Technical differences between Titan/JanusGraph and HugeGraph
What are the main technical differences between Titan/JanusGraph and HugeGraph?
Solution:Jump to solution
As answered by the HugeGraph Community:
- We have almost completely refactored the original data structure and serialization logic, and we have provided a complete set of toolchain tools, including visualization, spark/flink import, etc., and we are about to launch a separate flink/spark connector to help connect streaming data/different data sources and bulkload faster and better.
- We used to focus on Hbase and Cassandra as distributed storage backends, but now and in the future we will focus on high-performance RocksDB as the main backend. The current version has initially supported multi-instance Raft + RocksDB, and the next major version will fully support multi-replica + sharding, achieving a high degree of autonomy and control throughout the chain (no longer relying too much on third-party storage). The performance can also be greatly improved, and the single/multi-machine version will be more flexible.
- We have independently developed and open sourced HugeGraph Computer, which is a graph computing system based on the Pregel model with significant optimizations, and natively integrated with HugeGraph, thus bridging the storage and computing systems, providing a one-stop complete "graph service" ecosystem (rather than relying on spark-graphx or other external/limited solutions, which hugegraph also supported earlier)....
3 Replies
Solution
As answered by the HugeGraph Community:
- We have almost completely refactored the original data structure and serialization logic, and we have provided a complete set of toolchain tools, including visualization, spark/flink import, etc., and we are about to launch a separate flink/spark connector to help connect streaming data/different data sources and bulkload faster and better.
- We used to focus on Hbase and Cassandra as distributed storage backends, but now and in the future we will focus on high-performance RocksDB as the main backend. The current version has initially supported multi-instance Raft + RocksDB, and the next major version will fully support multi-replica + sharding, achieving a high degree of autonomy and control throughout the chain (no longer relying too much on third-party storage). The performance can also be greatly improved, and the single/multi-machine version will be more flexible.
- We have independently developed and open sourced HugeGraph Computer, which is a graph computing system based on the Pregel model with significant optimizations, and natively integrated with HugeGraph, thus bridging the storage and computing systems, providing a one-stop complete "graph service" ecosystem (rather than relying on spark-graphx or other external/limited solutions, which hugegraph also supported earlier).
The HugeGraph Community prefers to have user discussion here: https://github.com/apache/hugegraph/discussions so please direct further questions to their discussion forums on github, though I've extended an invitation for community members to participate here on our discord server.
GitHub
apache incubator-hugegraph · Discussions
Explore the GitHub Discussions forum for apache incubator-hugegraph. Discuss code, ask questions & collaborate with the developer community.
Thanks for @spmallette to post info here, if the users want to know more details about the difference, we could consider add more info / an article
Roughly but not rigorously speaking, the relationship between "Titan/JanusGraph" VS "HugeGraph" is more like "LevelDB" VS "RocksDB"