPOOL PAGES  23% COMPLETE
    As a result of customer feedback a PROGress_interval option has been developed and is available as a local CP modification. It may be obtained from the VM Download Library on Your Connection to VM (part of the S/390 Developer's Association). This modification allows tailoring of the progress report time interval with a command option. With the modification, the interval between progress reports can be set anywhere from 1 second to 9999 minutes or it can be turned off by setting it to 0. The default interval remains 15 seconds with the modification applied.

    Note: The PROGress_interval option was added to SPXTAPE in the base of the VM/ESA 2.1.0 and later releases.

    Other Improvements

    And that is not all! The following goodies have also been included in SPXTAPE.

    • Standard Tape Label Toleration - SPXTAPE will not write over an existing header label group of a standard tape label.

    • Huge file support - SPXTAPE allows you to dump a file that will not fit on one tape.

    • Files that are split across more than one tape are correctly put back together even when you load the tapes in a different order than they were dumped, as long as you load all of the tapes with the same SPXTAPE LOAD command.

    • NODUP option - helps you avoid the duplicate NSS problem. When loading files with SPXTAPE, duplicates of files that already exist on the system will not be created if you use the NODUP option.

    • Class G users - can use SPXTAPE to dump their own spool files.

    • QUERY VIRTUAL TAPE - support added as a result of customer feedback. This command now reports SPXTAPE status on the drive that is being queried if SPXTAPE is running there. It reports whether SPXTAPE is active, waiting for another tape to be mounted, or waiting for another command to be appended.

    Summary

    At first glance SPXTAPE may not appear to be much different than SPTAPE, but when you look a little deeper, there is a world of difference. SPXTAPE improves on SPTAPE in the areas of performance, tape usage, and usability. Spool back up with SPXTAPE will be much faster and easier than it ever was in the past.

    And now the answer to the question you've been waiting for, "What does the 'X' in SPXTAPE stand for?" To be honest with you, it doesn't have a real meaning. The SPXTAPE name came about because SPXTAPE is the XA-spool-format version of the SPTAPE command. But we in VM development hope that after you use SPXTAPE, you will believe it stands for X-cellent!


    The information provided, and views expressed on this site are my own and do not represent the IBM Corporation.