Speed has always been crucial to successful trading. However, for firms engaged in electronic and algorithmic trading, the speed of the trading infrastructure is almost as important as the trading algorithm itself. Most discussions involving latency optimization revolve around various equipment such as feed handlers or servers that process market data, order management, analytics and risk.
We are going to focus on optimizing the optical transport portion of the trading infrastructure, which deals with the speed of the physical communication connection between geographically separated execution venues and data feeds.
When it comes to connecting buildings, firms are faced with deciding among three options:
- Managed services
- Fractional or dim fiber
- Private fiber networks.
Let’s start by discussing the basics of optical transport and the building blocks that are common to all of these solutions.
The starting point for connecting buildings is the fiber-optic cable. The cable used inside a building is different from the cable used outside a building. Inside a building, you typically use multi-mode fiber and outside you use single-mode. Since this article deals with building-to-building connectivity, our discussion will focus on single-mode fiber.
Fiber-optic connectivity essentially works by sending and receiving pulses of light through a single-mode cable. The speed at which that pulse of light travels is governed by the laws of physics. Einstein determined the speed of light in a vacuum as ‘c,’ or approximately 186,282 miles per second. However, since the pulses of light in a trading scenario are not traveling through a vacuum but through a piece of glass, they are slowed down. The ratio between ‘c’ and the speed at which light travels in a material such as fiber is called the refractive index. The refractive index of light in a vacuum is one.
The refractive index for single-mode fiber can vary slightly based on a number of factors. However, a good estimate is around 1.467, meaning that light travels through fiber at 186,282/1.467 = 124,188 miles per second. As a result, if you can shorten your physical fiber route by one mile between exchange venues, then you would save around eight useconds (microseconds).
To make it simple, in order to get the fastest transport between buildings, you need to have the most direct fiber path. The traditional managed service solution of routing all traffic through an aggregation building (typically called a central office) results in extra cable length and, therefore, a slower solution. The good news is that most providers are now offering a ‘low-latency’ managed service that bypasses this central office and allows direct building-to-building connectivity. Also, remember that most providers, whether dark, fractional or managed, utilize different routes to get between buildings. As a result, you see different latencies. If you truly want the fastest connections between multiple venues, you will probably need to go with multiple providers.
This article is the first building block for optical transport and addresses the speed of the pulse of light in the fiber-optic cable. In the next installment, we will discuss the second building block of optical transport - the hardware necessary to transmit and receive the light.
Read more on low-latency optical transport here.