The SAP Workload Analysis Calculator (transaction AL08 and related tools like ST03N) is not a single built-in calculator but a set of performance-monitoring transactions that let you analyze the load (CPU, memory, response time) that users and programs place on an SAP system. You enter a time period and filter criteria, and the system returns aggregated metrics such as average response time, dialog steps per second, and CPU usage. This article explains what these tools actually calculate, how to read the numbers, and how to use them for capacity planning and troubleshooting.
Formula
The Workload Analysis in SAP does not use a single formula; it aggregates performance counters from the SAP kernel and the operating system. The core relationship you can compute yourself is throughput and CPU consumption:
- Throughput (dialog steps per second) = Total dialog steps / (number of hours × 3600).
- CPU time per step (estimated) = Average response time × CPU factor (typically 0.3 for a rough estimate, but varies by workload).
- Total CPU time = Average response time × number of steps × CPU factor / 3600 (to get hours).
Where:
- Dialog steps – a unit of work in SAP; one step is one round-trip between the frontend and the application server (e.g., a screen change).
- Average response time (ms) – the time the application server takes to process a dialog step, excluding network time.
- Users – the number of active users in the system (from
AL08).
The CPU factor is not a constant; it depends on the application (e.g., SAP GUI steps use less CPU than background jobs). The Workload Analysis tool itself shows the real CPU time from the operating system, so you do not need to estimate it manually unless you are doing rough capacity planning.
Worked Example
Using the default values from the calculator above:
- Number of dialog steps: 100,000
- Average response time: 250 ms
- Number of users: 50
Assume the system runs for 8 hours (a typical shift).
Throughput:
100,000 steps / (8 × 3600 s) = 100,000 / 28,800 = 3.47 steps per second.
Estimated CPU time per step:
250 ms × 0.3 = 75 ms = 0.075 s.
Total CPU time:
0.075 s × 100,000 steps = 7,500 s = 2.08 hours.
So the system processed about 3.5 dialog steps per second, and the application server spent roughly 2 hours of CPU time on these steps. This matches the calculator output (3.47 steps/s, 0.075 s per step, 2.08 hours).
Assumptions
- The CPU factor of 0.3 is a rough estimate; actual SAP systems vary from 0.2 to 0.5 depending on application type and code efficiency.
- The calculation assumes the system is the only one using the CPU; in a virtualized environment, other VMs may share the same physical CPU.
- Dialog steps are counted per round-trip; background jobs and RFC calls have different profiles.
What the SAP Workload Analysis Actually Measures
SAP’s Workload Analysis (transaction ST03N) collects data from the SAP statistics records. Each time a user performs an action (e.g., clicking a button), the SAP kernel writes a record containing:
- Response time (total processing time on the application server).
- CPU time (actual CPU consumption by the work process).
- Database time (time waiting for the database to respond).
- Queue time (waiting for a free work process).
- Load time (time to load ABAP programs into memory).
These components are displayed in ST03N under “Workload” → “Analysis.” The tool lets you filter by server, by user, by program, or by time. The key output is a breakdown of where time is spent, which is more useful than a single number.
For example, if you see that database time is 60% of response time, the bottleneck is likely in SQL queries or database locks. If queue time is high, you may need more work processes or a larger instance.
How to Use the Workload Analysis for Capacity Planning
Capacity planning is about predicting when your system will run out of resources. The Workload Analysis gives you historical data to extrapolate from. Here is a practical approach:
- Collect workload data for a representative period (e.g., a full month) using
ST03N. - Identify the peak hour (usually 10:00–11:00 AM) and note the peak dialog steps per second and CPU utilization.
- Compare the peak CPU utilization to the number of CPU cores. If the system is above 70% sustained, you are likely to see response time degradation.
- Use the formula above to estimate how much CPU you need for a projected increase in users or transactions.
- Check the trend: if the average response time is increasing over months while the workload is flat, you may have a performance degradation (e.g., database index fragmentation).
