Open Access
American Research Journal of Computer Science and Information Technology
ISSN (Online): 2572-2921
DOI: 10.46568/arjcsit
Pcm- Omars Algorithm: Parallel Computation of Median - Omniscient Maximal Reduction Steps
Associate Prof. Dr. Eng. Database Research Group, Department of Computer Science, University of Rostock
Abstract
The goal of a distributed computation algorithm is to determine the result of a function of numerical
elements, which are distributed in
multi sets.It is known that computation of holistic aggregation functions on
distributed multi sets indeed requires more work than non holistic aggregation functions. But with this article
we will prove that the computation of a holistic function, which named exact median, can be computed efficiently
by providing both a candidate finding and a deterministic location algorithms which computes the position
of exact median, dispelling the misconception that solving distributed median computation through parallel
aggregation is infeasible. Some of most important part in Big Data field is to evaluate massive data values. A
special case in this field is the calculation of
smallest values (specially the median) of distributed multi sets
containing enormous data. Many approximation algorithms and algorithms with iterative or recursive steps
of determination of median give solutions for the computation of median. But firstly sometime approximate value is dangerous for some data evaluation projects or researchs and secondly with other algorithms, the data
blocking time is too long through the iteration or the recursionbetweenglobal node and local nodes. This article
focuses on a solution that gives a best effectively computation for this problem named PCM-oMaRS algorithm.
The PCM- oMaRS algorithm guarantees the maximal reduction steps of the computation of the exact median in
distributed multi sets and proves that we can compute the exact median effectively without needing the usage
of recursive or iterative methods at the global communication level, which reducesthe blocking timemaximally.
This algorithm provides more efficient execution not only in distributed multi sets even in local multi set with enormous data.
