The Apache Giraph team is proud to announce our first release out of
incubation. The release version is 1.0.0 to reflect a lot of hard work
that went into making the code stable enough for production use, memory
efficient, and performant.
Apache Giraph is an scalable and distributed iterative graph processing
system that is inspired by BSP (bulk synchronous parallel) and Google's
Pregel. Giraph distinguishes itself from those projects by being
open-source, running on Hadoop infrastructure, and going beyond the
Pregel model with features such as master computation, sharded
aggregators, out-of-core support, no single point of failure design, and
more.
Here are some highlights of the release:
* Scales out to hundreds of machines easily and hundreds of billions of
edges (memory permitting)
* Efficient use of memory via fast byte-based serialization by default
and can use primitive specific types when better performance is required
* Multithreaded input and computation can take advantage of multicore
machines efficiently
* Simplified vertex API
* Vertex-based and/or edge-based input supported
* Master compute API for handling application-wide logic
* Sharded aggregators for handling large (memory) aggregators
* Easy access to/from Hive tables to integrate with your data warehouse
* Out-of-core graph and messaging support
* YARN support
For release details and download access, please visit:
http://giraph.apache.org/releases.html
Thanks so much to everyone for all their contributions. It is you who
made this release possible! We've also been investing in updating our
website as part of this release (http://giraph.apache.org), more
documentation/updates will be coming in the near future. We expect that
releases will happen more frequently in the future now that we are more
familiar with the process.
Regards,
The Apache Giraph team
