In my private opinion, = What's difference of the two projects? =
Hama pursues the general BSP computing engine, so it provides an chance of any additional programming models on top of BSP model. For example, Hama can be used for Pregel-like graph computing model, Machine Learning algorithms, and MRQL's complex query processing. Giraph is different in the sense that it focused only on "vertex-centric graph model" and it runs as a map-only job. So, it have a dependence of Map/Reduce framework. = What's the HAMA's target in the future? = A scalable and fast distributed computing engine for massive scientific computations such as matrix, graph and network algorithms. Also, we're trying to support C/C++ interface, easy-to-use query language (with MRQL project), and GPGPU acceleration. = HAMA's advantage VS Giraph at present? = In terms of graph computing qualities, Hama's graph package and Giraph support almost same APIs because we cloned a Google's Pregel. One different thing is a VertexInputReader for parsing the structure of Vertex from the arbitrary input records. And, I think it's hard to say at this stage which is the fast and reliable project. On Wed, Jul 3, 2013 at 12:33 PM, Changguanghui <[email protected]> wrote: > Hi, all： > > First, congratulations on the release of new version of HAMA！ > I have noticed HAMA for half year, but there are some questions about > HAMA's direction. For example, could the HAMA's team have discussed another > project which is called Giraph? and it focus on large-scale graph processing > too. > I just want to know What's difference of the two projects, and what's the > HAMA's target in the future, and HAMA's advantage VS Giraph at present? > Maybe, HAMA's team can give some advices to users like me, so we can make > a decision to choose a project as our work platform. > > Best Regards > > Chang -- Best Regards, Edward J. Yoon @eddieyoon
