BNY Mellon and Hazelcast: Gaining More Business Value by Expanding Use from Caching to Instantly Processing Real-Time Events

I love sharing our customer stories and celebrating the successes of our innovative customers and users. Today, I’m spotlighting BNY Mellon, a global authority at the center of finance. It has nearly $50 trillion in assets for its clients, which include more than 90% of the companies in the Fortune 100. The global financial services company is known for investing in what’s next and exploring every opportunity to give its clients an edge.

A few months ago, I had the opportunity to meet Michael Goldverg, Managing Director & Distinguished Engineer at BNY Mellon at the Current Kafka Summit in San Jose, where he presented. Michael is an innovative technology leader who’s been expanding the use of the community edition of Hazelcast Platform for more than 8 years to help BNY Mellon do more—faster. 

He shared his story and views about the rapidly evolving stream processing platform market, as well as how he and his team have evolved their use of our high-performance, resilient, and scalable unified real-time data platform. BNY Mellon uses the Hazelcast Platform for ultra-low latency data storage and compute to optimize business operations in real time and enable ultra-fast intelligent action in the era of AI.

Enjoy the video interview and keep reading for more highlights from the conversation:

BNY Mellon Requires Fast Data Access to Speed Up Application Response Times 

Access to the freshest, most up-to-date information about entitlements or permissions about who has access to what, when, and where is a mission-critical and enterprise-wide process for banks and financial services companies. It’s a highly regulated function (SOX, GLBA, and Basel II) and a value-added capability that employees use to gain business insights, boost customer experience, improve operational efficiency, mitigate risk, and reduce fraud. 

Hundreds of systems are integrated with the entitlements system, so it’s essential to enable fast access to the latest entitlement information using a high-performant, resilient, and scalable data store. Unfortunately, many entitlement systems rely on slow databases that don’t meet these requirements. 

At BNY Mellon, there was increasing demand on the entitlement system, and as a result, the system’s performance was slow. Like many of our customers, to speed up application response times, the bank started using the open source version of Hazelcast Platform as a real-time fast data store – or caching layer – by pulling entitlement data from two systems, one of which is an Oracle database, to continuously refresh it in real-time and ensure the most up-to-date information is available at all times. According to Michael, Hazelcast’s fast data storage capabilities reduced response times from “multiple seconds to a few milliseconds.” 

“Within a couple of weeks, we created a data abstraction layer with the Hazelcast Platform that provides extremely efficient data processing. It underpins business-critical applications, enabling instant action even during peak demand,” Goldverg says. “It was a very straightforward answer to a fairly complex problem.”

Instead of building, integrating, maintaining, and operating multiple technologies for our real-time architecture, we simplified our application development and modernization by using a unified real-time data platform.

About Hazelcast Platform
Hazelcast Platform is a unified real-time data platform that simplifies application development and modernization to accelerate time-to-market while reducing maintenance overhead and lowering TCO. The software platform uniquely combines a distributed compute engine and a fast data store into a single runtime to process both streaming data and stored data. 

It enables three approaches for application modernization: 

1. real-time data storage and compute with continuous data updates and enrichment, 
2. real-time action through an event-driven architecture, and 
3. real-time intelligent action with AI/ML-based automation including fast and easy-to-deploy ML inference capabilities. 

Hazelcast Platform supports the major cloud providers, namely AWS, Azure, and GCP.

“Instead of building, integrating, maintaining, and operating multiple technologies for our real-time architecture, we simplified our application development and modernization by using a unified real-time data platform,” says Goldverg. “It uniquely combines a distributed compute engine and a fast data store into a single runtime to process both streaming data and stored data. It minimizes operational disruptions and fast recovery with resilience and provides elastic scalability and faster time-to-market with a low TCO.”   

BNY Mellon Experienced the Value of Operating on Data In Real-Time
Over time, BNY Mellon’s usage grew beyond fast data storage to include high-speed, data-colocated compute and event processing, Goldverg says. “It was quite exciting to see what else we could do with the product. Over time, we were able to do a lot more.” 

The prime example of this is BNY Mellon’s use of the Hazelcast Platform is for stream processing and leveraging real-time events. BNY increasingly saw the need to operate quickly on data flowing in from real-time events. “Operating on data in real-time opened up a lot of opportunities,” said Goldverg, including fraud prevention use cases. 

Effective use of stream processing is a common challenge, and other companies will face many of the same difficulties overcome by BNY Mellon. 

For instance, to take advantage of stream processing capabilities, what’s needed is for data to come in continuously—not in batches.

“As we saw how data arrived, we saw more opportunities to perform processing as soon as data came in,” Goldverg says. “The seeds are there” for BNY to offer that business capability to clients now.

“Don’t think of Hazelcast Platform as just a cache for storing and reading data. It’s a high performance distributed storage and compute platform for ultra fast data processing. You can use it to reduce processing time from minutes to milliseconds. Instead of retrieving the data and then computing it, you can push the processing where the data resides. We are further optimizing data processing by analyzing streaming data, from multiple sources including Kafka, before it is stored instead of processing in batches.”  

What BNY Mellon has accomplished with stream processing is just the beginning.

Exploring Flink vs. Hazelcast for Stream Processing 

BNY Mellon is constantly reviewing new technologies. But “Hazelcast Platform is already doing what other new products are just now starting to do,” Goldverg says.

Flink, for instance, “is less mature and complex to ramp up and use,” as are other newer products. “The Hazelcast Platform is a solid streaming platform. It’s much more mature, stable, proven and highly performant for complex data operations for use in business-critical applications where you can’t afford to have downtime. It’s ready to use with enterprise-grade features, resiliency, scalability and high availability. And, it’s easier for developers to ramp up and use.”  

“What BNY Mellon has accomplished with stream processing is just the beginning,” says Goldverg.

Advice for Getting Started with Stream Processing
Goldverg’s tips for others when using the Hazelcast Platform, especially around stream processing:

• Start small. With success, you’ll become proficient with the Hazelcast Platform and learn to solve problems. The open source version of Hazelcast Platform is great for test, development, and experimentation.
• Test to validate. Don’t rely on manual testing. Write tests that define what you want the system to do so you can rapidly test.
  
• Define performance goals. The Hazelcast Platform is all about performance. Define goals and gradually iterate to success.
  
• At Hazelcast We Bring Passion for Customers
  

At Hazelcast, among our five values is the “customers for life” philosophy we embrace daily. We are committed to building strong relationships with our customers and users to ensure their application modernization initiatives are successful. From our users to customers – including our esteemed product strategy board (PSB) – we value all feedback to drive innovation and improvement.

We’re thrilled to have leaders like Michael as a customer, user, and member of our PSB. Innovators like him are simplifying building, running, and maintaining high-performance applications to accelerate their company’s time-to-market, at a lower TCO. 

For more on what makes Hazelcast so impactful, hear directly from Michael about BNY Mellon’s use of the Hazelcast Platform enabled it to shift data processing speeds from minutes to milliseconds.