Infosys helps banks leverage new computing technologies to lower cost and improve business agility

Banks, facing fierce global competition, are updating IT infrastructure to release new products faster and decrease operating costs. They are looking for greater computing scalability and flexibility to respond to continually changing business dynamics. Today, banks must cope with profit pressure, regulatory changes and new competitive entrants as they compete in all corners of the globe. "Banks recognize the need to use IT as a key tool to gain a competitive edge," says Mike Stephens, General Manager of Asia Pacific Financial Services Industry, Hewlett Packard.

Their biggest challenge is to improve customer profitability by marketing new services that retain and scale up customers, while attracting new ones. This is aided by offering differentiated products that increase customer loyalty. At the same time, banks must comply with more international regulations, such as Sarbanes-Oxley, that impose more rigorous financial reporting and capital management requirements. Additionally, banks must manage costs by increasing operational efficiency. Cost can be reduced by using systems that present a single enterprise level view of the customer as opposed to having data scattered across disparate accounts-based legacy systems.

To achieve greater success, banks are developing IT strategies that ease business workflow and new product introductions, helping them keep pace with the rapidly evolving business environment. In some cases, banks are eliminating their "silo systems" and turning to end-to-end banking solutions running on high performance computing systems. Financial institutions are also evaluating Service-Oriented Architecture (SOA), a software integration framework that simplifies deployment and reuse of software modules.

Infosys Technologies addresses these IT strategies with its Core Banking Solution, called Finacle. Infosys further enhances its financial software suite by running it on open system-based hardware that decreases total cost of ownership and incorporates software optimizations to maximize performance. This scalable and flexible solution provides banks with the infrastructure needed to adapt quickly to change, increase profitability and expand product offerings.

Challenge
• Identify and eliminate bottlenecks. Increase system performance by addressing issues that block the system's ability to deliver maximum performance.
• Optimize Resource Utilization. Maximize the throughput of key hardware and software system components.

Solution
• Perform benchmark tests. Run "Interactive Response Time Law" tests that identify performance bottlenecks.
• Tune the system stack. Modify OS parameters and database queries based on application specific banking transactions.

Migrating to Service-Oriented Architecture (SOA)
Most large banks throughout the world have heterogeneous computing environments comprising various applications running on a range of hardware platforms. But with a changing business landscape, along with a steady stream of mergers and acquisitions, banks are reluctant to upgrade the legacy systems that run their businesses.

However, a growing number of banks are considering replacing their existing core systems with next-generation vendor solutions, increasing business agility and lowering Total Cost of Ownership (TCO). SOA-based solutions, like Finacle, help banks achieve economies of scale by simplifying the reuse of software and IT technology. SOA is technology agnostic, not aligned to any specific hardware platform, development language or technology tool, which means it can be deployed into existing IT environments.

SOA-based applications are composed of logical units of work called "business services". These services are used to develop enterprise applications, like small business lending, which can be created in a plug-and-play fashion. Banking applications typically execute a large number of "business services" per banking transaction, which may be processed in parallel to boost performance. When processing transactions in parallel, it is essential to avoid execution bottlenecks that reduce performance.

Removing Bottlenecks
HP pulled together a virtual team to deliver Transaction processing requires the resources of many system stack components, such as hardware, network, storage, operating system and application software. Whenever two or more transactions need the same resource, some delay can result. The key is to identify potential bottlenecks and minimize them by adding more resources or adjusting how the resources are consumed.

Infosys identified bottlenecks using benchmarking tests that measured how transactions are processed by the system. They used performance laws, such as the "Interactive Response Time Law", to model the theoretical time required to execute transactions, and then compared these values to measured values. Response time models take into account system attributes like transaction volume, the number of users accessing the systems, the number of servers, and the resources per system.

With a thorough knowledge of the system throughput and resource utilization, Infosys pinpointed bottlenecks that resulted in unproductive dead time. This information was critical to determine which system components, both hardware and software, required further optimization.

Tuning the System
Infosys determined that system performance could be increased by adjusting operating system and software driver parameters. They controlled operating system behavior by modifying a) the maximum memory allowed per process, and b) the maximum number of open files, using the "maximum number of file descriptors" directive. These values are closely tied to the number of TCP (Transmission Control Protocol) connections a software process opens or the number of files the process accesses during the execution.

Comprehending the transactional dynamics of the system, Infosys tuned the amount of memory assigned to each transaction. This was a tradeoff between the number of concurrent transactions and the amount of system resources allocated per transaction. If the memory allocation per transaction was too small, transactions with a high volume of data would incur long delays. Conversely, large memory allocation per process reduced the number of transactions that could be executed concurrently.

In a transaction based system, such as a core banking solution, the database is a highly utilized resource and its performance can directly impact the throughput of the system. Infosys optimized database performance by using known best practices such as connection pooling. This reduces the amount of time users must wait to establish a connection to a database. After a connection is used, it is placed in a temporary storage location (cache) and can be used over again, avoiding the delay from establishing a new connection.

In addition, Infosys used database statistics to profile the application's database queries. They used this information to create indexes that provided the best performance for resource intensive table lookups.

Summary and Metrics
The combination of the Finacle core banking solution and servers built on the foundation of Intel Itanium processors achieved ground breaking performance in benchmarking tests. For online transaction processing, Finacle achieved a throughput of 11,180 transactions per second with an average response time of 0.29 seconds. Another benchmark achievement was performing 19,568 transactions per second for batch mode. These results out-performed industry standards by a wide margin. The benchmark tests have been reviewed by Ernst & Young.

Benefits of Using Itanium® 2-based Solutions
The transactions-based benchmark results demonstrates Finacle's capability as a highly scalable core banking solution. This open systems-based solution allows banks to choose a hardware platform that lowers their total cost of ownership, while end-to-end functionality increases their agility. With services oriented architecture and high levels of performance on HP superdome and Integrity servers powered by Intel Itanium 2 processors, Finacle is fast becoming the solution of choice for many banks worldwide.

Key Platform Components
• HP Superdome* and Integrity* servers with Intel® Itanium® 2 processors.
• Oracle9i Database providing efficient, reliable, and secure data management for high-end applications such as high-volume on-line transaction processing (OLTP).

Integral Answers
• Infosys used "Interactive Response Time Law" tests to locate performance bottlenecks.
• By tuning OS parameters such as the number of file descriptors and maximum memory size per process, system performance was increased.
• Database optimizations, such as connection pooling and query

© 2007 Itanium Solutions Alliance. *All other names and brands may be claimed as the property of others.

Note: Information and claims herein are provided by the award recipient and in now way warranted or endorsed by the Itanium Solutions Alliance. The Itanium Solutions Alliance does not control, verify or audit such information or claims and encourages all customers to independently obtain more information about the products.