Accelerating payment solutions with in-memory computing

By Nikita Ivanov | 13 June 2017

The shift to digital payments is taking place in many forms: bitcoins, mobile wallets, “tap and go” payment transactions, peer-to-peer money-transfer apps and more. Worldwide, the mobile payments market alone has grown from $235 billion in 2013 to a projected value of almost $800 billion in 2017 and over a trillion dollars by 2019.

This shift to digital payments, however, is not without challenges. To ensure a positive customer experience, service providers must be able to process payment transactions in real-time and ensure the highest levels of reliability. They also need to be able to scale capacity ahead of demand and employ sophisticated analytics to help prevent fraud, ensure regulatory compliance, and identify actionable insights from the data they are collecting.

Security is of particular importance to payment solution providers. Digital payment systems open the door to fraud, hacking, and other security risks on a massive scale. New cybersecurity regulations promote greater data protections, but they also add to performance challenges, requiring service providers to:

  • • Implement effective security and encryption techniques for user authentication and protection of transaction data
  • • Comply with complex and evolving security regulations, often in real time
  • • Access huge amounts of structured and unstructured data, such as historical behavior data that can help predict future fraud
  • • Identify and stop fraudulent transactions in real time through predictive modeling, machine learning and other techniques

Another key challenge for service providers is successfully mining the data they collect across a wide range of vendors, services, and payment recipients in order to generate insights that can drive new business opportunities. For example, they can combine payment data with geolocation data, customer history data, social media streams and more, and then analyze this aggregated data to predict possible future customer behavior and better understand how to improve customer satisfaction. They can also analyze real-time payment data to identify in-the-moment opportunities for upselling and cross-selling additional products and services or taking actions, such as suggesting a chat with a representative, that can lead to higher customer retention.

Both fraud prevention and analytics are data-intensive activities, requiring ever greater amounts of data, typically including structured and unstructured data types from a variety of sources, to be processed with unprecedented speed using machine learning and other A.I. techniques. In addition to scale and speed, the underlying technology infrastructure must deliver:

  • • Flexibility and interoperability, with the ability to incorporate legacy, open source and cloud systems and enable payment options using a range of devices and strategies.
  • • Fault-tolerance and high availability, ensuring the level of reliability and recoverability required for all financial data.
  • • Security to ensure customer financial data remains safe

To achieve all this, many payment solution providers are turning to in-memory computing (IMC) solutions. Modernize and Accelerate Payments Solutions with In-Memory Computing, a new whitepaper by GridGain Systems, discusses how in-memory computing is one of the key technologies powering new payment solutions.

By keeping data in RAM, IMC eliminates the bottleneck that slows down every other high performance computing strategy: painfully slow disk access. Instead, in-memory computing, coupled with parallel processing and data distribution across a computing cluster, enables service providers with heavy analytics and real-time or near real-time applications to process transactions about 1,000 times faster, performing hundreds or even millions of transactions per second.

IMC platforms are relatively easy to deploy. They are inserted between the application and data layers. The data in the underlying RDBMS, NoSQL or Hadoop database also resides in the RAM of the distributed IMC platform cluster, delivering a tremendous performance boost. Extreme scalability is also easy to achieve. Total system RAM can be increased simply by adding new nodes to the cluster. The system automatically rebalances the data across the nodes, adding the processing power and RAM of the new nodes. Today’s IMC platforms also offer the flexibility, interoperability and security payment solution providers need.

Until recently, IMC was considered too expensive because of the high cost of RAM. However, costs have dropped approximately 30 percent per year since the 1960s. Today, memory is still slightly more expensive than disk-based storage, but the increased performance improves ROI significantly. Many payment solution providers and other financial services firms that have implemented an IMC platform have seen a tenfold or more improvement in their ROI.

For example, Sberbank, the largest bank in Russia and the third largest in Europe, faced a problem of speed, scale, availability and security similar to what payment solution providers are facing: transitioning from traditional, human-based investment advising and trading environment to a real-time, 24/7 online access environment. Sberbank recognized the need for a next-generation data-processing platform to handle the expected massive rise in transaction volume and opted for one based on IMC. According to the bank, the IMC platform, which was built using industry-standard hardware, delivered very high performance and reliability while being much less expensive than the technology previously in use.