My research’s goal is to work out a MapReduce (short form: MR) strategy that provide a scalable data processing at a concrete reference mapping algorithm.
A 3 layers MapReduce algorithm is presented, that is capable to distribute workloads across multiple clusters, which is handling the special cases that came in with the 3 layered architecture. These exceptional case are for example: losing clients and therefore its already made computation, flexible client joining and disconnecting, dynamically changing task scheduling, or handling cases when there isn’t any available processing unit (client). In addition the implemented MapReduce strategy is optimized for a special pattern matching algorithm, accordingly creating a special scheduling strategy.
I touch upon, the result collection from different clients, that provide information that can be used up in the next part of the multiprocessed system.
In the thesis I mention, another existing distributed processing systems, compared with my technique, explaining my decisions. I discuss which data structures are used and what kind of functions do they provide for handling the actions appropriately at different network I/O operations
The introduced MapReduce strategy, shows a viable speed up for a special error detecting algorithm. We can use up this system for example at: VoLTE (Voice over Long Term Evolution) telecommunication failure detection or any real time multi-processing supported big complexity algorithm.