Turbocharge Your Web Applications

Web applications run the fastest when enough server resources are available to handle the workload. Today's Application Servers will automatically balance computing workloads, but balancing data operations has generally been left up to savvy development teams. To improve overall response time and performance, adding extra servers comes with the chore of distributing data, hard-coding the location of that data, and implementing a transactional model that guarantees data consistency and reliability across all nodes. Not a simple task, to be sure!

Web Data Processing is Different

Using CloudTran WebScale Data, data location is totally independent of the application, resulting in a highly scalable architecture that just about any developer can implement and maintain. CloudTran handles the scatter and gather of data throughout the distribution and provides a new approach to processing data transactionality called "CloudTP".

Unlike traditional distributed transactions, CloudTP commits data within the data grid first and then asynchronously to the data stores. This means fast response times since the application can respond back to the user immediately. Because CloudTran maintains its own transaction logs, there's no need for the application to wait for data store writes or for the cross-checking done by distributed transactions (two-phase commit). CloudTP has many appealing qualities for web-scale applications:

• fast - a CloudTran coordinator node can handle thousands of transactions per second
• reliable - all nodes can have one or more hot backups; all transactions exhibit ACID properties
• scalable - there is no limit to the number of nodes or records involved in a transaction

Furthermore, CloudTP is unique in that transactions are universal in that they span SQL and NoSQL persistent stores as well as messaging queues.

Fast In-Memory and to Fast to Disk

CloudTP can be implemented using two different types of memory-based data architectures. Both require data persistence, and CloudTran will keep everything in synch.

• Complete Database In Memory: For high-performance systems, this is the simplest approach and provides the fastest possible read and write performance. CloudTran provides automatic population of in-memory databases using data from the persistent data stores. If memory nodes fail, CloudTran will automatically reload the data, reapply transactions if necessary, and bring the system back to a consistent state.
• Cache Only What is Needed: Another approach is to use a local cache and bring data into the nodes' memory if and when necessary. This means the read performance of the system will be slower - being determined by the database performance on searches that have to go to disk. To support the caching approach, CloudTran provides lookup/load operations to integrate into a caching mechanism. This approach will have slower reads on average, but create/update performance will be the same as the in-memory method.