Kbase P58806: How should I address AS/400 performance problems?
| Autor |
Progress Software Corporation - Progress |
| Acesso |
Público |
| Publicação |
16/10/2008 |
|
Status: Verified
GOAL:
How should I address AS/400 performance problems?
GOAL:
What parameters should be evaluated for AS/400 connection performance problems?
FACT(s) (Environment):
IBM AS/400 (RISC)
Progress/400 DataServer
FIX:
The most important parameters that should be investigated and adjusted accordingly in case of connection performance are:
- TCP send and receive buffers (TCPSNDBUF and TCPRCVBUF) - the default value is 8KB and can be increased up to 8MB on V4R5 and higher. The optimal value depends on the status of the network and usually proves to be around 64KB. Decreasing this value may improve performance as well in situations where many retransmissions occur due to the overrunning of a network adapter.
- *TCPONLY parameter for line description can be set to *YES. It may improve the connection efficiency up to 50% on OS/400 V4R4 and higher. However it has to be remembered that this option can be set only when no APPC applications use the same interface an no SNA devices are utilized (like printers). In short, only TCP/IP network is in place.
- MTU - Maximum Transmission Unit should be set to its maximum value, which is *LIND.
- Message Buffer size (-Mm) of the client connection. The default value of 1KB is usually too small. This client startup parameter needs to be adjusted according to the average record length in the database. The optimal value usually ranges from 4KB to 8 KB. The idea is to pack as many messages into single message as possible. It may increase the advantage of the prefetch mechanism
- Switch on the record compression by setting the connection parameter -Dsrv COMPRESS=1
There are many other areas that need to be investigated as well: separate subsystems for Progress jobs, storage pools configuration, system value regarding performance adjustment (QPRFADJ), job management configuration (run priority, timeslice, etc). The optimization process is usually based on individual assessment and analysis. Performance Management may be extremely helpful in such situations. Under some circumstances the Consultancy Service should be considered as an essential step in problem isolation.
Some of the useful hints may be found in the Progress white paper:
Building High Performance Applications with the Progress/400 DataServer