Market data distribution is critical for financial markets. NovaTick not only processes the market data in FPGAs, but also distributes it directly from FPGAs.
Packets and messages (with the relevant information for the users) are duplicated and filtered before being delivered through a high number of Ethernet interfaces to trading servers or trading FPGAs.
Messages can be published in two ways: Either they are sent at the end of the incoming packet, or they are published as soon as they have been processed with a cut-through mechanism. The second publication approach offers a lower latency profile when the packets consist of multiple messages.
The performance of the NovaTick Ticker Plants remains unaffected whether each packet/message is sent out to one interface or twenty. Thanks the parallelism capability of FPGA and the NovaSparks FPGA Matrix architecture, a packet/message can be replicated as many times as there are output interfaces in just one clock cycle.
Software feed handlers typically copy the packets/messages as many times as they need to be available on the different output interfaces, creating high latency spikes, especially during market data peaks.
Coupled with instrument and exchange filtering, this one-clock cycle replication capability allows NovaSparks customers to customize a dedicated output stream for each of their trading applications, without impacting the overall latency of the ticker plant.