This document describes the Adabas Cluster Services 8.4 enhancements.
| Category | Enhancements |
|---|---|
| Adabas Cluster Services 8.4 SP1 Enhancements | PLXCB Structures in Dataspace |
| Collaboration Between Cluster Nuclei | |
| ADARUN Parameter Enhancements | |
| Operator Command Enhancements | |
| Utility Enhancements | |
| ADASMF Enhancements |
Adabas Cluster Services Version 8.4 introduces the ability to place the PLXCB structures in a dataspace, in order to provide relief from 31-bit ECSA storage constraints. The PLXCB structures maintain information about the active nuclei and users of a cluster database, which is used to route the commands of the users to their assigned nuclei in the cluster. If several different cluster databases have nuclei or users running on the same system, placing their PLXCB structures in dataspaces may substantially reduce the use of common storage (ECSA), if the configured maximum number of users (NU parameter) is large.
For a given operating system image, to place the PLXCB structures for a cluster database in a dataspace, specify the new parameter LOC=DSP in the ADACOM initialization task for that database. In this case, ADACOM must be started before the first cluster nucleus starts on the system image and it must stay active as long as a nucleus or users of the database are active on that system. For more information, read PLXCB Structures and LOC - Specify PLXCB Location.
Two enhancements in Adabas Cluster Services Version 8.4 make the Adabas nuclei in a cluster more independent of one another, by reducing the need for nuclei to collaborate in cluster-wide internal processes. This takes away two classes of scenarios where a temporarily unresponsive nucleus may impact the processing of the other nuclei in the cluster.
For more information about these enhancements than that given here, read Collaboration Between Cluster Nuclei.
The new ADARUN parameters CLUPUBLPROT and CLUWORK1CACHE may be used to configure the publishing of protection data within the cluster. This influences the way buffer flushes react if one nucleus in the cluster is slow or unable to respond when asked by a buffer flush to write its latest protection data to the WORK data set.
The publishing policy previously used by Adabas is set by CLUPUBLPROT=NO. This is the default value. With this setting, cluster nuclei do not publish their protection data within the cluster (i.e., make it available to the other nuclei) before they write related updated ASSO or DATA blocks to the cache. Instead, when a buffer flush is about to write updated blocks out to the database, all nuclei in the cluster must collaborate with the buffer flush and write their latest protection data to their WORK data sets.
A new publishing policy may be specified by setting CLUPUBLPROT=YES. With this setting, the nuclei publish their protection data in the global cache or on their WORK data sets before they write related updated ASSO or DATA blocks to the cache. In return, a buffer flush can proceed even if a nucleus is slow or unable to respond to a request to write its latest protection data to WORK. In this case, the buffer flush reads the latest protection data of the unresponsive nucleus from the cache and writes it to WORK by itself.
The new CLUWORK1CACHE parameter configures how many WORK blocks a cluster nucleus may keep in the global cache at the same time, if CLUPUBLPROT is set to YES. CLUWORK1CACHE implicitly regulates the use of cache writes versus WORK I/Os for the publishing of protection data and is performance-sensitive for update-intensive workloads.
Setting CLUPUBLPROT=YES makes the nuclei in a cluster more independent of one another by allowing any nucleus to perform a buffer flush without the collaboration of all of its peer nuclei.
For more information about these parameters, read CLUPUBLPROT Parameter: Publishing of Protection Data and CLUWORK1CACHE Parameter: Number of WORK Blocks in Global Cache.
The new ADARUN parameter UPDATECONTROL may be used to set the scheduling policy for update commands at the end of buffer flushes.
This parameter is available for both cluster and non-cluster nuclei. It is introduced in the Release Notes for Adabas Version 8.4.
Setting UPDATECONTROL=NODELAY makes the nuclei in a cluster more independent of one another, by not forcing the runout of all update commands (update command synchronization) after every buffer flush. These synchronization processes require the collaboration of all nuclei in the cluster.
For more information about the UPDATECONTROL parameter, read UPDATECONTROL Parameter: Control Scheduling of Update Commands During Buffer Flush.