Remember that the Workload Analysis only shows the application server’s view. It does not include network latency or frontend rendering time. For end-to-end response time, you need additional tools like SAP NetWeaver Browser or third-party APM.
Common Mistakes When Interpreting Workload Data
One common mistake is confusing “response time” with “user wait time.” The response time in ST03N is the time the server takes to process a step, not the time the user perceives. The user’s perceived response time includes network round-trip and frontend rendering, which can add 100–500 ms.
Another mistake is ignoring the number of dialog steps per user. A single user can generate many steps per minute; the “users” count alone is not a good workload indicator. For example, a power user doing data entry may generate 10 steps/min, while a casual user may generate 1 step/min. The Workload Analysis shows the step count, so always look at steps per second or steps per user per hour.
Finally, do not use the average response time alone. The average can hide spikes. Use the “distribution” view in ST03N to see the 95th percentile. A system with an average of 250 ms may still have 10% of transactions taking 2 seconds, which is poor for interactive users.
Workload Analysis vs. Other SAP Monitoring Tools
SAP offers several monitoring tools, and it is easy to confuse them:
- AL08 – Shows current active users and their work processes. It is a real-time snapshot, not historical.
- ST03N – The main Workload Analysis tool. It provides historical data (retention period configurable, default 30 days).
- ST04 – Database performance monitor. Shows SQL statement response times, buffer hit ratios, and database locks.
- ST05 – SQL trace. Records individual SQL statements for a specific user or transaction.
- ST22 – ABAP dump analysis. Shows program errors, not workload.
For a complete picture, you would use AL08 for current issues, ST03N for trends, and ST04 to drill into database bottlenecks. The Workload Analysis is the starting point because it tells you where time is spent, which guides you to the next tool.
Costs and Licensing Considerations
The Workload Analysis is included in SAP NetWeaver Application Server ABAP, which is part of most SAP licenses. There is no extra cost for using ST03N. However, the data collected consumes disk space in the database (the statistics records are stored in tables like MONI). The default retention is 30 days, but you can extend it, which increases storage costs. In large systems, the statistics tables can grow by several gigabytes per month, so plan accordingly.
Also, the Workload Analysis itself has a small performance overhead (about 1–2% of CPU) because every dialog step writes a statistics record. This is usually negligible, but in very high-throughput systems (e.g., 1000 steps/sec), you may want to reduce the level of detail (e.g., only collect for specific users).
Frequently Asked Questions
What is the difference between AL08 and ST03N?
AL08 shows a real-time list of users currently logged on and their work process usage. ST03N shows historical workload data aggregated over time. Use AL08 to see who is on the system right now, and ST03N to analyze performance trends over days or weeks.
How do I find the average response time in SAP?
Run transaction ST03N, go to “Workload” → “Analysis,” select a time period and server, and the output shows the average response time in milliseconds. You can also break it down by transaction code or by user.
What is a good response time in SAP?
For interactive dialog steps, a response time below 200 ms is generally good, 200–500 ms is acceptable, and above 1 second is noticeable to users. However, the target depends on the application; background jobs can tolerate longer times.
Why does the Workload Analysis show high DB time but the database is not busy?
High DB time can be caused by inefficient SQL statements that lock rows or by buffer misses. Check ST04 for buffer hit ratios and ST05 to trace the specific SQL. Sometimes the time is spent in the database network or in lock waits, not in actual CPU.
Can I use the Workload Analysis to predict hardware needs?
Yes, but only as a rough guide. Use the historical peak CPU utilization and dialog steps per second to extrapolate. For example, if your system uses 50% CPU at 1000 steps/sec, then 2000 steps/sec would likely require 100% CPU, so you would need to double the cores. But you must also consider memory and database I/O.
Is there a way to automate workload analysis reports?
Yes, you can schedule the report RSSNAPDB or use the standard SAP report RSSNAPDB to generate a workload analysis report in the background. You can also use transaction SE38 to run the report and send the output via email. Additionally, SAP Solution Manager can collect and display workload data from multiple systems.