CloudTran & Coherence

Tap into the full value of Oracle® Coherence

So far, systems requiring high throughput and low latency have used the Oracle Coherence in-memory data grid to cache and scale. This incremental change to architecture has moved data one layer ahead and closer to the request as a quasi-in-memory database.

Turning this architecture inwards and utilizing Coherence to provide ACID transactional capabilities requires that grid data be persisted while maintaining high performance to incoming requests. Currently, applications either take an slow down waiting for updates to disks, or they sacrifice ACID and do asynchronous write-behind. Distributed databases and 2-phase commits make the dilemma even worse.

Introducing CloudTran for Coherence

CloudTran is the industry's only solution for managing the complete data layer in distributed computing environments. We integrate Coherence with back-end data stores asynchronously AND provide ACID transactionality across the entire architecture. This means you can now have applications that run at "grid speed" but that still maintain reliable backup of all data in a disk-based data store.

The CloudTran system runs alongside Coherence and is integrated with TopLink. No special coding is required to use CloudTran; just a few entries to your persistence.xml file and one annotation change on entities. Since CloudTran acknowledges the Coherence grid as the system of record, all transactions are committed in the grid nodes first. Then, committed transactions are sent in parallel to multiple data stores asynchronously for persistent backup.

Aren't Distributed Transactions Slow?

It's true that the old 2-Phase Commit ("2PC") protocol is slow, and it certainly wasn't designed for today's web architecture. CloudTran does not use 2PC and instead has invented a new way of handling ACID-property transactions across a highly distributed architecture. You can read more about the genesis of our 2PC replacement in this technical whitepaper.

Why is CloudTran so fast

  • Because CloudTran maintains its own log files for all transactional data, there's no need for a "prepare" phase.
  • Since CloudTran recognizes the Coherence grid as the system of record, all constraint processing occurs there first which allows commits to back-end databases to happen in parallel and be re-ordered for best performance.
  • Unlike 2PC schemes where everything waits if one participating database is unavailable, the CloudTran approach accommodates "eventual" consistency and can continue operating if a database is down, buffering updates in the grid until it comes back up.

CloudTran for Coherence Features

  • Data persistence - CloudTran writes all committed transactions in the Coherence grid to distributed data stores including Oracle and JDBC. A single grid transaction can include data that spans heterogeneous data stores.

  • Distributed transactions - CloudTran coordinates transactions that modify data stored in multiple Coherence grid nodes. Persistent data stores can be sharded across LANs, WANs, data centers, public clouds, and hybrid clouds.

  • Scalable architecture - As greater performance is required, additional Coherence nodes and persistent data shards can be added and data distributed across these entities. With just a configuration change, CloudTran will manage data in its new location with no modifications necessary to the application. In fact, in a running system, if you need more power, you can just start-up more nodes. The transaction manager and data grid scale up together.

CloudTran Architecture With Coherence

Click to enlarge

© Copyright 2015 CloudTran, Inc.| All Rights Reserved.