In-Memory Computing Delivers Nearly 2 Billion Transactions Per Hour

- Tsipi Erann, Marketing Programs Manager, GigaSpaces Technologies, says: 

When Gresham Computing needed to select an infrastructure for Clareti, their financial transaction control solution, they chose GigaSpaces XAP elastic application platform. According to Gresham, they selected XAP because it had a proven record for reliability, scalability and high availability. But even Gresham had no idea at that point of the surprises yet to come. 

In July 2012, Gresham performed a benchmark running Clareti on XAP at Intel’s computing lab, and the results were outstanding; the tests, which included load and match into a database, performed over 50,000 equity trade transactions per second, or about 200 million transactions per hour.

It’s pretty well-known by now that data in the enterprise is growing at a much faster pace than the ability to process it effectively and in time to get insights for immediate decisions. So these high numbers were greeted with much enthusiasm. However, Gresham believed that they had not yet reached the best possible results given the advances in hardware, including multi-core processors such as the Intel® Xeon® Processor E7 family used for the benchmark, combined with GigaSpaces’ in-memory computing capabilities. The two together enable both extreme performance that supports terabytes of data with near-zero latency, and the scalability and agility that in-memory computing is designed for.

So Gresham took the next step, and performed another series of tests – this time entirely in memory, with no relational database in the stack.

If the previous results were outstanding, these were nearly unbelievable: Clareti was able to process over 500,000 transactions per second, or more than 1.8 billion per hour, using in-memory matching.

For many financial services enterprises, the IT holy grail is gaining the ability to process masses of data in real time, while also maintaining transactionality, scalability and availability. By co-locating the app’s business logic (such as Clareti’s) with the data and messaging – all in-memory, all in a single platform – GigaSpaces places that ability squarely in their reach; ACID transactions, real-time event processing and elasticity of both processing and data, all come at a fraction of the cost compared to traditional approaches.

Of course, financial services organizations are not the only businesses that need the real-time factor for their Big Data processing: e-commerce, transportation and logistics, social networking, SaaS providers and others all need to process increasing volumes and types of data, from multiple sources and for multiple applications – a degree of complexity that is not easily overcome. Or hadn’t been – until now.