Midwest CMG

Each user on shared computing systems desires for his or her job to be processed by the system as quickly as possible. This requires each user's job to be run in an environment where resources (memory, CPU, disk, and network I/O) are fine-tuned to highly optimize the system's response on a per job basis. Such is, of course, prohibitive by conventional methods. This presentation introduces the design of a rapid response neural network method for resource tuning on a per job basis for time-shared computing systems. Future work will yield results from concept demonstration, research, and implementation.

Rapid advances in research and technology now allow data analysis and modeling of extremely complex systems. Methods from artificial intelligence (AI) such as neural networks have been explored to determine how they apply to systems management. In going to the next step-from analysis to action, we believe that software systems can now be automated to

• Analyze measurements being generated by a complex system in real time
• Automatically model its different facets, and then
• Initiate actions automatically to maintain effective service time and resource utilization

Such automation and control is being driven by the demand for enterprise-wide information integration and the need to accelerate development and deployment of complex eApplications. By combining emerging technologies from AI and PM and Workload modeling, data centers can begin to enjoy a new level of availability and guaranteed service levels. Eventually, the automatic configuration and deployment of new servers in real time will provide minimization of spare capacity on individual systems and the reality of capacity on demand across platforms.

This presentation examines the practical aspects of performing a migration to z/OS V1R4, from either OS/390 or prior releases of z/OS. The focus of this presentation is to provide you with as much planning information for a z/OS migration as is possible within an hour, and to position you to implement your migration in the immediate future. Topics will include:

IBM support and availability considerations by release level
IBM coexistence, fallback, and other maintenance considerations
Hardware considerations
Migration issues, such as withdrawn function
Key or major migration tasks
Summary of migration tasks by element
z/Architecture considerations and migration tasks
Migration tools

The behavior of a WLM-managed MVS system is sometimes best understood by looking at the goals rather than workloads. A WLM goal occupies a space of many dimensions. The system administrator is free to burden the goal with an assortment of workloads, assign it a high or low importance, and set it a target that is somewhere between lenient and stringent. In execution, the goal may be met well or poorly, at a cost that may be cheap or dear. Mr. Olcott will review the WLM metrics and present several maps to facilitate navigation through its multi-dimensional space.

One of our board members will raise some issues such as tracking components of poor performance. An example: PeopleSoft, which is distributed, but often has a DB2 mainframe component. When response is slow, where does one look? There is no longer a unified package that will provide the info we previously enjoyed in the mainframe environment. Shouldn’t we encourage the vendors to provide solutions?