We have an iSeries 820 with V5R1 OS/400. For the last few days we've gotten complaints from our users regarding slow response times. We never received complaints after the 820 installation. When this randomly happens, the iSeries becomes erratic and even becomes slower then our earlier 500 model.
After close observation, I found that one system's job, by the name "CFINT01", was active at that moment and was "eating" up the complete CPU time (approximately 92%). After moving to the 820, we never found CPU utilization above 3 or 4% even on full workload. However, most of the time, it was found below 1%. This is a very serious issue for us and requires immediate attention.
There are many reasons systems run slow. It's not always simple to know exactly why a single program gets the bulk of CPU resource at any point in time. Proper performance optimization on iSeries for a given resource and goals is NOT a trivial task for the average developer or administrator. If you have serious performance problems you should consider contacting a third-party iSeries performance consultant who can properly analyze your system and give you unbiased recommendations on what to do. In most cases, the consultant will use the IBM performance monitors to collect performance data.
That said, I believe your problem is due to the interactive workload running higher than allowed by IBM, according to the interactive CPW rating for your 820. When that happens, IBM has a "neat"(?) program that kills response time (CFINTxx). It grabs CPU resource in addition to the actual interactive jobs running, limiting the additional work that can be done by both interactive and server workloads. In effect, CFINT has been called an "interactive governor" limiting the processing power of your server. You can create strategies to manage or allocate workload to interactive and server processing to better balance your performance. You can also control tasks manually to a degree.
The actual behavior of your system depends on several factors beyond your specific hardware and workload: OS version, PTFs applied, system configurations and the system values: QDYNPTYSCD, QDYNPTYADJ.
There have been problems at different OS versions, where CFINT didn't function correctly "eating" your CPU. One IBM PTF created a CFINT problem. The fix was to remove the PTF.
Model 720 experiences 85-95% CPU utilization by CFINT01 and interactive utilization is not the cause. System also experienced srcB6000103 on IPL. The cause of the problem appears to be PTF MF25592. Removing this PTF fixed the problem.
MORE INFORMATION ON THIS TOPIC
Ask your Web development questions--or help out your peers by answering them--in our live discussion forums.
The Best Web Links: tips, tutorials and more.
Check out this live Q&A with Dave Slater and Jim Mason, How to best get started with WebSphere Development Tools.