Burstream – a new entrant to the market data space – has gone live with the first market on its NanoSpeed Mesh, a managed service leveraging FPGA-based technology from NovaSparks (see separate post here). The company is now delivering Nasdaq OMX data to customers co-located at the exchange’s Carteret, NJ data centre, typically in 600 nanoseconds.
As well as very low latency, by providing a turnkey, managed service, Burstream is seeking to make it easy and inexpensive for trading firms to deploy their proprietary trading models, says Paul Barringer, CEO of the company.
At Nasdaq’s co-lo, Burstream has built an infrastructure comprising NovaSparks’ data feed appliance, routers from Arista Networks and servers from Super Micro Computer. Customers can “show up with their algo and turn it on in a day,” says Barringer.
If customers don’t want to run their code on Burstream’s servers, they can cross connect to their own, says Barringer, who notes that even some major firms are choosing to go with the completely managed service in order to avoid the effort, expense and time it takes to provision their own equipment.
The NovaSparks appliance offers very low latency by dint of its total FPGA – Field Programmable Gate Array – design (no use of a traditional CPU). It provides data feed processing and order book creation, with highly deterministic (that is, very low jitter) processing even when inbound message volumes are very high. Burstream pegs this at 600 to 900 nanoseconds.
Next up for Burstream is to bring on line its infrastructure at the 350 East Cermak data centre in Chicago to handle Chicago Mercantile Exchange data. That data will be fed not only to customers that co-locate at that facility, but will also be routed via fibre connectivity to Carteret. In due course, as Burstream adds co-lo/data collection sites, data will be fed to all nodes on its mesh network. It will make use of technology from Vello Systems in that endeavour.