Regarding the implementation - without knowing much about the algorithm - I would guess you would at minimum need to implement a new scheduler - see here <https://github.com/apache/hadoop/tree/trunk/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-resourcemanager/src/main/java/org/apache/hadoop/yarn/server/resourcemanager/scheduler> for some examples (fair/capacity/fifo). You may also be able to modify one of the existing schedulers if the new algorithm isn't too different. You would likely also need to change the MRAppMaster <https://github.com/apache/hadoop/blob/trunk/hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/MRAppMaster.java> and the Mapper <https://github.com/apache/hadoop/blob/trunk/hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapreduce/Mapper.java> and Reducer <https://github.com/apache/hadoop/blob/trunk/hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/src/main/java/org/apache/hadoop/mapreduce/Reducer.java> classes to work with this algorithm.
HTH, Hariharan On Sat, Apr 22, 2017 at 7:04 AM, Ahmed Altaj <[email protected]> wrote: > Hi, > > I'm trying to deploy PRISM algorithm which proposed in IEEE paper title > "prism: Fine-Grained Resource-Aware Scheduling for MapReduce", for phase > level scheduling work in homogeneous cluster. > I'm stack with first step how to implement this algorithm and configure > it, which parts i have to change. > > Best Regards... > Ahmed >
