Your Browser is not longer supported

Please use Google Chrome, Mozilla Firefox or Microsoft Edge to view the page correctly
Loading...

{{viewport.spaceProperty.prod}}

Parallel and serial processing in ARCHIVE

&pagelevel(4)&pagelevel

The following section describes how you can achieve better performance when working on storage level S2.

Each SAVE, RESTORE and COPY request is processed by a single HSMS server task. This HSMS server task divides the request into so-called packets. Each packet can be processed individually by an ARCHIVE subtask. ARCHIVE subtasks can run in parallel. At the beginning, each ARCHIVE subtask receives a packet. When a subtask has processed a packet it receives a new one.

To ensure good performance, the number of ARCHIVE subtasks generated should be the same as or less than the number of packets created. If multiplexing operation has not been activated, each ARCHIVE subtask requires one tape device for SAVE/RESTORE activities and two tape devices for COPY activities. The number of ARCHIVE subtasks that can successfully run in parallel is limited by the number of available tape devices.

If multiplexing operation is active, multiple ARCHIVE subtasks can share the same devices for SAVE/RESTORE activities in order to achieve better performance and device utilization. For COPY activities, twice as many devices are required for the backup.

In HSMS the division into packets is performed implicitly. The number of ARCHIVE subtasks required for RESTORE and COPY activities is calculated automatically. The user can control the number of packets created for a BACKUP-FILES request as well as the number of ARCHIVE subtasks generated for BACKUP-FILES or BACKUP-NODE-FILES (see "User control of packet generation").

If preprocessing/postprocessing is activated (PRE-POST-PROCESSING operand) in the case of operations involving the BACKUP-NODE-FILES and RESTORE-NODE-FILES statements, the ARCHIVE subtasks await the results of the preprocessing action before they commence data exchange with the workstation.

Multiplexing operation

In the multiplexing operating mode, multiple ARCHIVE subtasks can simultaneously sharethe same MTC devices and magnetic tape cartridges in parallel. This improves performance during save and restore operations and ensures optimal utilization of tape devices and streaming. In addition, the magnetic tape cartridges are optimally filled, thus reducing storage costs.

In multiplex operation the number of tape devices (or tape device pairs when copying) used in parallel for a request is preset, as is the mulitplex factor which determines the number of ARCHIVE subtasks which share a tape device.

In BS2000 save operation multiplex operation is well suited for configurations with high-performance cartridge devices (as of TAPE-C5/-C6).

In HSMS, multiplexing operation is achieved by connecting multiple ARCHIVE subtasks to a single ARCHIVE tape drive. The ARCHIVE tape drive is responsible for all accesses to the connected device. The number of subtasks for a device driver task determines its multiplexing factor. As a result, the subtasks refer to the device driver task when writing to or reading from tape.

At restore time, automatic parallel demultiplexing is performed. The number of subtasks required depends on the number of devices, the multiplexing configuration during backup and the files that are to be restored (see also section "Multiplexing operation").

Multiplexing operation is activated by specifying PARALLEL-RUNS=*MULTIPLEXING(...) in the HSMS statement BACKUP-NODE-FILES, the suboperands NUMBER-OF-DEVICES and MULTIPLEXING-FACTOR determining the number of devices and the multiplex factor. The settings for multiplex operation can also be specified at archive level using the HSMS statements CREATE-ARCHIVE and MODIFY-ARCHIVE-ATTRIBUTES. During the backup these are evaluated by specifying the PARALLEL-RUNS=*STD option.

If data is backed up to disk, no multiplexing is performed. However, the number of ARCHIVE subtasks is exactly the same as the planned number of devices (NUMBER-OF-DEVICES).

In the case of MULTIPLEXING-FACTOR=*AUTOMATIC (default setting) ARCHIVE determines a favorable multiplexing factor by itself. The result of this operation gives the multiplexing factor per drive. For further information, refer to section "Multiplexing operation".

Specifying node files via PATH-NAMES in multiplexing operation

If the path names of the node files to be processed are taken over from a file or a library element (PATH-NAMES=*FROM-FILE or *FROM-LIBRARY-ELEMENT operand), they are processed in the sequence in which the user has entered them in the specified file or library element. Consequently, BACKUP-NODE-FILES operations can be optimized by planning the sequence of files carefully. See also section "Selection of node files of the BS2000-UFS".

New format for multiplexed save files

Save files created in multiplexing mode have a slightly different format from save files which were not created in multiplexing mode. Among other things, additional control information is added to the tape security blocks.

Continuing save files in multiplexing operation

A non-multiplexed save file cannot be continued with a multiplexed save version since the tape formats are different. The entire save file must have a uniform format if restore operations are to be performed correctly.

However, it is possible to continue a multiplexed save file with a non-multiplexed save version. ARCHIVE automatically creates a multiplexing environment. In this case, the employed multiplexing factor is 1.

Selecting files via *LATEST-BACKUPS-OR-S0 in multiplexing operation

If the files for backup are selected by means of the specification FROM=*LATEST-BACKUPS-OR-S0, different save files can be used as input during the offline copying phase. This means that both multiplexed and non-multiplexed save files can be copied. When a multiplexed save version is copied, the entire new output save file must be created as a multiplexed file. As a result, non-multiplexed input save versions are multiplexed in the new output save file.