Code

There are totally three algorithm in code/algorithm folder, including PAS, AHP and SJF. And we also change the source code of hadoop yarn resource manger. To use our code. Firstly, you should export the hadoop-yarn-server-resourcemanager to the jar and replace it on the yarn. Then, you can use our code by running the socket_server.py, because the communication between hadoop and python is via socket.

Data

Prediction

Here is data of job completion time under different Resource Utilization Ratio (RUR), which including wordcount, sparkpi and SVM.

Experiment

Here is data of six groups of workloads, including the submission sequence, makespan, turnaround time and RUR.