This comparison is an outdated comparison. I think Neo4J has improved a lot with the time. But I'm posting this because a person who wants to compare both of these technologies, can get an idea about the aspects they need to focus. If you know something is outdated please feel free to suggest using comments. I'll update the blog post accordingly.
| Feature | Neo4J | Titan |
| License | GPL/AGPL/Commercial | Apache 2 License |
| Commercial Support | Available Advanced: Email 5x10 USD 6000/yr Enterprise: Phone 7x24 USD 24,000/yr | Available (Prices and availability of support not published officially.) |
| Graph Type | Property Graph | Property Graph |
| Storage Backend | Native Storage Engine | Cassandra, Hbase, Berkeley DB |
| Dependin on the requirement we should select Database Backend (eg : Cassandra for Availabilty and Partitionabilty, Hbase for Consistency and Partitionabilty) | ||
| ACID Support | Yes | ACID is supported on BerkeleyDB Storage Backend |
| Has Transactions in Java API. | On Cassandra Eventually consistent | |
| Scalabilty | Can't Scale out like Titan | Owns very good scalabilty |
| can scale like Cassandra if storage backend is cassandra | ||
| High Availabilty | Replication is the only way to have high scalability | Titan is like API because of that Availability of Storage backend is the availabilty for graph database |
| Failover is not smooth | If we are using cassandra with Titan No-Single-Point of failure. Extremely Available | |
| Query Language | Cypher and Gremlin | Gremlin |
| Cypher easy to learn but only suitable for simple queries. | Gremlin has good algorithms to retrieve data in optimal way. (+ More generic) | |
| Graph Sharding | Not Available, under development | Not Available, under development |
| Support for languages | Java/.NET/Python/PHP/NodeJS/Scala/GO | Java |
| Written in | Java | Java |
| Protocol | HTTP/REST | Can expose REST using Rexster |
| Use cases | more than 10 available | 0 use cases exposed officially |
| Number of edges vertices supported | 2^35 (~34 Billion) Nodes (Vertices) | 2^59 Vertices |
| 2^35 (~34 Billion) Relationships (edges) | 2^60 (quintillion) edges | |
| 2^36 (~68 Billion) Properties | ||
| 2^15 (~32 000) Relationship types | ||
| Limitations | Key Index must be created prior to key being used | |
| Unable to drop key indices | ||
| For bulk graph operations we have to use Faunus otherwise storage backends get OutOfMemoryException | ||
| Types cannot be changed once created | ||
| Web Admin | Available | Not Available |
| Embeddable | Yes | Yes |
| MapReduce | - | Yes with Faunus |
| Lucene Indexing Support | Yes | Yes |
| Backups | Yes | Yes (+Titan Parellel backup) |
This link is also very usefull - http://db-engines.com/en/system/Neo4j%3BTitan
one confusion is created....
ReplyDeleteTitan can store up to a quintillion edges (2^60) and half as many vertices.what it interprets???
2^60's half of 60's half??
means titan supports vertices upto 2^30 or 2^59????
Please refer this link http://s3.thinkaurelius.com/docs/titan/0.5.4/limitations.html I think what they mean by it is half of a quintillion. Not something else.
Deletehalf of quintillion means (2^60)/2 then it is 2^59.....not 2^30
DeleteThanks a lot! I corrected it. I didn't see that! :)
Delete