OPRA Raw Filtering and Feed Handling with NovaTick™
A Comprehensive Solution
OPRA feed is one of the most challenging data feed to process for a ticker plant.
NovaTick solution for OPRA feed is fast, scalable and customizable, resulting in higher reliability for software applications trading on US Options.
NovaTick supports 96 lines/channels along with the 40Gbit input (100Gbit soon).
NovaTick offers two use models: Raw Filtering and Feed Handling.
Raw Filtering is used to reduce the amount of market data going into the trading application and does not require any change in the trading infrastructure other than inserting NovaTick as a front-end for the application.
Feed Handling is used to accelerate and facilitate the ingestion of the market data by the trading applications. Our solution simplifies and normalizes the OPRA raw feed so that it can be processed in a single thread on one core of the trading server. Trading algorithms access the market data faster, and process exactly the data they need. Furthermore, the whole software trading stack is accelerated as more processing resources are freed up for its execution.
Two Deployment Options
1 – OPRA Raw Filtering on FPGA PCIe Card
- Filtering on Exchange (Participant), Symbol (underlying equity/index and related options contracts), Message type, Quote/Trade conditions, NBBO
- Output in OPRA Raw (Pillar format) on PCIe bus or Ethernet 10G
- No change in your trading infrastructure other than inserting NovaTick PCIe Card as a smart NIC card in your trading server
- Very small additional latency for filtering raw data largely compensated by the boost of the trading applications
2 –OPRA Feed Handler on appliance with 4 FPGA boards (2U appliance)
- Split on each board is done based on multicast channels and symbols (underlying equity/index and related options contracts)
- Support entire universe (Up to 2,2M symbols x 17 Exchanges)
- A/B arbitration on all 96 lines
- Output distribution features: NBBO on specific interface or multicast channels, split quotes & trades…
- Normalization in NSWF or NSMDA C++ API
- Average wire to wire latency at 746ns
- Distribution to multiple servers through 40Gbit or multiple 10GBit Ethernet outputs
