The landscape of electronic trading is in constant flux, driven by a relentless pursuit of speed and efficiency. As execution times shrink from milliseconds to microseconds – and even nanoseconds – the underlying infrastructure has become paramount. A recent webinar hosted by A-Team Group and sponsored by LSEG Data & Analytics gathered industry experts to dissect the critical challenges and evolving strategies in this dynamic environment.
The Evolving Definition of Low Latency
The definition of ‘low latency’ is constantly evolving, driven by technological advancements and changes in market structure. What was once measured in milliseconds is now discussed in terms of hundreds or even tens of microseconds, pushing trading firms firmly into the low latency space.
This acceleration is partly driven by increased market fragmentation across different execution venues and asset classes. For firms operating globally, achieving low latency requires not just raw speed but also the agility to navigate a multitude of market structures with varied order types, trading mechanisms, complex architectures, and frequent infrastructure upgrades. As one speaker noted, the requirement for low latency now exists across a broad spectrum of use cases, from high-frequency trading to compliance.
Key Challenges and Considerations
The Quest for Determinism
For practitioners, the biggest network challenge isn’t just speed, but achieving consistent, deterministic low latency. This means minimising jitter (the variation in network latency) which is crucial for predictable trading performance. While resilience, cost, and global market access are significant, an audience poll confirmed that consistency and determinism topped the list of concerns.
Navigating Critical Decisions
Pursuing the lowest possible latency forces firms to navigate complex decisions. A fundamental dynamic exists between speed and cost, as investing in marginal, microsecond or nanosecond-level improvements requires expensive infrastructure upgrades. Firms must critically assess how fast they truly need to be and whether the cost can be justified.
A similar balancing act arises between speed and resilience. Maximising speed often involves streamlined architectures like UDP multicast, which prioritises low latency over guaranteed delivery. In contrast, more resilient approaches like TCP introduce latency through error-checking and retransmissions. As one speaker highlighted, organisations must decide “the purpose of what you’re doing” and determine their risk tolerance based on business drivers, not purely technological desires.
The Interplay of Technology Components
Hardware Acceleration: FPGAs and ASICs
While the network is critical, it exists within a broader performance ecosystem. High-frequency trading strategies often depend on dedicated hardware like FPGAs (Field-Programmable Gate Arrays) and ASICs (Application-Specific Integrated Circuits) for ultra-low latency functions. These devices deliver performance gains by operating at wire speed, but they require specialised skills and significant investment. Moving to a low latency-based strategy can take months or even years from decision to deployment, forcing firms to weigh the latency benefits against development time and operational costs.
The Rise of Hybrid Architectures
A hybrid architecture is increasingly the norm. This approach blends on-site, low-latency applications for execution with more compute-intensive workflows. Tasks like AI model training are often handled in offsite GPU clusters or cloud data centres, with the resulting models deployed back into colocation facilities for live trading.
Network Engineering for Performance
The underlying network must be meticulously engineered to support this mix. Best practices for reducing jitter include over-provisioning bandwidth to handle microbursts and minimising cable lengths. For critical paths, firms may use Layer 1 connectivity or direct exchange handoffs. Implementing physically diverse paths into data centres and leveraging technologies such as cut-through switches and optimised hardware and software solutions can further improve speed and consistency.
Cloud and Cross-Market Connectivity
The Cloud Question
The rise of hybrid and multi-cloud architectures presents both challenges and opportunities. While some experts believe the cloud is not yet suitable for ultra-low latency use cases, the fact that major exchanges such as CME and Nasdaq are exploring it raises questions about its evolving capabilities. The cloud offers benefits like centralised data storage and access to global networks, but adapting it for high-performance trading may require deployments that more closely resemble a traditional infrastructure model.
A prime example of how firms are currently leveraging the cloud for testing of low latency use cases through using LSEG’s Tick History – PCAP – a cloud-based, 70+ petabyte repository of ultra-high-quality global market data. Tick History – PCAP’s low-latency capture process uses proprietary technology to maintain the highest fidelity with the exchange data coming off the wire. Geared for flexibility, data is distributed in both raw and normalised formats (CSV and Parquet) to support multiple use cases which can save bandwidth and compute costs.
The Need for Global Interconnects
Cross-market connectivity is essential for accessing liquidity across diverse regions and asset classes. As one expert noted, there is “no substitute for having really good interconnects.” While geographic distances impose inherent latency floors, firms can focus on improving determinism within their own infrastructure. Wireless technologies, such as microwave and millimetre waves, are becoming more economically viable for key low-latency routes like the ‘New Jersey Triangle,’ particularly for transmitting small, time-sensitive signals.
Looking Ahead
Looking forward, several factors are expected to shape the low-latency landscape. The first is exploding data consumption. Evolving algorithms and AI will require a significant increase in both bandwidth and processing capabilities.
Market fragmentation will also continue to add complexity. The emergence of new exchanges and venues, particularly in the US equity and options space, makes achieving comprehensive, low-latency access more challenging.
The primacy of measurement and verification will only grow. As one speaker put it, “if you can’t measure it, you can’t verify it,” highlighting the critical importance of network telemetry and monitoring tools. Firms are already exploring AI to analyse historical data and proactively predict potential issues.
Future advancements may include next-generation technologies like photonic computing, as performance gains become harder to achieve. Finally, external factors like power availability in financial hubs, the geopolitical classification of data centres, and the prospect of 24-hour trading in markets like US equities will all place sustained new demands on low-latency infrastructure.
In essence, navigating the sub-microsecond frontier demands a holistic approach. Success requires balancing the need for speed with considerations of cost, resilience, maintenance, and complexity across an integrated technological stack, all while keeping a sharp eye on evolving market structures and innovations.
Accessing high-quality low latency and real-time market data is essential in the financial services sector. Find out more about how LSEG’s solutions can help you solve your low latency use cases: https://lseg.group/466occX
Subscribe to our newsletter
