HFT

The GATElab suite for ultra-low latency High-Frequency Trading (HFT) provides sophisticated technological tools for trading. It is characterized by several state-of-the-art features that can be summarized as follows:

  • Ultra-low latency software risk gateway for normalizing markets access, controlling end-user operational risks and forwarding orders to the specified market. All actions are performed in less than 20 µs wire-to-wire
  • High-level IDE to build, test and deploy high-performance low-latency algorithms, distributable on different collocated sites, the algorithms can analyze incoming market data and implement proprietary trading strategies in less than 50 µs.

The GATElab HFT suite can be used by:

  • proprietary firms / proprietary trading desks (for example market making, event arbitrage, statistical arbitrage):
    • analyzing market data to capture trading opportunities that may open up for a fraction of a second
    • holding a position for very brief periods of time and rapidly trading into and out of those positions, a large number of times a day
    • competing on the basis of speed with other high-frequency traders, for small but consistent profits
    • closing investment positions at the end of a trading day
  • brokerage firms for their HFT customers and their internal proprietary trading desks:
    • implementing “sponsored access” offering low-latency direct access to markets, with/without any pre-trade risk management, with/without any algo development support.

HFT Risk Gateway

The HFT Risk Gateway is the GATElab software low-latency high-throughput solution for normalizing markets access, controlling end-user operational risks and forwarding orders to the specified market.

For both software and hardware version, the users of the GATElab Risk Gateway HFT can benefit from multiple FIX connections (Full 4.2 and 4.4 Dialects) and multiple market connections to:

  • receive end-user orders
  • process them against end-user risk limits (closed-form-rules), using market data
  • store for audit and trail purposes
  • execute orders into desired market
  • return back status to broker and end-user

There are more than 15 risk checks for Equity, Forex, Futures and Options including real time initial margin calculations.

Latency is all important: speeding up the process of receiving orders, validating them to control risk, storing them for auditing purposes, and then sending them to the target market. Therefore, in these scenarios:

  • the HFT Risk Gateway Software solution can show:
    • a maximum latency of less than 30 µs (NIC to NIC) one way
    • <30 µs to receive, store, analyze, validate and send orders to the market
    • <30 µs to receive, store and forward trades to end-users and brokers

From a technology point of view

  • the software version is a Linux based multi-threaded single process with:
    • multiple Risk Gateways that can be deployed to the same server following a CPU core ratio to sustain a continuous less than (30+30) µs latency without internal queues
    • active-passive nodes, with node switching <500 µs: implementation has been adopted for handling any kind of faults

HFT algorithmic environment

The HFT algorithmic environment represents the evolution of algorithmicpath (more info...) to stress two main concepts in terms of high-performance low-latency:

  • Complex Event Processing (CEP) environment which is the core of the architecture, focusing also on its distribution and cooperation over a network (LAN, WAN) in order to provide a unique, distributed, consistent layer to the running strategies. In this scenario the CEP architecture is based on a high-performance distributed blackboard (across several servers) used not only as a distributed repository for low-latency market data and historical data, but also to share internal data produced by each strategy implementing local or remote cooperation between them. The CEP engine can process high volumes of fast-moving market data [notified through the traderpath (more info...) DMA platform] from several sources and perform actions in the market in tens of microseconds to decide, monitor and analyze execution activities.

This allows traders to write collocated cooperating distributed strategies: i.e. one reading market 1 data events, signalling checked markets behaviour to remote collocated strategies to trade on their monitored markets and vice versa.

  • Graphical Integrated Development Environment (IDE) to further enhance the process of creating, testing and deploying strategies. This allows traders to:
    • create strategies that can cooperate between them either locally or over a network, while graphically deciding the testing and deploying-running locations (markets collocation)
    • compile the tested strategies into machine language to decrease the latency of strategies’ computational timeframe: from the 300-500 µs of the canonical environment to the 30-50 µs of the HFT environment.

The IDE is made up by three modules: Point & Click Event Editor, Guided Action Editor and Run-Time Environment.

The Point & Click Event Editor easily allows to graphically depict AND-ed or OR-ed Events-Actions behavior of the strategy with input, state and relation parameters. The Action Editor allows writing actions in a language sensitive environment using enhanced Python. Using this IDE, end-users can focus on business logic (described in terms of events and related actions) rather than on other relevant programming aspects.

HFT FEATURES

  • Low-latency infrastructure for high-frequency trading
    GATElab delivers a high-performance low-latency distributed trading solution including risk control of a brokerage or proprietary trading activity
  • Message rate and performance
    The technology infrastructure can handle millions of incoming messages per second, analyze them either for risk control or to identify market opportunities and respond to market changes within microseconds
  • Fully redundant environment
    Resiliency and reliability are key factors in this platform’s capacity to handle huge amounts of transactions per second
  • Software and hardware technology
    GATElab chose to build a mix of software components, tightly coupled with processor cores, memory, I/O and hardware components and implementing Protocol Offload Engines
  • Speed-speed-speed
    GATElab has achieved a substantial increase in overall application throughput (algo environment in conjunction with market risk gateway), end-to-end median latency under 100µs for a software based architecture