| Hybrid Channel Alloction Scheme Continue... |
A natural extension of channel borrowing is to set aside a portion of the
channels in a system as dynamic channels with the remaining (nominal) channels
being fixed to specified cells. If a cell requires an extra channel, instead of
borrowing the channel from a neighboring cell, the channel is borrowed from the
common "bank" of dynamic channels. An important consideration in hybrid systems
of this type is the ratio of dynamic channels to fixed channels. Analysis by Cox
and Reudlink [Cox - 1973] showed that given ten channels per cell, an optimum
ratio was 8 fixed channels and 2 dynamic channels. In general, the optimum ratio
depends upon the traffic load [Zhang]. In addition to BDCL, a second channel
allocation method was presented by Yum and Zhang [Zhang]. Referred to as Locally
Optimized Dynamic Assignment Strategy (LODA), this method is best described as a
purely dynamic channel allocation procedure as opposed to a hybrid method. In
this strategy there are no nominal channels; all channels are dynamic. When a
given cell needs to accommodate a call, it chooses from among the bank of
available channels according to some cost criteria. The channel with minimum
cost is assigned. In a general sense, the cost is a measure of the future
blocking probability in the vicinity of the cell given that the candidate
channel is assigned. A more detailed description of the cost function will be
addressed below.
Dynamic Channel Reassignment
Similar to the goals of dynamic channel assignment is the process of Dynamic
Channel Reassignment (DCR). Whereas a DCA scheme allocates a channel to an
initial call or handover, a DCR system switches a cell's channel (that is
currently being used) to another channel which is closer to the optimum
according to frequency reuse or other cost criteria. Thus, for example, a user
communicating with channel n may be switched to channel m during
the middle of her/his call if channel m is a more efficient use of the
available bandwidth from a frequency reuse point of view. Philosophically, DCR
is equivalent to DCA.
Simulation and Comparison of Channel Allocation Schemes
A great deal of work is available comparing various realizations of channel
allocation schemes [Cox, Elnoubi, Jiang, Katzela, Yue, Zhang]. In comparing
performance, typical system metrics include blocking probability of new calls
and blocking probability of handover calls. These metrics are written as
functions of offered traffic (where the traffic may be written in a variety of
forms). It is generally assumed that a blocked new call is preferred over a
blocked hand-off call. The idea being that with a blocked hand-off, users are
forced to terminate communication in the middle of their session. If this
blocking happens at a particularly inopportune time, the results could be
disastrous (e.g., business partners cut off in the middle of a vital
negotiation). In the case of a blocked new call, at least the business
negotiation hasn't started and the involved parties aren't interrupted. Blocking
probability is an important metric throughout the field of queueing theory and
in the case of M/M/1 queues, the Erlang-B formula is often used
for analysis of blocking probability. Because blocked calls can be very
disconcerting, systems are typically designed to have blocking probabilities of
no more than 1% or 2%. This is consistent with the assumption of small offered
traffic loads.
|
