| Budget Amount *help |
¥4,290,000 (Direct Cost: ¥3,300,000、Indirect Cost: ¥990,000)
Fiscal Year 2013: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2012: ¥650,000 (Direct Cost: ¥500,000、Indirect Cost: ¥150,000)
Fiscal Year 2011: ¥2,990,000 (Direct Cost: ¥2,300,000、Indirect Cost: ¥690,000)
|
|---|
| Research Abstract |
To improve MapReduce, a popular large-scale distributed programming framework, we developed a high-reliable framework and techniques for accelerating a MapReduce application.
Firstly, we extend an existing MapReduce framework to improving fault-tolerance. In the existing framework, the master computer is the single point of failure. Our extended framework provides multiple master computers and switches them to avoid a system failure.
Secondly, we developed optimizing techniques for intermediate data processing to improve application performance. One of our technique reduces the amount of intermediate data by compressing data from an alternative viewpoint. Another technique reduces intermediate data transfer by improving data locality in task assignment.
|
