IAS Performance and Tuning Guide

Order Number: AA-H848C-TC

This guide provides advice and general guidelines for optimizing the performance of an IAS (Interactive Application System) system. The goal of tuning is to enhance system throughput and reduce terminal response time, acknowledging that trade-offs between these objectives often exist. The document emphasizes understanding the system's current performance and workload before making modifications.

Key areas for performance tuning are detailed across several chapters:

  1. System Generation: Careful selection of features during system generation is crucial. Excluding unnecessary components, such as certain memory management directives, terminal handler features (bells and whistles, escape sequences, specific interface definitions, dial-up, block mode support), or batch streams, can significantly save memory. Adjusting the size of the System Communication Area (SCOM) and its node pool is vital, as insufficient SCOM can lead to system hangs. Configuring appropriate partitions (user-controlled, system-controlled, or timesharing) for tasks, especially real-time applications, impacts performance.
  2. Efficient Memory Use: Maximizing available memory for user tasks improves response time. This involves judiciously installing tasks and loading device handlers; rarely-used tasks/handlers should not be permanently resident in memory.
  3. Effective Disk Use: Disk performance is critical for I/O-intensive applications. Recommendations include redirecting spooler and work files from the heavily-used system disk to less frequently accessed or faster fixed-head disks. Optimizing disk usage further involves balancing files across multiple disks, strategically placing index files, and using memory-resident overlays for faster access (though this consumes more memory). Regularly releasing free space and using the Disk Save and Compress (DSC) utility to defragment files also improve access times.
  4. Swap Files: Efficient swap file management directly impacts performance. It is recommended to use fast fixed-head disks for primary swap files and dedicated controllers/units for swapping disks. The document provides a "swap ratio" calculation to assess swapping efficiency and suggests improvements like adding memory, limiting users, or stopping tasks if the ratio is too high. Specific tuning for MSCP disk and T/MSCP tape configurations, including using port servers and multiuser handlers, is also covered.
  5. File System Tailoring: The file system's configuration significantly affects IAS performance. Tuning involves selecting the appropriate Ancillary Control Processor (ACP) task (FCP.TSK for small systems, BIGFCP.TSK for multi-user, RESFCP.TSK for memory-resident overlays), and potentially using multiple ACP tasks for concurrent file operations across different disks. Adjusting file system data areas (FCPCOM and ACP task private areas) reduces reliance on the SCOM node pool. Reducing disk accesses is achieved by setting parameters at volume initialization or mount time for the directory LRU cache, default file extension size, and window block size.
  6. IAS Scheduler Tuning: The scheduler balances response time and throughput for timesharing/multiuser systems by managing tasks across scheduling levels (interactive, I/O-bound, compute-bound, batch). Tuning involves adjusting quantum parameters (CPU time allocated), promotion time (between scheduler promotions), and batch scheduling parameters (batch quantum, time between batch schedules). Specific strategies are provided for systems with rarely-used batch, frequently-used batch, compute-bound tasks, I/O-bound tasks, and real-time systems. Real-time tasks benefit from higher priorities and potentially separate partitions.

The document concludes with a practical example of tuning an IAS system, demonstrating how adjustments to scheduling parameters (allocation factor, promotion time, batch quantum) improved throughput for I/O-bound tasks. It advises making one parameter change at a time to accurately assess its impact. A workbook is included for system managers to record tuning changes.

AA-H848C-TC
May 1990
64 pages
Quality

Original
2.4MB

Site structure and layout ©2025 Majenko Technologies