It is not difficult to find MR models and many are asking whether or not these should replace parallel database systems altogether. Parallel DBMS are available from several vendors and do provide a high-level programming enviornment that is parallelizable. While there are differences, some contend that it is feasible to write almost any parallel processing task as either database queries or MR jobs. According to ACM’s research, using MR systems to carry out work that is best suited for DBMSs often leads to less desirable results as MR is more associated with extract-transform-load (ETL) systems than a DBMS because it can quickly handle large amounts of data off the cuff. It is for this reason that some argue that it complements DBMS technology versus competes with it.
MapReduce and Parallel DBMSs: Friends or Foes?
June 21, 2010