Global Banking and Financial Services Firm Drives Down Cost, Risk, and Complexity with Azul Compute Appliance

When your clients include some of the world's most well-known companies, wealthy individuals, and institutional investors, your mission-critical trading applications had better perform flawlessly. For this global banking and financial services firm, whose trading values were doubling every year, ensuring high application performance was beginning to be a problem. The bank, active in over 50 countries and employing approximately 40,000 people, could no longer ensure its trading applications would deliver their intended business outcomes. This posed significant risk to the firm. Trading applications provide a consistent revenue stream and are the lifeblood of the business. If these and its other Java™ applications suffer from poor scalability or response times, the entire bank suffers from loss of revenue, loss of margin, and added cost due to compliance reporting penalties.

Key applications unable to meet business needs

For example, its Fixed Income program, which takes trades from customers and routes them to the market, could only support 1000 trades per second. With user demand growing rapidly, the bank could start losing customers if this application was unable to scale with the business. The bank’s application for buying and selling trading companies’ debt also had performance problems. It could only support eight simultaneous users with 24-second response times on average. This was unacceptable. A key business requirement of this application is the ability to handle split second buy/sell decisions. Clearly, the bank needed to optimize its trading applications or risk the business consequences. And it had to act fast. The banking industry is highly competitive and financial institutions are always looking for ways to steal customers away from a competitor. To solve this challenge, the company had to find a way to make its Java environments run faster while at the same time making it easier for the company to manage. It considered the traditional horizontal/scale-out or vertical/scale-up scaling models, but these approaches would be costly and add huge amounts of IT complexity.

The company needed a new technology that would deliver extraordinary performance, break free from application memory constraints, and fit into the bank’s IT environment with minimal disruption.

Azul Compute Appliances deliver breakthrough performance

For help, the bank turned to Azul Systems. After performing due diligence, the bank opted for a solution based on the Azul Compute Appliance, a server technology that provides massively multicore processing and power density. Azul Compute Appliances provide unprecedented capacity and scalability for the bank’s Java-based workloads while offering revolutionary power, density and cooling economics.

Azul Compute Appliances are pooled together to deliver capacity as a shared network service, which multiple applications can tap into at the same time – an approach known as network attached processing. This approach is similar to network attached storage (NAS). It provides all the benefits of scale-out with a lot fewer servers to manage. And like NAS, it does not require disruptive changes to applications and eliminates speculative capacity planning.

Azul Compute Appliances free applications from memory constraints by providing massive heaps (up to 200 GB) per application instance, which don’t pause due to the company’s exclusive hardware-assisted Pauseless Garbage Collection. This special hardware assisted algorithm offers parallel concurrent garbage collection and virtually eliminates application pause times, clearing the path for more predictable and consistent SLAs.

The Azul Compute Appliance also introduces Optimistic Thread Concurrency (OTC), a capability that eliminates the unnecessary contention for shared objects in memory that typically has a dramatic impact on parallel performance and can cause long delays in large applications. OTC lets threads speculate through lock requests, while the hardware monitors and protects against possible memory contentions.

Trading applications meeting and exceeding SLAs

As a result of the differentiated Azul approach, the bank has dramatically improved the response times and scalability of its mission-critical applications. Today, the bank can ensure that its trading applications will continue to deliver predictable results in the face of growing demand and that its customer SLAs will remain intact.

In addition, the bank has enough spare computing capacity to handle unexpected requirements. And long pause times, poor throughput, and inconsistent responses are things of the past. For example, the bank’s Fixed Income trading application now supports 6000 trades per second, with flat, consistent response times. Before, it maxed out at 1000 trades per second. The bank’s application for buying and selling trading companies’ debt also showed dramatic performance improvements. Now, running on Azul Compute Appliances (and with virtually no changes to the application), this trading application scales to 88 simultaneous users – with just 2-second response times. Previously, it could only handle eight users with 24-second response times.

At the same time, the improved response times for key internalfacing applications allows the bank to submit regulatory reporting in a more timely fashion and avoid severe financial penalties from the SEC. Those fines could have of exceeded millions of dollars and put the company in jeopardy on Wall Street.

Moving forward, the Azul technology will help the bank as it begins building its next-generation data center. The bank now has a technology strategy to support a centralized, shared, or utility computing model. With Azul, the bank can guarantee resources, while reducing the number of servers and operational complexity across all data centers globally. This will pave the way for the global banking and financial services firm to achieve its most important objective: to become the world’s premier bank.