Introduction

The TABB Group estimates that financial firms spent $15 billion in 2009 on data center relocation and consolidation to reduce electronic trading latency. This latency is reduced by either shortening the geographic distance between their trading engines and the major exchanges, or by collocation.

The next logical step is reducing latency on Wide Area Networks (WANs), which explains the flurry of activity in private network builds and new ultra-low- latency managed services in the first half of 2010. Much of this activity has focused on new or upgraded long-haul networks between Chicago and New York. These networks and services feature ultra-low-latency connections promising as little as 13.3ms of round-trip latency—crucial to reducing propagation delay and delivering a competitive advantage for high-frequency traders.

To address this market demand, network equipment vendors and service providers continue to innovate and squeeze latency out of parts of the network using various techniques, such as:

• ‑Ultra-low-latency Gigabit Ethernet (GbE), 10GbE, and eventually 40GbE and 100GbE connections over fiber

• ‑Ultra-low-latency optimization of line systems, including amplification, regeneration and dispersion compensation

• ‑High-performance optics for longer reach

A New High-Frequency Trading Battleground

While lowering network latency at all costs makes sense, eventually it will reach its limits. At that point, competitive advantage in ultra-low-latency networking will shift to a new competitive differentiator: network availability.

While reducing network latency is today’s urgency for high-frequency traders—to the extent they’ll swap out recent equipment investments or switch network providers—the focus will soon turn to eliminating the cost of downtime. Reducing latency becomes much less relevant to high-frequency traders if their network or service is down. Simply put: a high-frequency trader could have the lowest latency trading connection in a market, but if the network is down the transaction doesn’t just get delayed, it never makes it.

In some cases the quest for ultra-low latency can turn traditional network design notions on their head. For example, network resilience concerns the network’s ability to diagnose itself, identify possible anomalies, and recover or “fail-over” following a disruption.

For soft failures, like laser aging or a pinched fiber, it’s critical to determine if these are occurring before they become hard failures in the middle of the trading day. Some transport solutions will strip out performance monitoring because it adds latency, leaving networks vulnerable. Fortunately, more innovative techniques are available today to provide performance monitoring without sacrificing latency.

For hard failures, such as a fiber cut, a traditional network design will include a <50ms failover to a route that is geographically different from the lowest latency primary path. In this case, however, the new route will be slower than the primary path, so rapid failover loses its significance if the new route is too slow to begin with for the trader to compete.

So, just as reducing latency requires that various techniques be employed in different parts of the network, protecting a network from downtime requires a similar complementary approach using highly-reliable network equipment, professional services, technical support and management software. Critical building blocks of this approach include:

• ‑Network equipment – the foundation of network reliability, equipment must provide at least 99.999% availability

• ‑Professional services – a diverse portfolio of services delivered by network experts that can address the full lifecycle of network and service design, management, and ongoing optimization

• ‑Technical support – 24x7x365 on-site or remote proactive network monitoring and fault detection with fast response time

• ‑Intelligent management software – provides traders access to the historical or real-time health, status, and performance of their network

The combination of these approaches ensures that high-frequency traders are served by networks that minimize both latency and downtime, while maximizing their trading advantage.

Malcolm Loro is Director Product Marketing at Ciena,

613-591-7854; email: mloro@ciena.com; web: www.ciena.com.