Principle (Buffer Position): If all non-bottleneck machines are identical and all buffers are of equal size, then the decrease in line WIP level from adding an additional buffer space will be largest either directly before or after the bottleneck station.
As we discussed in the buffering principle, restriction of buffer size causes blocking, which in turn causes the average WIP to increase. All other things (i.e., buffer sizes and variability) being equal, the station that will cause the most blocking is the bottleneck station. This implies that placing a buffer space in front of the bottleneck will be most effective at reducing blocking. In addition, the worst station to block is the bottleneck (because it causes lost throughput). Therefore, the most valuable blocking elimination occurs from a buffer space immediately after the bottleneck. Thus, depending on the position of the bottleneck in the line, the greatest reduction in line WIP level will come from either adding a buffer space directly before or after the bottleneck.
Example
As an illustration of this principle we use the same
system as for the other buffering principles. The bottleneck is the
third station as before. The following figure plots the average WIP when
the throughput is fixed at 1/1.5 jobs per minute when a buffer is placed
before the bottleneck (between Station 2&3), after the bottleneck (between
Station 3&4) and at a non-bottleneck (Station 1&2).
|
|
|
This figure shows that the smallest line WIP level is achieved when the extra buffer space is added immediately before the bottleneck station. The next lowest line WIP level occurs when the buffer space is added immediately after the bottleneck. The largest line WIP level results from adding the additional buffer space at a non-bottleneck station. This confirms the buffer position principle that the most effective use of buffering is adjacent to the bottleneck.
In this example, adding an extra buffer space before the bottleneck is larger than that from adding it after the bottleneck. The reason for this is that we assumed an unlimited buffer at station 1. As a result, the buffer before Station 3 buffers against starvation by both Stations 1 and 2, while the buffer after Station 3 buffers against blocking by Station 4 only (i.e., because the unlimited buffer in front of Station 1 prevents it from ever blocking Station 3). If all non-bottleneck machines and buffers were identical, then by symmetry the impact of a buffer space upstream and downstream from the bottleneck would be equivalent. In practice, the relative benefit of a buffer upstream or downstream of the bottleneck will depend on the capacity and variability of the stations. If a downstream station is slow or highly variable (e.g., is prone to failures), then a downstream buffer will be helpful in preventing blocking. Conversely, if an upstream station is slow or highly variable, then an upstream buffer will be helpful in preventing starving. In any case, it makes sense to look first for buffering opportunities in close proximity to the bottleneck.