-
Notifications
You must be signed in to change notification settings - Fork 4
In Memory Storage Backends
Matthias Broecheler edited this page Nov 18, 2013
·
2 revisions
All of the primary storage backends for Titan persist the data durably to disk.
This is has two primary benefits:
- Parts of the graph that are not accessed frequently can be held on disk and not in RAM which significantly reduces the cost of storing large graphs.
- The data is persisted durably and can be recovered in case of single machine failure.
However, if the entire graph is frequently accessed and low latency access is required, Titan additionally supports in-memory storage backends. Those hold all data in-memory and thereby avoid the additional complexities and access cost of disks.
| Name |
storage.option |
Consistency | Availability | Scalability | Replication | Persistence |
|---|---|---|---|---|---|---|
| Hazelcast | hazelcast | transactional consistency | highly available | linear scalability | yes | in-memory |
| Infinispan | infinispan | transactional consistency | highly available | linear scalability | yes | in-memory |