Adabas Cluster Services Version 8.4 introduces two new ADARUN parameters: CLUPUBLPROT and CLUWORK1CACHE. Both are related to the Buffer Flush Independence enhancement described above under Collaboration Between Cluster Nuclei. See that section for more information.
Adabas Cluster Services Version 8.4 introduces or enhances the following operator commands:
| Command | New or Changed | Enhancement Description |
|---|---|---|
| CLUPUBLPROT=YES|NO | New | This new command dynamically modifies the setting of the ADARUN CLUPUBLPROT parameter. |
| DPARM | Changed | In cluster nuclei, the parameters listed include the new CLUPUBLPROT and CLUWORK1CACHE parameters. |
| DRES | Changed | In cluster nuclei, the resources listed include the new protection data area extension on the WORK dataset ("Work Pt 1b"). This is also shown in the Adabas session end statistics. |
| DSTAT | Changed | In cluster nuclei, the statistics include
the extra WORK I/Os for publishing protection data and the timeouts of buffer
flush-related V2 commands, if at least one of these numbers is nonzero or
CLUPUBLPROT is set to YES. This is also shown in the
Adabas session end statistics.
|
| DXCACHE | Changed | This includes the cache I/Os for WORK blocks, if at least one of the related numbers is nonzero. This is also shown in the Adabas session end statistics. |
| DXSTAT | Changed | This includes the cache I/Os for WORK
blocks, if at least one of the related numbers is nonzero, and it also includes
the report by the new DXWORK command.
|
| DXWORK | New | This new command displays the I/O statistics for the WORK datasets of all nuclei in the cluster. See WORK Data Set I/O Statistics for more information about these statistics. This is also shown in the Adabas session end statistics. |
Adabas Cluster Services Version 8.4 introduces or enhances the following ADADBS OPERCOM functions:
| Utility Function | New or Changed | Enhancement Description |
|---|---|---|
| ADADBS OPERCOM CLUPUBLPROT=YES|NO | New | This new function dynamically modifies the setting of the ADARUN CLUPUBLPROT parameter. |
| ADADBS OPERCOM DPARM | Changed | For cluster nuclei, the parameters listed include the new CLUPUBLPROT and CLUWORK1CACHE parameters. |
| ADADBS OPERCOM DRES | Changed | For cluster nuclei, the resources listed include the new protection data area extension on the WORK dataset ("Work Prt1b"). |
| ADADBS OPERCOM DSTAT | Changed | For cluster nuclei, the statistics
include the extra WORK I/Os for publishing protection data and the timeouts of
buffer flush-related V2 commands, if at least one of these numbers is nonzero
or CLUPUBLPROT is set to YES.
|
| ADADBS OPERCOM DXCACHE | Changed | This includes the cache I/Os for WORK blocks, if at least one of the related numbers is nonzero. |
| ADADBS OPERCOM DXSTAT | Changed | This includes the cache I/Os for WORK blocks, if at least one of the related numbers is nonzero, and it also includes the report by the new DXWORK function. |
| ADADBS OPERCOM DXWORK | New | This new function displays the I/O statistics for the WORK datasets of all nuclei in the cluster. See WORK Data Set I/O Statistics for more information about these statistics. This is also shown in the Adabas session end statistics. |
These functions are also available via the Adabas Online System (AOS).
Furthermore, for each WORK dataset in a cluster, the ADAWRK utility prints information from the protection data area extension as part of its summary report.
Adabas Cluster Services 8.4 adds the following information to the SMF records created by cluster nuclei, if ADARUN parameter SMF=YES:
If SMFDETAIL includes ‘CSHB’, the cache statistics include those for block type ‘WORK’.
If SMFDETAIL includes ‘PARM’, the parameter settings include those for the new CLUPUBLPROT and CLUWORK1CACHE parameters.
If SMFDETAIL includes ‘IODD’, the I/O statistics include those for other WORK datasets in a cluster. The DD names have prefix ‘ADW’.
If SMFDETAIL includes ‘SESS’, the session statistics include the blocks, I/Os and waits for the extra publishing of protection data on WORK Part 1 induced by CLUPUBLPROT=YES, as well as the buffer flush V2 timeouts.