Juniper Networks is putting more compute capabilities into its networking hardware to help financial services institutions accelerate their business-critical applications while reducing the latency in their environments.
The networking vendor in the third quarter will launch the QFX5100-AA switch and QFX-PFA accelerator module, which will enable users to distribute more compute capabilities across the network infrastructure and leverage Intel’s x86 Xeon server processors, Broadcom’s switching silicon and Altera’s field-programmable gate arrays (FPGAs) to significantly improve the performance of applications used for data and social media feeds as well as execution routing and big data analytics.
Financial services institutions will be able to more quickly make real-time decisions in a space that features rapidly changing business opportunities.
“High-frequency trading is declining and may soon go the way of the Merchants of Venice,” Andrew Bach, chief architect for Juniper’s financial services team, wrote in a Feb. 24 post on the company blog, foreshadowing the QFX5100-AA and QFX-PFA hardware. “The new advantage comes from the ability to analyze breaking news and social media sentiment in real time and make trades based on a broad base of instantaneous information. Miss the increasingly narrow window—and miss the market.”
Juniper officials call the trend “hyper-contextual trading,” which means organizations must be able to quickly analyze market data, news and social media before making decisions, and the amount of data that must be analyzed continues to grow—as many as a billion messages a second, Bach wrote. The demands call for more distributed computing that focuses less on zero latency.
“In hyper-contextual trading, ultra-low latency is presumed—not a differentiator,” he said, noting that the industry is reaching the physical limits in how fast packets can be switched.
An emerging approach called applications-embedded networking calls for embedding application code onto the switch fabric, which reduces network latency for applications and the costs by cutting down on the number of servers and switches that are used.
“There are several ways to embed applications into the network fabric, including integrating a virtual machine in the switch, inserting an FPGA in the data path and switch, and incorporating an ASIC-packet forwarding engine,” Back wrote. “At Juniper, we think that the FPGA approach is ideal for financial services, because firms can embed their own application code directly into the switch.”
The QFX5100-AA and QFX-PFA—which expand Juniper’s QFX family of products—will enable financial services providers and their customers to make quick decisions based on an analysis of huge amounts of data, according to Juniper officials. The QFX5100-AA is a 40 Gigabit Ethernet switch that uses Xeon E3-1125C v2 processors with Broadcom switching silicon, while the QFX-PFA is based on Altera’s multi-100G FPGA. Both work in combination with Maxeler Technologies’ data flow programming technology.
The products also run Juniper’s networking operating system, Junos, and integrate with the vendor’s Junos Space Network Director for automated management from a single point. They support Java as the programming language. Juniper’s Cloud Analytics Engine gives users greater visibility and an application-centric view of the network through such capabilities as high-frequency network monitoring and data analysis.
Juniper officials, including Bach, are in New York April 6 demonstrating the company’s cloud networking solutions for the financial services industry.