TD Securities Case Study
TD Securities Achieves 100% SLA Compliance Using the Azul JVM
With more than 2,700 people in 15 offices around the world, TD Securities is one of Canada's leading providers of advisory and capital market services to corporations, governments and institutions, and the premier liquidity provider. TD Securities is a subsidiary of TD Bank Financial Group, which offers a full range of financial products and services to approximately 17 million customers worldwide and ranks among the world's leading on-line financial services firms, with more than 5.5 million on-line customers.
TD Securities employs a wide range of business-critical Java™ applications worldwide for its algorithmic trading and credit and risk analysis operations, as well as various types of ad-hoc queries and data mining. As a result, the company's centralized datacenter must be available on a 24/7 basis to support global operations.
In this demanding environment, application performance and SLA compliance are essential. Any delay in analysis and decision-making can potentially mean the loss of millions of dollars. The challenge for the company was ever-increasing transaction volumes that threatened its ability to meet performance and service level goals. In addition, the company was growing more concerned with the amount of physical hardware and resources needed to support operations.
Performance Demands and Higher Volumes Present Difficult Challenge
TD Securities relies on both internally developed and commercial Java applications to run its credit derivatives and fixed income trading platform—the foundation for the company's reporting and analysis capabilities. Unfortunately, the increase in demand from users within the bank for additional reporting, more complex ad hoc queries, and requests for faster reporting and analysis turnaround became a major burden. In addition, trading volume had increased significantly, causing the application to hit Java scalability limits and requiring additional computer and human resources just to maintain system performance.
At the root of the problem were Java deployment platform issues, namely garbage collection pauses and memory limitations. If ignored, these would continue to degrade performance and cause the company to miss its SLAs. This shortfall could be particularly severe for TD Securities because it impedes overnight application processing, thereby delaying financial analysis and decision making by several hours, or even an entire day.
In addition to application performance issues, the TD Securities team was concerned over the growing amount of physical hardware needed for their IT operations, as well as the additional personnel required to manage the growing pool of compute resources.
Through their Java application partners, TD Securities learned of Azul Systems and the benefits their products have provided to other major financial services companies. After a thorough evaluation, TD Securities moved forward to deploy a solution based on the Azul JVM.
Rapid Validation of Azul's Capabilities
TD securities' first concern was how Azul would perform with existing applications, given that these consist of a mix of single-threaded and multi-threaded code. In the first test, this issue was resolved. The Azul JVM and Compute Appliances delivered excellent performance right out-of-the-box, and offered the additional advantage of no garbage collection pauses and no limits on memory resources.
Very quickly, it became clear that the Azul Compute Appliance—which included up to 864 processors and 768GB of memory—is an optimized platform for multithreaded applications. As TD Securities continued to work with Azul Systems products, the team was impressed with the simplicity and speed of deployment as well as the quality and effectiveness of the software tools provided with the Azul solution.
An example is Azul's RTPM (REALTime Performance Monitoring) tool, which provides real-time production monitoring and diagnostics of all application activity processed through the Azul Java virtual machine (JVM). Within the first week of using RTPM, TD Securities was able to find and fix difficult-to-reproduce—and therefore debug—problems within their database tier. Unlike traditional diagnostic tools, RTPM gave the bank real-time production visibility into the application's operations, even down to the thread/method level. In addition, the bank could use RTPM to optimize multi-threaded performance by quickly finding concurrency bottlenecks using real production data.
The next implementation stage was focused on performance and reliability in achieving and exceeding SLA thresholds. Not only were all of TD Securities' performance and reliability goals met, but they also found that their existing infrastructure could support several times more transaction load than before, with no major application changes.
The Promise of Java Redelivered
With the deployment of the Azul solution, TD Securities now has embraced a revitalized view of Java and its role within the datacenter. Azul Systems has helped to deliver all the advantages of Java the bank had originally targeted–simplicity, portability, and efficient internal development. Further, the issues that prevented the bank from realizing these values—long garbage collection pauses, increasing numbers of servers needed to support operations, difficulty in debugging internal applications, limits on speed and the depth of ad-hoc queries—all have been resolved by incorporating the Azul JVM and Compute Appliances into the existing infrastructure.
In terms of tangible results, TD Securities is now able to meet their demanding SLAs 100% of the time using the Azul JVM and Compute Appliance with improved overall performance.
In addition, the bank has gained substantial datacenter efficiency, both in terms of physical and human resources, for lower Total Cost of Ownership. For example, twenty data blades can be replaced by a single blade connected to the Compute Appliance. This reduction in blades has resulted in significant savings in power, cooling, and floor space requirements.
Overall, the bank can now consistently achieve tight SLAs, perform faster and deeper analysis and queries, provide their internal developers a better Java development platform, and hold down the human and operational expenses of managing the datacenter.
By turning to Azul, TD Securities not only achieved its original goals, but now is able to focus on improving application capabilities to support business needs, not working around Java limitations. The Azul Compute Appliance model keeps the performance and capabilities of Java "ahead of the curve" for TD Securities.
Ever-increasing transaction volumes that threatened its ability to meet performance and service level goals
The Azul JVM delivers all the advantages of Java the bank had originally targeted–simplicity, portability, and efficient internal development
100% SLA compliance
Improved datacenter efficiency
Faster queries and deeper data analysis