|
Our approach depends on taking the historical system information and feed it to
an engine where we try not only to keep latency, reliability, and storage within
constraints, but in addition we try to optimize latency, reliability, and storage
in order to find different trade offs between these objectives. Since we are using
more than one objective, we are doing multi-objective optimization.
A solution in our system is a combination of nodes that will hold replicas of
a given data-item. We use a simple network configuration in which a value of 1 means
hosting a replica and the value of zero means no hosting of a replica. Fig. 4 shows
the binary representation of one chromosome for a system of n nodes and m dataitems.
The first row has nodes labels, the second row has data-items labels, and
the third row has the general binary representation of the chromosome.
|
