Estimate - Approximation

Enhanced Definition

In the mainframe context, an estimate or approximation refers to a reasoned judgment or calculation of a future value, quantity, or duration, often made with incomplete information. It is a prediction used for planning, resource allocation, and decision-making, acknowledging inherent uncertainty. In the mainframe context, an estimate or approximation refers to a reasoned judgment or calculation of a quantity, cost, time, or resource requirement for a task, project, or system component, often made in the absence of complete or precise data. It provides a foundational basis for planning, resource allocation, and decision-making within z/OS environments.

Key Characteristics

- Uncertainty: By nature, estimates carry a degree of uncertainty and are not guaranteed exact values, often expressed as a range (e.g., "50-70 CPU seconds") or with a confidence level.
- Basis in Data/Experience: They are typically derived from historical performance data, expert judgment, benchmarking, or simplified models of system behavior and resource consumption.
- Purpose-Driven: Primarily used for project planning, capacity planning, resource provisioning (CPU, memory, storage, I/O), and setting expectations for performance or delivery.
- Refinable: Estimates can and should be refined as more detailed information becomes available, moving from high-level conceptual estimates to more detailed, accurate ones.
- Resource-Specific: Often pertains to specific mainframe resources like CPU time consumption, elapsed time for batch jobs, DASD or tape storage requirements, or CICS transaction rates.

Use Cases

- Batch Job Performance: Estimating the CPU time, EXCP count, or elapsed time a new or modified COBOL batch job will require to complete, crucial for scheduling and workload management.
- Storage Allocation: Approximating the DASD or tape storage space needed for new datasets, VSAM files, or DB2 table spaces, impacting storage provisioning and management.
- Project Planning: Estimating the effort (person-days/weeks) and duration for developing, testing, and deploying a new mainframe application or significant enhancement.
- Capacity Planning: Approximating the future MIPS or MSU requirements for a z/OS LPAR based on anticipated workload growth, informing hardware upgrade decisions.
- CICS Transaction Volume: Estimating the peak transaction rates and average response times for a new CICS application to ensure adequate region sizing and resource availability.

Related Concepts

Estimates are fundamental to Capacity Planning and Performance Tuning, providing the initial data points for resource allocation and identifying potential bottlenecks. They are critical inputs for Project Management, influencing timelines, budgets, and resource assignments for mainframe development and maintenance. Accurate approximations help in setting realistic Service Level Agreements (SLAs) and managing expectations for system performance and availability. They also guide Workload Management by informing job scheduling and priority settings within JES2 or JES3.

Best Practices:

Document Assumptions: Clearly state all assumptions made when generating an estimate, as these are critical for understanding and validating the approximation.
Use Multiple Techniques: Employ a combination of historical data analysis, expert judgment, analogy, and parametric modeling to cross-validate estimates.
Involve SMEs: Engage experienced mainframe architects, developers, and system programmers (Subject Matter Experts) who have deep knowledge of the system and workload.
Iterative Refinement: Treat estimates as living documents, refining them continuously as project phases progress, requirements become clearer, and actual performance data is collected.
Add Contingency: Include a contingency buffer (e.g., 10-25%) to account for unforeseen issues, scope changes, or inaccuracies inherent in the estimation process, especially for early-stage estimates.
Track Actuals: Compare actual resource consumption and project durations against estimates to improve future estimation accuracy and identify areas for process improvement.