Data that is transferred between a VTAM® application program and another logical unit in the network passes through VTAM I/O buffers. VTAM temporarily holds the data until an application program requests it or until it can be sent to a logical unit.

Figure 1 shows how VTAM uses buffers for input operations.

Figure 1. How VTAM uses input buffers

1. A logical unit in the network can send data to an application program independent of the application program requesting the data. The data is placed in a buffer in the fixed buffer pool of data buffers.

2. One of the following occurs:

   1. If the application program has an outstanding request for data from the logical unit, VTAM moves the data from the buffer in the fixed buffer pool to the specified application program data area. Later, data can be paged out of main storage by the operating system.
   2. If the application program does not request data from the logical unit, VTAM places the data in a data space.
3. When the application program issues an input request, the operating system pages the data into main storage (if necessary) and VTAM moves the data to the specified application program data area.

In general, an application program should be written so that at least one request for input is always outstanding. This reduces the chance of VTAM using excessive amounts of an application program's dataspace storage.

Figure 2 shows how VTAM uses buffers for output operations.

Figure 2. How VTAM uses output buffers.

1. The application program requests data transmission. When VTAM has sufficient buffers in the fixed pool, it moves the data to the fixed pool.
2. VTAM sends the data to the NCP or the logical unit, freeing the buffer.
3. If HPR is used, VTAM holds onto the I/O buffers in pageable storage until the other RTP endpoint acknowledges receipt of the data.

If you notice that your I/O buffers have expanded and you are running out of CSA, it is possible that an application program is not pacing data. Investigate the application programs that have a lot of traffic moving through the I/O buffers to determine whether pacing is being performed. For general information about pacing, see Session-level pacing. For specific information about application program pacing, see Application program pacing.