The company, part of SBI Holdings, had more than 2.6 million securities accounts as of March 2013, processing transactions in the United States, China (Hong Kong), South Korea, Russia, Vietnam, Indonesia, Thailand, Singapore and Malaysia.

The company offers transactions through its website, app and mobile site.

With 1,312 funds available in its trading system, it is a mission-critical system.

As a result, downtime not only results in lost revenue, but it can also lead to serious public issues and adversely affect customer satisfaction.

Expanding the multi-site system, the database is still the bottleneck. SBI Securities' online trading system was built using the common three-layer architecture at the time. These three layers were the Web server, application server and database server.

As the company's number of securities accounts grew rapidly, the technical team began to discover system performance problems and realized that the database server had become a bottleneck.

To solve this problem, the team decided to deploy a duplicate system specifically to serve new customers.

However, maintaining multiple sites and systems adds burden from a risk, complexity, capital cost, and operational overhead perspective.

SBI needed new ways to extend the data layer.

SBI needed to find a solution that would maintain low latency even as load increased and could scale out as accounts and transaction volume grew.

The company also wants to reduce costs by removing the replication systems it originally created to enable expansion.

Of course, SBI hopes that data will have redundancy, recoverability, reliability, consistency, security, etc.

After starting to work with NRI Financial Solutions and Hitachi, Ltd., SBI's technology team realized that other options did not meet their needs and ultimately selected Pivotal GemFire ??after multiple comparisons.

Building an in-memory distributed data grid with Pivotal GemFire ??When major systems integrators and application developers NRI and Hitachi, Ltd. began defining system requirements, they looked at in-memory systems and distributed caching solutions.

They believe that only these two architectures and scenarios are suitable for SBI.

They then began evaluation and testing to validate the performance of multiple solutions.

They first tried a competitor's product, and while performance improved, they found some limitations in scale-out configurations.

Additionally, they were unable to port existing applications due to lack of API compatibility.

After testing Pivotal GemFire, the evaluation team found that the product outperformed other systems and could scale out.

While existing applications still needed to be converted, they saw that GemFire ??could deliver better performance at a lower cost.

Pivotal's engineering team worked with NRI and Hitachi, Ltd. to go from build to production deployment in just a few weeks.

When the system was launched in January 2011, only 150 servers were deployed. Compared with the previous 400 server system, the scale was reduced by nearly 1/3.

They also reduced the size of the data center in five areas, and an assessment performed by SBI after the project showed that cost reduction goals were achieved.

In addition, the system can now process twice as many transactions as before, while reference latency has been reduced to 1/20 (now 0.05 seconds compared to 1 second) and execution throughput has increased three to four times.

Processing loads with reduced latency compared to traditional databases Pivotal GemFire ??is an in-memory distributed data platform that can be used as a primary data store or cache for existing databases and hosts, dramatically accelerating the processing of all types of data through parallel processing.

SBI uses GemFire, a wireless sharing architecture, which utilizes CPU, memory, network and hard disk to achieve extremely low latency under increased load.

"When you understand how Pivotal GemFire ??distributes and replicates data and adds dimensions in memory faster than disk, you quickly realize why this data platform model outperforms traditional solutions," said Ito. "Disk cannot deliver

The speed that memory can provide, centralized data cannot run in parallel like distributed systems. Moreover, running on commodity hardware can also bring financial advantages, while simplifying the system and reducing risk and cost compared to other solutions.

solution, Pivotal GemFire ??requires less space, and this is the case at SBI.