z/OS DFSMS Using Data Sets
Previous topic | Next topic | Contents | Contact z/OS | Library | PDF


Using Hiperspace Buffers with LSR

z/OS DFSMS Using Data Sets
SC23-6855-00

If you are using local shared resources (LSR), you can specify multiple 4-KB Hiperspace buffers for each buffer pool in the resource pool. The size of the Hiperspace buffers must be equal to the CISIZE of the data sets being used.

The use of Hiperspace buffers can reduce the amount of I/O to a direct access storage device (DASD) by caching data in central storage. The data in a Hiperspace buffer is preserved unless there is a central storage shortage and the central storage that backs the Hiperspace buffer is reclaimed by the system. VSAM invalidates a Hiperspace buffer when it is copied to a virtual address space buffer and, conversely, invalidates a virtual address space buffer when it is copied to a Hiperspace buffer. Therefore at most there is only one copy of the control interval in virtual address space and Hiperspace. When a modified virtual address space buffer is reclaimed, it is copied to Hiperspace and to DASD.

For the data pool or the separate index pool at OPEN time, a data set is assigned the one buffer pool with buffers of the appropriate size—either the exact control interval size requested, or the next larger size available.

You may have both a global resource pool and one or more local resource pools. Tasks in an address space that have a local resource pool may use either the global resource pool, under the restrictions described below, or the local resource pool. There may be multiple buffer pools based on buffer size for each resource pool.

To share resources locally, a task in the address space issues BLDVRP TYPE=LSR, DATA|INDEX. To share resources globally, a system task issues BLDVRP TYPE=GSR. The program that issues BLDVRP TYPE=GSR must be in supervisor state with key 0 - 7.

You can share resources locally or globally, with the following restrictions:
  • LSR (local shared resources). You can build up to 255 data resource pools and 255 index resource pools in one address space. Each resource pool must be built individually. The data pool must exist before the index pool with the same share pool identification can be built. The parameter lists for these multiple LSR pools can reside above or below 16 MB. The BLDVRP macro RMODE31 parameter indicates where VSAM is to obtain virtual storage when the LSR pool control blocks and data buffers are built.

    These resource pools are built with the BLDVRP macro TYPE=LSR and DATA|INDEX specifications. Specifying MACRF=LSR on the ACB or GENCB-ACB macros causes the data set to use the LSR pools built by the BLDVRP macro. The DLVRP macro processes both the data and index resource pools.

  • GSR (global shared resources). All address spaces for a given protection key in the system share one resource pool. Only one resource pool can be built for each of the protection keys 0 - 7. With GSR, an access method control block and all related request parameter lists, exit lists, data areas, and extent control blocks must be in the common area of virtual storage with a protection key the same as the resource pool. To get storage in the common area with that protection key, issue the GETMAIN macro while in that key, for storage in subpool 241. If you need to share a data set among address spaces, multiple systems, or both, consider using record-level sharing (RLS) instead of GSR.

    The separate index resource pools are not supported for GSR.

    The Hiperspace buffers (specified in the BLDVRP macro) are not supported for GSR.

    Generate ACBs, RPLs, and EXLSTs with the GENCB macro: code the WAREA and LENGTH parameters. The program that issues macros related to that global resource pool must be in supervisor state with the same key. (The macros are BLDVRP, CHECK, CLOSE, DLVRP, ENDREQ, ERASE, GENCB, GET, GETIX, MODCB, MRKBFR, OPEN, POINT, PUT, PUTIX, SCHBFR, SHOWCB, TESTCB, and WRTBFR. The SHOWCAT macro is not related to a resource pool, because a program can issue this macro independently of an opened data set.)

Go to the previous page Go to the next page




Copyright IBM Corporation 1990, 2014