A single set of VTAM® tuning statistics can be enough to indicate how a network is operating. However, these statistics become more valuable as you compare sets of values over time to see trends or the effects caused by changing buffer pool specifications and operands. You should analyze tuning statistics before and after making any change that might affect system performance.

Use tuning statistics to analyze the effects of VTAM host processor usage.

1. Compare the number of attention interrupts with channel READs. If the number of attention interrupts is the same as the number of channel READs, each READ is the result of the NCP sending the host a stand-alone attention. This condition causes increased CPU utilization because the host processor and VTAM must handle each attention interrupt. Conversely, if the number of attention interrupts is small in relation to the number of channel READs, the impact on the host is much less. If the number of attentions is large in relation to the number of channel READs (CHRD or RCH), then this could be a normal condition. Some controllers (for example, those running NCP Version 5 Release 2 and later releases) send attention polls to the primary logical unit during periods of no traffic. VTAM does not count the number of channel reads when no data is transferred.
2. Analyze your IOBUF size. If RDBUF is much larger than IPIU, the value specified for the I/O buffer size might be too small. However, you need to know the message characteristics of your network before you make decisions. Many systems receive the bulk of their inbound data in very short messages and respond with longer outbound messages. For details on choosing an appropriate value for the I/O buffer size, see Guidelines for setting UNITSZ.
3. Analyze data transfer operations between VTAM and an SNA controller. To analyze inbound data transfer, multiply MAXBFRU by the number of channel READs (CHRD or RCHD). MAXBFRU is specified on the GROUP, HOST, LINE, or PU definition statement. The product should be close to the value of RDBUF (the number of read buffers used). Otherwise, data is not being transferred for all the channel command words (CCWs) in the read channel program. Reducing the value of MAXBFRU should solve this problem. However, the value of MAXBFRU times UNITSZ must be at least as large as the largest PIU received. For more details on setting the value of MAXBFRU and UNITSZ, see Guidelines for setting UNITSZ and Setting the MAXBFRU operand.
4. Examine the read attention (RDATN) information that is available in VTAM tuning statistics to understand the efficiency of the inbound data transfer operations. RDATN is the number of times that VTAM, after reading data, is requested by an attention to read more data. This happens when one of the following occurs:
   • The READ channel program CCW string is not long enough to contain all of the data sent.
   • Enough additional data comes into the communication controller during the read operation to cause it to request VTAM to do another read.
   • The controller is short on buffer space and is requesting a read to confirm VTAM receipt of data sent by the controller, so that the controller buffers can be reused.

   Although this coattailing of PIUs is preferable to stand-alone attention interrupts, a large RDATN value is not desirable. To increase the size of the CCW string, increase the MAXBFRU value. For more details on setting the value of MAXBFRU, see Setting the MAXBFRU operand.

5. Analyze the outbound data transfer operation between VTAM and an SNA controller by analyzing VTAM blocking of outbound PIUs. To determine the average number of PIUs for each write operation, divide the outbound PIU (OPIU) count by the number of channel WRITEs (CHWR or WCH). This average number of PIUs indicates the effectiveness of the VTAM blocking algorithm. Increasing DELAY can also increase inbound coattailing. For more information about setting the value of DELAY, see Guidelines for setting DELAY.