Avatar

Today, April 8, 2013, the brightest of Wall Street are meeting at the 10th Annual High Performance Computing Linux for Wall Street event in New York to learn about the latest technology innovations that give financial trading firms a competitive edge. With network latency achieving speeds of less than 50 nanoseconds, this year’s event focuses not only on speed but also on cutting-edge advances in network visibility, analytics and instrumentation. One way Cisco is helping traders gain a competitive edge in this area is by offloading the gathering of analytics data to the network devices, and working with industry leading monitoring firms like Corvil and TS-Associates to decode these new sources of data.

No doubt, high performance trading systems are becoming faster, processing more data than ever before. Network performance is of paramount importance in high performance trading markets where obtaining and acting on real-time equities, options and other financial market information, nanoseconds faster than competitors, may be worth millions of dollars. While complex trading analysis and risk mitigation still need to be performed outside the network layer, significant cost savings and scale advantages can be obtained by leveraging embedded intelligence in the network to enable key functions such as fabric congestion monitoring and intelligent traffic mirroring.

With the recent introduction of the Cisco Nexus® 3548, Cisco provides customers with highly granular visibility into congestion in the trading fabric using a feature called active buffer monitoring. This feature relies on purpose-built silicon to collect packet buffer utilization data at frequencies as low as 10ns, presenting it to the user in detailed histograms. The feedback on active buffer monitoring has been very positive with customers and monitoring vendors requesting that the collected data be actively pushed to monitoring systems at a high frequency interval to enable more scalable data gathering. Further investigation revealed that there are many different intervals and export formats that would provide value to customers looking to consume this data. Instead of using an approach that would mandate a rigid export protocol, Cisco chose to enable this functionality through our python scripting capability, using the sockets library to send custom-built packets to analyzers at a configurable interval. The work we are doing in this area can be found at our github page, here.

Additionally, the Cisco Nexus® 3548’s new IEEE-1588 precision time protocol (PTP) feature uses purpose-built silicon to propagate highly synchronous time data through the network. This allows the 3548 to offer intelligent traffic mirroring using the embedded remote switch port analyzer version 3 (ERSPANv3) feature, carrying mirrored network traffic across multiple layer-3 hops to an analyzer as well as value-added metadata. Nanosecond granularity timestamps, source interface, and traffic direction are examples of metadata that facilitates differentiation between identical packets.

By correlating the data provided by these features, traders gain greater visibility into why gapping, slippage and slow order situations occur, enabling smarter trading decisions. Most importantly, by integrating this functionality into the network devices themselves, this flexibility and intelligence can be delivered while managing instrumentation device costs.

We’re excited that these innovations are being adopted by leading ecosystem partners such as Corvil and TS-Associates to provide direct correlation between congestion identified by active buffer monitoring and trading data captured by ERSPANv3 and other sources such as network TAPs. This integration allows users to leverage one dashboard to view all relevant data to determine the health of a trading environment, especially in areas like capacity planning and troubleshooting.

A global connected marketplace has elevated the demands placed on trading networks, making visibility and control as critical to deterministic performance as raw speed. Proactive adaptation to market volatility with real-time monitoring and precision timing accuracy of trade events provides trading participants greater control and visibility at peak opportunity times across the globe. By providing these capabilities, the Cisco Nexus 3548 with Algo Boost enhances the Cisco High Performance Trading Fabric architecture to help enable greater business agility.



Authors

Dave Malik

Cisco Fellow and Chief Architect

Customer Experience