In the news yesterday as a result of its global partnership with Reuters, Endace’s technology – coupled with that of partner Trading Metrics – underpins the Reuters Latency Monitor. But just who is Endace, and what does their technology really do? IntelligentTradingTechnology.com talked to Endace vice president of marketing Steve Gleave to find out more …
IntelligentTradingTechnology.com: Let’s start with a quick intro to Endace.
Steve Gleave: We’re quite an interesting little company really. We were founded some ten years ago out of a research project being performed at the University of Waikato, in New Zealand. It was evident at the time that no solutions existed that could reliably capture every single packet from a high speed network link, time stamp them accurately and write them directly into CPU memory. We needed to build these network cards for the project, and inadvertently, a company was born. Subsequently, these Data Acquisition and Generation or DAG cards have been sold into numerous platforms and applications worldwide. Financial service customers have always been attracted to this technology given our highly accurate, hardware time-stamping and the ability to use that for accurate latency measurement. We’ve also moved up the value chain by integrating these cards into servers and starting to ship managed appliances. During this time, we’ve listed on the London AIM and grown to over 100 employees, with offices and partners worldwide.
ITT: The Reuters Latency Monitor offering involves Reuters, Trading Metrics and yourselves. What does each company bring to the party?
Steve: Obviously Reuters brings market reach and instant brand recognition. One of the many services that Reuters delivers are consolidated market data feeds which collect trade and quote feeds from multiple market exchanges and ECNs and serve them up to customers worldwide. These feeds are then distributed to trading applications at the customer premises using something called RMDS, or Reuters Market Data System. There are thousands of RMDS customers worldwide and RLM, or Reuters Latency Monitor, can be used by any of them to measure latency in their market data system.
Trading Metrics is a software company out of New York City. They offer a suite of products for the real time measurement of trades, market data, and exchange latency. In this case, they are licensing their Market Latency Metrics software to Reuters to use as part of RLM. It’s really cool software – providing a sharp GUI for monitoring market feed message delay in real time and then being able to compare the data between different exchanges or at different times on different days.
Endace supplies the hardware. We load the Trading Metrics software onto one of our platforms and ship this solution as a latency measurement appliance to Reuters’ customers, and it’s called Ninjabox-LM. So behind the Trading Metrics front end, you’ve got our solution monitoring packets, extracting and applying time-stamps, staying synchronized – giving all that data to the Trading Metrics software so that they can turn the data into information – which in turn allows Reuters customers to fine tune their algo trading solutions.
ITT: Can you describe your Ninjabox-LM in more detail. What it is, and what does it do?
Steve: Ninjabox LM is a combination of four port GigE DAG cards, quad-core CPUs, lots of RAM and some SATA hard drives. It comes with Trading Metrics software pre-loaded. Ninjabox is really a high performance network monitoring device, operating at gigabit speeds, with copper or fiber interfaces. Our DAG cards time-stamp every single packet, in hardware, using CDMA or GPS clock sources, add a header into an extended record format, and write this directly into memory without interrupting the CPU. All of our DAG cards support a common API through which other applications, namely Trading Metrics, can access the stored data and use accordingly. In this case, the software uses the time-stamping to calculate delays from exchange to premise, or at any point within the customer’s RMDS installation. Graphs are drawn, reports are generated and alerts are raised depending on the parameters that you’ve set.
ITT: What does a typical Ninjabox-LM installation consist of?
Steve: It’s pretty straightforward. Ninjabox can connect to the SPAN port of an enterprise switch or router. Or if you’re short of SPAN ports, we can supply you with a tap that’ll unobtrusively and passively monitor Ethernet links without adding any delay. These connections are made at different points within the RMDS network; typical locations include data feed termination, market data hub, point to point servers or consuming applications. The Trading Metrics software observes the time-stamp at each location and then analyzes the performance and measures latency. This allows Ninjabox-LM to measure the latency from the exchange to the customer, or at any transit point within the customer’s network.
ITT: How accurately can Ninjabox-LM measure latency?
Steve: Very accurately. Our time stamps are eight bytes long and offer granularity down to 10 nanoseconds. We believe that accurate measurement is the key right now. Too often, customers upgrade their entire network in the hope of shaving off a few more milliseconds in the quest for ultra low latency solutions. But the first step has to be accurate measurement. Without knowing where your delays are, what’s the point in wholesale upgrades?
ITT: How is Trading Metrics’ analysis software integrated with Ninjabox-LM?
Steve: Tightly. It comes pre-installed. Trading Metrics have been working with our DAG API for a while now, and we’ve been supplying cards for years. But this is the first time that we’ve shipped a platform with latency measurement software pre-installed. Software upgrades can be downloaded from Reuters and we’ll obviously be shipping the latest release whenever it’s been through our internal QA.
ITT: Within an RMDS installation, where might latency occur, and what can Reuters Latency Monitor monitor? What kind of instrumentation is available?
Steve: Latency can occur pretty much anywhere – and of course at different times of day. There’s always more latency at market opening and closing. And different exchanges exhibit different delay characteristics. Of course, RMDS installations are traditionally low latency solutions anyway. But bottlenecks can occur at data feed entry to the customer, across the market distribution hub, and within the servers that provide access to the data that’s on the hub. We measure across any and all of these transit points.
Consuming apps can get slow as well. RLM measures the delay pretty much anywhere from exchange to consuming app – but it’s for market data feeds, not for trade execution. The Trading Metrics software facilitates all kinds of graphical representation for charting exchange and RMDS transit point delays. These can be absolute numbers, or shown as distributions in and out of user-defined percentage bands.
ITT: How will Reuters, Trading Metrics and yourselves be working together to market, sell and support Reuters Latency Monitor?
Steve: Clearly, Reuters will be pushing RLM as part of their many service offerings. Obviously there’s a pressing need for latency measurement given the amount of algo trading that’s going on and the need for “fresh data” to support best execution compliance mandates. Reuters will direct their customers to Endace for order fulfillment. Endace takes orders from the customers directly, ships the NinjaBox-LM and then has the end user work with Reuters to license the software. Endace will then support the hardware, and Reuters the software.
Endace and Trading Metrics will further complement the relationship by co-marketing the solution to those clients that we typically reach directly.
ITT: What if a client does not use RMDS. Is Ninjabox-LM still relevant?
Steve: No. NinjaBox-LM is a solution that works exclusively with Reuters LM for RMDS. That said, the Trading Metrics Market Latency Metrics software can also be used for other feeds and Endace has the ability to power that application. But NinjaBox-LM is a solution specifically tailored for RMDS customers.
ITT: What else is Endace doing in the financial markets world?
Steve: Lots that we’d love to tell you about but probably can’t! Our technology has been used to benchmark a lot of networks; both directly and with reference architectures. The guys at Stac talked about the latter during their Council meeting in December 2007. And we have a number of other partners who are also integrating our technology into their own latency measurement solutions.
But I think this is just scratching the surface for Endace. Our mission is still to deliver horizontally integrated probes to a broad customer base; probes that capture all data and then provide it to a multitude of performance measurement, network security or traffic monitoring applications. We want our customers to know everything that’s happening on their network – which is why we’re still selling the “power to see all.”
For more information, see www.endace.com/rlm.
[tags]endace,reuters,trading metrics,rlm,reuters latency monitor,ninjabox-lm,latency measurement,network management,network timestamping[/tags]
Subscribe to our newsletter
