Based on several component products, NYSE Technologies has introduced its Enterprise Ticker Plant, designed to deliver market data to a wide variety of applications across a trading firm, from low-latency algorithmic trading systems to desktops running Microsoft Excel spreadsheets.
Says Brian Doherty, global product manager for Data Fabric, the offering is essentially a productisation of what has become a “standard deployment model” for the components, including its Data Fabric, data feed handlers, messaging APIs, legacy platform bridge, data entitlements (DART) and systems management (from partner ITRS)
Key to its enterprise scalability is the 6.0 release of Data Fabric, which supports MultiVerb – allowing high fanout Remote Direct Memory Access (RDMA) communications. Applications that require low-latency access to data connect directly using RDMA, while others that are less latency sensitive would typically interface via a bridging daemon, and be fed via TCP or a third-party messaging bus.
Across all receivers, NYSE Technologies’ Middleware Agnostic Messaging API (MAMA) and the Middleware Agnostic Market Data API (MAMDA) provides a common and stable messaging interface, across different operating systems and programming languages.
Also notable is the implementation of kernel bypass within its data feed handler components – transferring data such as UDP packets from the network direct to application memory – using InfiniBand verbs and leveraging Mellanox Technologies’ network interface cards. This approach cuts latency compared to providing a traditional sockets interface. Mellanox’s ConnectX NICs can connect to both 10gE and InfiniBand networks.
Doherty notes that while some firms demanding the lowest latency will opt for co-located data feed handlers, there is also increasing demand by trading firms for an infrastructure that can deliver data with both low latency where required and more widely with reduced cost.