When you are accessing data sequentially, increase performance by increasing the number of data buffers. When there are multiple data buffers, VSAM uses a read-ahead function to read the next data control intervals into buffers before they are needed. Having only one index I/O buffer does not hinder performance, because VSAM gets to the next control interval by using the horizontal pointers in sequence set records rather than the vertical pointers in the index set. Extra index buffers have little effect during sequential processing.

   Index buffers = 3
   Data buffers = 2 * (number of Data CI/CA)

For straight sequential processing environments, start with four data buffers per string. One buffer is used only for formatting control areas and splitting control intervals and control areas. The other three are used to support the read-ahead function, so that sequential control intervals are placed in buffers before any records from the control interval are requested. By specifying enough data buffers, you can access the same amount of data per I/O operation with small data control intervals as with large data control intervals.

When SHAREOPTIONS 4 is specified for the data set, the read-ahead function can be ineffective because the buffers are refreshed when each control interval is read. Therefore, for SHAREOPTIONS 4, keeping data buffers at a minimum can actually improve performance.

If you experience a performance problem waiting for input from the device, you should specify more data buffers to improve your job's run time. More data buffers let you do more read-ahead processing. An excessive number of buffers, however, can cause performance problems, because of excessive paging.

For mixed processing situations (sequential and direct), start with two data buffers per string and increase BUFND to three per string, if paging is not a problem.

When processing the data set sequentially, VSAM reads ahead as buffers become available. For output processing (PUT for update), VSAM does not immediately write the updated control interval from the buffer unless a control interval split is required. The POINT macro does not cause read-ahead processing unless RPL OPTCD=SEQ is specified; POINT positions the data set for subsequent sequential retrieval.

   Index buffers = STRNO
   Data buffers = 3 + STRNO (minimum)