Databases Reference
In-Depth Information
FIGURE 4.6: Decomposition of an HGN onto a number hosts in a physical
network.
layer index is also required to indicate the neuron's level in the hierarchy.
These additional indices and parameter increase the complexity of the
processes involved in the HGN recognition scheme. In addition, these
values must be pre-assigned before performing the actual recognition
process. Nevertheless, the change in complexity is minimal.
2. The port number assigned to each neuron must be pre-determined. In
addition, each neuron must be able to calculate the port number of
its adjacent neurons before any communication can occur. This pre-
arrangement requires all neurons to evoke an additional pre-processing
step to identify and calculate the destination ports of its preceding and
succeeding neurons.
3. All of the hosts involved must stay“alive”for the duration of the recogni-
tion process. This effect will create high-interdependency between hosts.
The recognition process will be prone to a total failure if any single host
fails. In addition, the massive and rapid passing of messages between
neurons with different port numbers results in a cost of communication
between hosts.
