Network latency is the time needed to effect a communication between two network nodes. It is typically associated with the distance the signal must travel, but it can also be affected by delays introduced in network routing or inefficiencies in the trading or communications protocol. Inefficient protocols create excessive amounts of data which can take longer and be more expensive to transport.
Network latency is a key area of interest at present, given the potential for latency to increase with the number of intervening systems (trading rules engine, OMS, FIX server, etc.) and physical distance travelled between systems. Every processing and distribution system that data interacts with on its journey to and from a trader's desk adds latency. Data packets can also be delayed by slow routers queuing due to lack of capacity or inefficient paths that pass through multiple routing points before reaching their destination. For example, the bandwidth of an area within a trading firm through which the data packets would normally pass quickly, may be disrupted during activity from other users sharing that bandwidth during the trading day. As a result, the
packet would need to 'queue' at each router before processing. A suitable analogy would be the effects of rush hour on everyday road traffic. Using this scenario, there are cases where travelling on a toll road such as BT Radianz, which is limited to paying customers, may result in a faster connection than using public roads such as the internet.
Combating the Issue
It has been common practice in the equity markets for some time for trading servers to be co-located with close proximity to the main stock exchanges. For the reasons outlined in the last paragraph, similar practices are now being adopted by high volume FX players. If a trading platform's data centre is in the same physical location where a market maker already has a presence, or alternatively co-location space is available, algorithmic market participants can minimize latency by cross-connecting from within the data center. Co-location means a counterparty's trading infrastructure is hosted at, or very near to, the same data center as the trading platform. The co-location facility will typically take care of power and connectivity, while the market participant takes care of all the data and applications required. For example, a market participant based in Chicago may use an algorithmic trading server he/she has in New York to get to the main exchanges and FX trading platforms.
For those players who do not have the capability to co-locate themselves, technology companies are moving into providing services to meet their needs. BT Radianz is introducing a new service offering outsourced data center facilities to trading customers co-located at their major network nodes in New York, London and Tokyo. The service will offer customers the ability to place their trading servers in much closer proximity to the communication tunnels served by the BT Radianz network, thereby eliminating key causes of network latency.