gtpd1m15Database Reference

3880 Record Cache RPQ

This section provides an overview of the 3880 Record Cache RPQ and a description of the TPF functions you can use in conjunction with that RPQ.

Hardware

The 3880 Record Cache RPQ contains a large subsystem random access electronic storage (as many as 64 megabytes) used to store active DASD data for quick access. This means that frequently used DASD data can remain in the cache, reducing the number of repetitive DASD accesses to a single initial I/O operation. In addition, data stored in the cache can be accessed at a much higher speed than data stored in the DASD device.

Modes of Operation

The 3880 Record Cache RPQ operates in either record access mode or direct mode. All DASD devices attached to the storage director must operate in either mode. To directly support TPF, the 3880 Record Cache RPQ (operating in record access mode) stores single DASD records for accesses when requested by the TPF system. Direct mode can access data directly from the DASD devices.

Initial Program Load

One function of the TPF IPL is to initialize the record cache control unit and place the devices in record access mode, the normal operating state of 3880 attached record cache devices. However, initializing to record access mode does not take place for a general file IPL or an IPL of other than the first processor in a loosely coupled complex. When storage or hardware failures prevent initialization or ready status, or a device cannot be placed in record access mode, IPL processing notifies the operator and continues the IPL by placing the devices in direct mode.

If the system is IPLed before a cache allocation request completes, the request terminates and no allocation processing takes place. A warning message is issued to notify the operator and the IPL continues. It is the system operator's responsibility to insure that the pending allocation operation is completed to the 3880 Record Cache Control Units in the complex using either the prior (current) or new (target) allocation values.

DASD Processing and Error Recovery

TPF DASD support routines handle the new channel command words (CCW) for record cache as well as any new error conditions that may result. DASD processing CCWs are built by segment CJII from indicators set by IPL processing. Normally a buffered CCW string is built to access record cache DASD.

I/O Retry

On a buffered CCW I/O failure with unit check and command reject, the DASD program CCSONS retries the operation once using direct access CCWs. If the retry is successful, direct access CCWs are used to address the device until either a TPF IPL occurs or the device's status changes. If the retry is unsuccessful, a catastrophic error occurs.

Module Up/Down Processing

The ZMCPY DOWN command returns the record cache device to direct mode if it was in record access mode. This allows other processors not running TPF to access the device.

The ZMCPY UP command attempts to place the record cache device into record access mode. If the device is being placed online on an available and initialized record cache control unit, an attempt is made to place the device into record access mode. If this attempt fails, the device is placed in direct mode and an error message is issued.

If the device is the first device to be placed online for the record cache control unit then the cache must be made ready before the device is put into record access mode and initialized. If any failures occur, the device is left in direct mode.

The processing of a general file mount and dismount is similar to module up and down.

Data Collection and Data Reduction

Performance statistics are collected for 3880 Record Cache Control Units and devices attached to the control units. The ZMEAS command is used to offload the collected data from each device. These statistics are displayed on the reports generated by data reduction.

Restrictions

Nonretentive data is not supported by this RPQ. Loader general files cannot be attached to this RPQ. Devices are not sharable among subsystems, but are sharable among processors running the same subsystem.

Command Description

The TPF commands that support 3880 Record Cache operations are:

You are not required to modify the record cache, but you can adjust the ratios for block sizes in keypoint 0 from the default values.

Note:
Throughout the TPF publications, the term ratio refers to the hardware term weighted values.

In addition, you can monitor data block usage and control unit status as explained later in this section.

Specifying Block Size Ratios

You can partition the record cache into the 3 TPF record sizes (381, 1055, and 4096 bytes) by assigning block size ratios. Defaults for all 3 ratio values are set in keypoint 0 at one, dividing the cache equally among the TPF block sizes. In addition, one full track buffer is defaulted in keypoint 0.

You can change the block size ratios by using the ZBUFC ALLOCATE command. The allocation is a 2 step process:

  1. First use the ZBUFC ALLOCATE RC381-rr,RC1055-rr,RC4096-rr,RCBUF-bb command to specify the target allocation values.
  2. Next use the ZBUFC ALLOCATE IMPLEMNT command to implement those values. These values will then be applied to all the record cache control units in the complex.

The ZBUFC ALLOCATE DISPLAY command displays the current and target cache allocation settings for the specified record cache control unit.

Monitoring Control Unit Status

The ZBUFC STATUS command lets you monitor the status of a control unit. This display informs you of device operating modes, storage and interface status, and the storage size allocated to each TPF block size.

For example, record access mode status information can help you determine which devices to take offline. Placing these devices online again requires the ZMCPY UP command and each record caching device should be brought online in record access mode. If the device being placed online is the first device on a specific control unit, then TPF issues an initialization sequence for the control unit before placing the device into record access mode.