The total order based protocol is a lock free commit protocol that relies on the concept of totally ordered delivery of messages which, informally, implies that each node which delivers a set of messages, delivers them in the same order.
This protocol comes with this advantages.
- transactions can be committed in one phase, as they are delivered in the same order by the nodes that receive them.
- it mitigates distributed deadlocks.
The weaknesses of this approach are the fact that its implementation relies on a single thread per node which receives the transaction and its modification from other nodes in the cluster, and the slightly higher number of messages exchanged by JGroups.
Thus, this protocol delivers best performance in scenarios of high contention, in which it can benefit from the single-phase commit and the thread that delivers/applies transactions is not the bottleneck.
Currently, the Total Order based protocol is available only in transactional caches for replicated and distributed modes and it is available in Infinispan 5.3.0 Alpha1.
If you are interested in know more, please take a look at the user documentation where it is explained in more detail how it behaves and how you can configure it. Since this is a recent work, if you find any incorrect behavior please create a JIRA.
Please find below some benchmarks comparing the total order based implementation with the locking based implementation in two different scenarios:
- Contention: 1000 keys in a shared pool, 8 threads per node and 5 writes (on average) per transaction;
- No Contention: 1000 keys in a thread private pool, 8 threads per node and 5 writes (on average) per transaction.