Irregular applications are hard to scale
Irregular applications arise in:
- Social network analysis
- Semantic databases
- Combinatorial optimization
- Circuit simulation
- Machine learning
Low locality and frequent random communication make these applications difficult to scale on commodity clusters.
Grappa makes scaling irregular applications easier
Programming models like MapReduce and GraphLab can be restrictive; for example, parallel recursive algorithms are hard to express.
Grappa provides a familiar multithreaded programming model, making it easy to scale up irregular applications on commodity clusters.
Grappa's core features
Grappa’s runtime system consists of three key components:
- Tasking layer
- Supports millions of threads across the cluster
- Automatically balances load with work stealing
- Distributed shared memory (DSM)
- Provides a single system, shared memory abstraction
- Supports high-throughput small message communication
About the project
Grappa is a project group in the Sampa Group at the
University of Washington.
or join or mailing list to be informed about software releases: