NEW YORK, March 20 -- Wombat Financial Software Inc., the high speed trading technology experts, today demonstrated further leadership in the provision of high-performance market data solutions. An independent study conducted by the Securities Technology Analysis Center (STAC) recently provided evidence that latency spikes common to traditional market data environments can be substantially reduced by implementing the Wombat solution on an infrastructure comprised of Cisco low latency InfiniBand Server Fabric Switches (SFS) and AMD Opteron processor-based servers. Cisco Server Fabric Switches use standards-based InfiniBand technology to provide a high performance unified fabric for connecting servers together into grids of compute resources.

Increasing volumes from securities exchange feeds like OPRA need high-performance, high-capacity infrastructures. Improving performance in automated trading environments is a critical requirement for companies in the financial services industry to stay competitive.

The STAC study consisted of a rigorous benchmarking exercise designed to simulate the throughput demands of a real world data feed. Results demonstrated that latency spikes of over 30 milliseconds -- typically induced by the “initial image” request occurring at the instantiation of new clients -- were eliminated when using a Cisco InfiniBand SFS networking infrastructure. Single client mean latency for update sizes of 214 bytes was reduced from 220 microseconds to 80 microseconds, a 63 percent reduction. Reductions of over 50 percent were demonstrated with payload sizes of 1KB and 6KB. A full description of the testing methodology and results may be found at www.STACresearch.com.

Peter Lankford, director of STAC, said: “This initial comparison of Wombat transport latencies indicates that Cisco’s InfiniBand infrastructure has several benefits over gigabit Ethernet. In addition to reducing the average latency of market data updates, it also reduced the variability of latency. We were especially impressed with the way that the Cisco SFS solution improved the software’s predictability under extreme load. Under conditions that caused latency on the Ethernet-based system to spike upward by tens of milliseconds, the Cisco InfiniBand solution held steady. Predictability is valuable to firms that engage in automated trading. Unexpected increases in latency can throw off an algorithm and hurt profits.”

“We are deeply engaged with industry leaders including Wombat and AMD, in offering compelling low latency solutions for accelerating automated trading platforms, for the financial services industry,” said Bill Erdman, director of product management for the Server Virtualization Business Unit at Cisco. “This recent study is a clear proof point regarding the applicability of InfiniBand, as the premier low latency technology for this market.”

Cisco Server Fabric Switches provide up to 288 ports of non-blocking, InfiniBand 4X that can operate in either 20-Gbps double data rate (DDR)) or 10-Gbps single data rate (SDR) mode for server and inter-switch connectivity. InfiniBand takes advantage of Remote Direct Memory Access (RDMA) technology greatly reducing latency, latency jitter and CPU utilization while increasing the throughput and message capacity of automated trading environments.

“Distributed applications demand the unprecedented performance of AMD Opteron processors,” said Patrick Patla, director of the Server and Workstation Division at AMD. “When combined with leading high-performance solutions like Cisco SFS and Wombat, customers have a tested solution that delivers greater business agility and a competitive advantage.”

The Testing Process

The STAC study rigorously followed established Wombat test procedures on both Cisco SFS infrastructure and a Gigabit Ethernet infrastructure. Wombat unicast middleware was implemented on Cisco’s Remote Direct Memory Access (RDMA) technology without any modification to the Wombat executables. In the tests, a Wombat feed-handler simulator had to respond to large batches of requests from consuming applications while sending out a heavy stream of update traffic. On the Ethernet-based system, these request/response bursts caused large latency spikes, while on the system using the Cisco InfiniBand SFS networking infrastructure, there was no difference from the steady state. Overall standard deviation of latency was approximately one third lower. STAC employed their benchmarking best practices for the testing, including recurring runs of the tests to ensure repeatability of the results. A full description of the test methodology and results can be found at www.STACresearch.com.

“Messaging platforms are only as valuable as the performance they can extract from the multiple levels of technology they depend on,” said Danny Moore, COO of Wombat. “We believe InfiniBand enables a significant increase in performance and have invested our resources to make this technology accessible to the capital markets sector. Collaborating with Cisco and AMD has allowed us to make the Wombat Platform the one solution that allows customers to realize the benefits of InfiniBand in their market data systems today.”

About Wombat Financial Software

Wombat enables limitless connectivity to the global markets through the world's most advanced high speed market data infrastructure, the Wombat Platform. With offices in the US, UK and Japan, Wombat software underpins automated market-making systems, electronic-trading floors and tick-capture platforms in financial institutions around the world.

The Wombat Platform includes Wombat Feed Handlers, providing sub-millisecond connectivity to more than 80 direct exchange and aggregated vendor feeds; Wombat Entitlements, providing market data compliance, analysis and reporting services; Wombat Transport, enabling high speed, multi-protocol messaging throughout the enterprise; and Wombat value added servers such as the Wombat SuperBook virtual order book consolidator and the Wombat DBAC tick-capture adapter. For more information, visit www.wombatfs.com.

-----

Source: Wombat Financial Software Inc.