Creating a RECON data set

Create a RECON data set using the DEFINE CLUSTER command.

The keywords that are recommended when defining a RECON VSAM KSDS are discussed in the following list. Information on all the keywords can be found in the z/OS® DFSMS Access Method Services for Catalogs.

CONTROLINTERVALSIZE

The values used with this keyword affect the total amount of storage used by DBRC for VSAM and internal buffers. DBRC uses the Local Shared Resources (LSR) option of VSAM to process the RECON data sets. If the number of index and data buffers created by DBRC is allowed to default, the amount of storage used for RECON buffers is:

(60 X index_ci_size) + (120 X data_ci_size)

This amount of storage is used when the index and data CI sizes are the same for all RECON data sets. You can change the default number of index or data buffers used by DBRC in an online or batch environment with the DSPBUFFS Buffer Size Specification Facility.

DBRC uses one set of internal buffers for each RECON data set to build logical records from smaller segments. The size of these buffers depends initially upon the CI size or the VSAM maximum record size (LRECL), whichever is smaller. After the initial allocation of these buffers, the logical record buffers grow over time as the largest logical record on the RECON data set grows. Because only the initial allocation of buffers is determined by the VSAM data set definition, the storage allocated to internal buffers cannot ultimately be affected by changes to either the CI or the record size.

DBRC divides its own records into segments, each of which is always smaller than a single control interval and which is seen by VSAM as a complete physical record. VSAM record spanning is not used. Segmenting allows a logical RECON record to be as large as 16 MB independent of the VSAM RECORDSIZE parameter.

Recommendation: Initially set your minimum CI size to a minimum of eight KB. The allowable CI size is affected by the value you select for RECORDSIZE. Also, ensure that the smallest data CI size exceeds the largest index CI size by at least 2048 bytes. Failure to do so can seriously degrade your DBRC performance.

Related reading: See IMS Version 15.2 Exit Routines for further details about using the Buffer Size Specification Facility (DSPBUFFS).

CYLINDERS

Specifies the amount of space to allocate to the cluster.

FREESPACE

The default values of FREESPACE(0 0) must not be used. While you are entering initial information in the RECON data set, you must specify a high control-interval percentage (for example, 70%) as free space. Later, you can lower the percentage with an Access Method Services (AMS) ALTER command.

INDEXED

Specifies that the cluster is defined for key-sequenced data.

KEYS

KEYS(32 0) is required.

LOG

When parallel RECON access is used, let DBRC manage the LOG parameter on the AMS DEFINE CLUSTER command. When parallel RECON access is enabled, DBRC sets the value of the LOG parameter to UNDO (meaning that the data sets are recoverable) and sets the ACCESS= field in the RECON header to PARALLEL. When parallel RECON access is disabled, DBRC sets the value of the LOG parameter to NONE and sets the ACCESS= field in the RECON header to SERIAL. You should specify the appropriate value for the LOG value whenever you delete and redefine a RECON data set. For example, if you are using parallel access, you should specify LOG(UNDO)

NAME

Defines the cluster's entry name and is required.

NONSPANNED

See RECORDSIZE.

RECORDSIZE

DBRC always writes physical VSAM records that are less than the CI size in length, even though the logical RECON records can be a long as 16 MB.

Recommendation: Set the maximum record size to be, at most, equal to the CI size minus seven bytes and ensure that NONSPANNED is specified. For example, if the CI size was defined as 8192 bytes, then RECORDSIZE (4086,8185) is appropriate.

SHAREOPTIONS

SHAREOPTIONS(3 3) must be specified. The first value is required with single-host processors. Both values are required with multiple-host processors.

SPEED

This is recommended because the initial load is faster.

NOWRITECHECK

Avoid using WRITECHECK. It can degrade the RECON data set I/O performance. Using dual RECON data sets eliminates the need for WRITECHECK.

Recommendation: Do not use authorization keywords because frequent operator prompting results.