You are on page 1of 2

Enhanced Catalog Sharing

"Enhanced catalog sharing (ECS) is a new catalog sharing protocol available with DFSMS V1.5 and OS/390 R7. As an alternative to the
existing VVDS-based sharing protocol, enhanced catalog sharing uses the coupling facility to exchange catalog control information
between systems in a sysplex. The VVDS-based catalog sharing mode remains the default. In this mode DFSMS stores catalog sharing
control information in a VVDS residing on the same DASD volume as the catalog itself. The control information is used to ensure catalog
consistency between the sharing systems. Consequently, when catalog changes occur, along with the I/O request that writes the actual
changes to the catalog data set, additional I/O is performed to update the control information in the VVDS. The additional I/O overhead can
be significant and has the potential to adversely impact catalog performance throughout the sysplex. Instead of using a DASD data set,
enhanced catalog sharing uses a new coupling facility cache structure, SYSIGGCAS_ECS, to hold the shared catalog control records. By
using the coupling facility, enhanced catalog sharing offers improved catalog performance by eliminating the extra DASD I/O to the
VVDSs."

When a new ICF catalog is defined, it is marked shared by default -------------------------- Volume WACCAT Device Data --------
<SHROPTIONS(3,4)> , assuming it is on DASD capable of being Number: 826B Active: 23% Pending: 7%
shared and defined as SHARED in the IODF. When a catalog is Device: 33903 Connect: 16% Delay DB: 1%
shared, it might get updated from another system, possibly Shared: Yes Disconnect: 0% Delay CU: 0%
invalidating any in-storage records that the catalog address space has Delay DP: 1%
chached directly (ISC) or through VLF (CDSC). To check for this, -------------- Data Set Name --------------- Jobname ASID
catalog management maintains a series of catalog update entries in the -- N/A -- *MASTER* 0001
VVR for the catalog, indicating which control intervals have been CATALOG 0032
updated. SYS1.VVDS.VWACCAT CATALOG 0032

In order for one system to know if there have been updates by any sharing system, CAS has to read the catalog VVR record from the VVDS, so
any access to a catalog marked shared without ECS requires at least one read to the VVDS to determine if any updates have taking place. As
standard search order starts with the master catalog, every lookup requires then at least one read of the master catalog which in turn requires
I/O with RESERVE of the relative volume. Enhanced Catalog Sharing addresses and replaces only this "sharing protocol" and does not
substitutes or changes the actual caching of catalog records, therefore all catalogs managed by VLF can and should remain so even after ECS
implementation.

The current working implementation of ECS has only been carried out on our IXCMIAPU
test sysplex. The above structure was defined in our coupling facility by adding
it to our current policy which can be found in the member CFTADMIN of the STRUCTURE NAME(SYSIGGCAS_ECS)
data set SYS1.MVS.CFRM . INITSIZE(256) SIZE(1024)
The structure has been sized using the IBM CFSIZER PREFLIST(IBM#R06T,IBM#R06J)
at http://www-1.ibm.com/servers/eserver/zseries/cfsizer/ecs.html and
specifying 12 as the number of active catalogs.

After defining the CF structure and activating the new policy containing its definition, the command < MODIFY
CATALOG,ECSHR(CONNECT) > was issued on every member of the sysplex to get the catalog to connect to the newly created structure.
At this point, having verified that the correction had taken place, we made - only some user catalogs to start with and later all the others - our
master and user catalogs ECS eligible. This must be done only once by issuing the IDCAMS command < ALTER 'catalogname'
ECSHARING > and does not cause enhanced catalog sharing for this catalog to begin immediately, rather it allows it to be switched later
automatically to ECS. The status of the various catalogs can be displayed by issuing the command <F CATALOG,ECSHR(STATUS)>.

F CATALOG,ECSHR(CONNECT) F CATALOG,ECSHR(STATUS)
IEC351I CATALOG ADDRESS SPACE MODIFY COMMAND IEC351I CATALOG ADDRESS SPACE MODIFY
ACTIVE COMMAND
IXL014I IXLCONN REQUEST FOR STRUCTURE IEC380I ENHANCED CATALOG SHARING
SYSIGGCAS_ECS 961 *CAS*****************************************
WAS SUCCESSFUL. JOBNAME: CATALOG ASID: 0023 * CF Connection: AutoAdd ç autoadd enabled
CONNECTOR NAME: IXCLO01D0002 CFNAME: IBM#R06J * ------------------CATALOG------------------- -----STATUS---
IEC377I ENHANCED CATALOG SHARING: CONNECT * UCAT.ATST16 Inact(NotElig)
COMPLETE * UCAT.ATST15 Inact(NotElig) *
UCAT.ATST14 Inact(NotElig)
* UCAT.ATST13 Inact(NotElig)
* UCAT.ATST12 Inact(NotElig)
At this point, having verified that all sysplex members have successfully * MCATICF.VA81CAT Inact(NotElig)
connected to the ECS structure, the automatic add of all eligible catalogs was * UCAT.ADB210 Inact(NotElig) *
requested through the command F CATALOF,ECSHR(AUTOADD). The ...
command can be issued from any member of the sysplex and causes ECS to
be started for all those catalogs whose cluster contains the parameter
ECSHARING. Again, the new status can be verified by issuing the
ECSHR(STATUS) modify command.

To revert back to not using ECS, the modify command ECHR (DISCONNECT) must be issued on every member of the sysplex. Failing to
do so could lead to inconsistencies if one catalog is not using ECS on all members of the sysplex. Because such a situation could compromise
catalog integrity due to different protocols being simultaneously used (ECS and VVDS), ECS would reject catalog request on some systems,
causing data sets not to be found on those members not using ECS protocol.

Note that the command ECSHR(AUTOADD) must be issued again after all members of the sysplex are brought down. In other words, at
IPL time each member checks whether the active members of the sysplex are using AUTOADD and, if so, enables automatically all eligible

1
catalogs on the IPL'ed system. If the system being IPL'ed is the first in the sysplex, this piece of information is no more available and the
command must be re-issued.

Enhanced Catalog Facility connects to its structure SYSIGGCAS_ECS through IXLCONN with ALLOWREBLD=NO which means that, in
order to move these structures to a different coupling facility, ECS must be temporarily stopped on all systems on which it is active.

All the possible catalog statuses and their actual meaning, in response to the modify catalog command <ECSHR(STATUS)> , are listed in the
explanation for message IEC380I.

You might also like