| Bshells for Parallel ProcessingYou may divide the planning run over multiple bshells in LN, this allows parallel
processing of plan items. Each bshell has its own processing capacity, so when multiple
bshells are in use, the total processing capacity for planning run
increases. The bshells are assigned plan items starting from the top
planning level, until the complete level has been assigned and calculated. On
each pass an average of 500 plan items is assigned to a bshell. Lower level
plan items are only assigned to a bshell after the above level has been fully
assigned. The number of plan items that is allocated to each bshell is
set in the Workload per Server field of the Performance Parameters (cpcom0100m000) session. Setup To setup multiple bshells for parallel
processing: - Select the Parallel Processing check box in the Generate Order Planning (cprrp1210m000) session.
- Select the relevant session from the Session field in the Performance Boosters (tcmcs0597m000) session.
- In the Value field give the number of bshells that are
running.
-
To optimize performance the parameters in the Performance Parameters (cpcom0100m000) need to be set. Enter the refresh interval for the progress bar when
running the Generate Order Planning (cprrp1210m000) session. Enter the maximum number of plan items that is
included in each batch distributed over the parallel bshells. If multiple bshells are used, LN uses a dynamic
calculation to determine the optimal workload per server. The workload is used
as the upper limit for each of the bshells. A workload of 500 plan items per
bshell on average is recommended. If this check box is selected, the routing composition of each item is
used as a parameter for workload distribution over the available
bshells. Note The more operations a routing of an item is
composed of, the more weight it has. If this check box is selected, performance increases in case of large
differences in the number of operations between the different items. If this
situation doesn't apply, this parameter should be disabled. If this check box is selected, the total workload is equally
distributed over the available bshells.
Note The planning process can use multiple bshells, but the
transfer process can only be implemented in one at a time. | |