ETD

• building-blocks
• distributed
• distributed-memory
• FastFlow
• runtime

In recent years, the large volumes of data and stringent bandwidth and latency requirements have intensified the need for new scalable algorithms and programming interfaces for distributed memory platform. The thesis aims to introduce a distributed-memory runtime system to FastFlow’s building-block. This RTS allows to split a FF application in multiple processes deployable to different interconnected machines (e.g., clusters). The experimental evaluation reports considerable performance gains with respect to shared memory application which saturate all the resources of a single machine.