Parallel processing concept

To improve processing speed, some planning runs can run in parallel on multiple processors (CPUs). To do this, set these sessions to run in multiple bshells. Each bshell can run on its own CPU.

This section describes the general concept. For information on the setup, refer to To set up parallel processing for planning.

Application of the parallel processing concept

The sessions for which parallel processing is available include these planning sessions:

  • Generate Order Planning (cprrp1210m000)
  • Initialize, Roll, and Update Scenario (cprpd4200m000)
  • Generate Pegging Relations (cprrp0240m000)
  • Generate Planned Supply based on Forecast (cpvmi1211m000)
Note

The process that transfers planned orders to the execution level can only use one bshell.

Distribution of items across bshells

The order planning process takes the plan item's phase number into account. The planning of plan items with phase number 1 cannot start until all plan items with phase number 0 have been planned; otherwise, some dependent demand might be missing for the second level.

If you use multiple bshells, LN completes the following steps:

  1. LN dynamically calculates an optimal workload for the bshell, but the Workload per Server field in the Performance Parameters (cpcom0100m000) session restricts this number to a maximum, such as 500 different plan items.

  2. LN assigns a batch of plan items with phase number 0 to the first bshell. LN assigns a second batch of plan items with phase number 0 to the second bshell, and so on; in this way, all plan items with phase number 0 are distributed across the bshells.

  3. To plan the assigned plan items, all available bshells work in parallel.
  4. When all plan items with phase number 0 have been planned, LN assigns batches of plan items with phase number 1 to the bshells, followed by the plan items with phase number 2, and so on.
Performance aspects
  • Effect on CPU: Yes
  • Effect on database growth: Not applicable

Related topics