| Zhang and Yum compared four channel assignment |
- Fixed Channel Assignment (FCA),
- Borrowing with Channel Ordering (BCO),
- Borrowing with Directional Channel Locking (BDCL) and
- Locally Optimized Dynamic Assignment (LODA).
With respect to uniform offered traffic, their results showed that BDCL had
the lowest blocking probability followed by BCO, LODA and FCA. With non-uniform
offered traffic, the relative performance of the four methods was the same with
the exception that in this case, LODA performed better than BCO. It makes sense
that the ordering for BDCL, BCO and FCA was as found. Indeed, BDCL was
specifically designed as an improvement over BCO and BCO was designed as an
improvement over FCA [Zhang, Elnoubi]. The fact that the performance of LODA
varies under uniform versus non-uniform traffic is rather interesting however.
The reason behind this phenomenon is that LODA provides optimal channel
allocation only in local regions. Given non-uniform traffic which consists of
dense regions in certain local areas, LODA will accommodate these regions of
high traffic offering. However, in a global sense, the LODA algorithm will not
necessarily provide the optimal allocation. With uniform offered traffic, LODA
does not have any regions with peak traffic to optimize; i.e., no local regions
within which the benefits of LODA can be realized. Furthermore, with respect to
the entire region, the optimization is generally not optimal in a global sense.
The result is that with uniform traffic, LODA does not have any advantage to
offer over BCO. From the previous discussion we see that one general result of
all of the comparisons is that dynamic channel allocation outperforms fixed
channel allocation for low blocking rates (below 10% in most cases). Blocking
rates above 1% or 2% are generally not tolerated. This is generally an accepted
guideline throughout the telecommunications industry and we will adhere to this
design constraint as well.
|
