To your question "But what is the reason to make it comparable to giraph?"
I started a review on availlable graph processing tools, so I found this:
"So, to summarize, what Hame, GoldenOrb and Giraph have in common is: Java platform, Apache License (and incubation), BSP computation. What they differ for: Hama offers BSP primitives not graph processing API (so it sits at a lower level)..."
In a separate project, it is called "hadoop distributed graph space" (HDGS) a Metadata store for graph processing (like Hive warehouse) and a generic workflow framework (compatible to Oozie) is under construction which is running on top of HDFS and HBase in a hadoop cluster.
I would like to see a common way to start these jobs based on HAMA and Giraph or simple MapReduce (for example for preprocessing graph data sets by analysing time series data in a comparable way, so it would be easy to have applications on a higher abstraction level which build workflows by combining different tasks using different computational concepts in one single Hadoop cluster.