2.1 CF Structures Overview
Within the CF, storage is dynamically partitioned into structures. OS/390 (XES)
services can then manipulate data within the structures. Each of the following
structure types has unique functions:
Cache structures
Sharing CPCs can use the CF to cache shared data. Three types of caching
are done: cache
directory-only
, cache
store-through
, and cache
store-in
. All
cache structures have directory entries. The store-through and store-in
structures also have data entries.
Directory-only caching consists of a directory element that records which
users have an interest in a particular shared piece of data. The actual
shared data does not exist in the CF structure.
For cache store-through, the data is written to the CF structure and also to
DASD immediately.
For cache store-in, DASD is written to the CF structure, and to DASD at some
later time.
For shared data caches using the CF, cross-invalidation of local buffers is
done automatically. Examples of the use of caching are:
For members in the data sharing group, IMS/ESA V5 uses cache
structures for OSAM and VSAM buffers. Here, the cache structures are
used to manage the
integrity
of data in the local buffer pools. This is an
example of directory-only usage of the cache.
CICS/VSAM RLS uses cache structures to maintain buffer and data
consistency at the control interval (CI) level of all systems in a Parallel
Sysplex. CICS/VSAM RLS uses the cache structure for store-through
caching; that is, data is written to DASD as well as to the cache
structure. Status is returned to the subsystem when both the CF write
and DASD write have completed.
IMS/ESA V6 has an option to use a store-through cache for OSAM data
sharing in a similar fashion.
RACF or OS/390 Security Server uses a cache structure to quickly access
frequently used information located in the database. Store-through
caches are used for the RACF primary and backup databases.
DB2 uses a cache structure to store buffer pool data, which is later read
back to DB2 to be written to DASD. This is an example of cache store-in.
DFSMS/MVS V1.5 Enhanced Catalog Sharing (ECS) uses a cache
structure to store catalog sharing information that was previously kept in
the VVDS.
The ECS use of the cache structure is a unique implementation in that
information that is written to the cache structure is never written to
DASD. The cross-invalidation feature of cache structures is used to
inform the Catalog Address Spaces if another system updates a catalog
record that they have an in-storage copy of. If the cache structure
becomes unavailable, all in-storage catalog records are marked as being
invalid, and the Catalog Address Space reverts to using the VVDS to
store catalog integrity information.
60
Parallel Sysplex Configuration, Volume 2: Cookbook
Lock structures
Centralized lock tables are used in the CF to provide synchronous
serialization and contention detection. Following are examples of the use of
locking:
The CF is used as a high-speed locking facility by the Integrated
Resource Lock Manager (IRLM), which in turn supports IMS DB and DB2
data sharing.
CICS/VSAM RLS uses the locking function of the CF to provide
record-level locking.
GRS star support uses the contention detection and management
capability of the lock structure to determine and assign ownership of a
particular global resource.
List Structures
List structures provide shared queues and shared status information.
Messages can be exchanged using CF list structures to implement
high-performance data sharing. The following are examples of list structure
usage:
JES2 uses the list structure for shared work queues. JES2 can place the
checkpoint data set in the CF to exchange job information between
systems. When a system accesses the checkpoint, it can review all
changes from other systems and add its own updates, without having to
serialize on a DASD volume.
OS/390 Communications Server and VTAM use a list structure for shared
status generic resource information.
DB2 uses a list structure for the Shared Communication Area. This is
used for inter-DB2 communication and also to record status information
about shared databases.
The system logger component of OS/390 may be used by, for example,
OPERLOG to create a single logstream in a list structure.
XCF may use the list structure for message passing. Prior to the CF,
channel-to-channel (CTC) links were required between each
communicating processing unit.
XCF message passing exploitation includes:
- GRS ring processing (not applicable to GRS star configurations)
- Workload Manager (WLM), to pass performance management data
among systems in the sysplex
- IRLM uses XCF to exchange lock information between IRLM
instances on multiple systems
- CICS, for MRO requests between images
- JESXCF, which in turn may be used by
JES2, to send messages for checkpoint error or reconfiguration
JES3, to do cross-system communication
TCAS, invokes JESXCF as part of the generic logon processing
- Console, for multisystem message passing
- VLF, to send messages
Chapter 2. A Structured View of CF Structures
61

Get OS/390 Parallel Sysplex Configuration, Volume 2: Cookbook now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.