Every FX trading firm has the equivalent of one or more major multi-lane highways running between it and other market participants. Increasing the number of lanes and their width is a simple way to ensure the flow of trade is continuous and smooth. However, smart network operators are increasing trade by changing the way traffic moves, how it moves and by adding new services that they can offer to traders, via intelligent networks.
“Intelligent networking is about allowing customers to access the applications and services they need when they want to,” says Yousaf Haafeez, director for capital markets development at BT. “So if we take a look at FX, we have connected almost all of the big FX venues around the world, giving our customers the ability to trade on all these FX venues when they like and to trade the currency pairs they like, when they want to. The intelligence for us is about giving customers the ability to access the services they need to trade or receive the required market data in the FX environment.”
Both buy and sell side are starting to adopt the global connectivity model in foreign exchange trading. Its use is broadening from the equity market which has led both adoption and innovation in electronic trading, adopting the idea of single connectivity to one extranet and multiple destinations within it. That has been facilitated by the use of common messaging standards such as the FIX Protocol which allows buy- and sell-side firms to communicate with one another for transactional purposes.
“Over the last few years we have seen a lot of take-up, in several ways,” says Alex Walker, executive vice president and managing director at TNS. “In the FX market there has been a lot more FIX connectivity. In the equity markets it’s been going on a very long time but it’s really only been in the past few years that FX has embraced it as wholly.”
Rapid adoption of messaging standards has been supplemented by both sides of the street moving away from the bilateral model with dedicated trading connections to counterparts.
“They have wholeheartedly embraced the extranet model which in addition to the usual benefits of super-fast connectivity and reliability gives them the ease and flexibility of being able to connect to a variety of direct trading counterparties as well as ECNs” says Walker. “What we have witnessed over the last few years in the FX world is that the adoption of electronic trading has really accelerated from a fairly low level to now, where the overwhelming majority of trades are done electronically. In addition we have observed the global expansion, in particular of electronic communication networks (ECNs) who typically originate in the US, build up a lot of traction there and then expand into Europe and beyond to Asia.”
Simple yet effective
Adding more trading superhighways creates cost and complexity for the firms that operate them. However several factors mean having a greater number of trading connections – via as few pipes as possible – is highly desirable, including: the growth in number of potential counterparties; an increased awareness about the risks involved with the world’s largest banks; and a growing awareness of and concern about operational risk.
“The theme we are seeing is a much higher demand for connectivity to more venues, to more sources of liquidity, to more counterparties,” says Andy Young, specialist sales director, capital markets, at network operator Colt. “So there is a whole network connectivity piece becoming much more complex and bigger. Clients are asking us for easy access to multiple different sources now, whether that be venues themselves, or electronic communication networks (ECN), single bank platforms, or other members of the community in the ecosystem. We need to connect to those multiple sources in an easy way over a single access mechanism can flexibly get them to many different sources.”
The firms demanding access to intelligent networks have been changing – where the demographic was once traditional Tier 1 banks, who continue to be the major providers of liquidity in the FX space, Hafeez notes that smaller firms, particularly algorithmic traders, are proving particularly interesting often in locations not associated with the major financial centres.
“We are seeing customers in places like Israel and Turkey, coming onboard to trade FX 24/7,” he says. “One of the key requirements for our customers is the ability to trade on a global basis. It’s not just about trading in the four key FX centres, London, Singapore, Tokyo and New York; it’s about the ability to trade globally. They want to be able to access venues or brokers where they are located. So from our perspective globalisation of the FX market is key for our customers and they want the flexibility to trade in any location and any currency pairs they need to.”
Flexibility must be reflected in the capacity of the network to support changes in trading volume, expanding to meet the level of trading messages that can be passed back and forth between two counterparties. Walker notes that a price maker does not necessarily know the changing dynamics of its price takers requiring a considerable reserve capacity in order to best service its clients. The ‘intelligent’ part of the network is ensuring there is no need for time consuming interaction between the network provider and the client.
“It becomes a big thing where they want to ensure that if the counterparty does need more liquidity our client is able to meet that need quickly and they don’t have to come to us reporting any packet loss [of data] and asking to be authorised for more,” he says. “That is why we have a dynamic bursting model whereby a provider of pricing pays for a certain bandwidth, but he isn’t limited if that taker starts using it more.”
The dynamic system automatically allows them to have as much bandwidth as they want up to the line speed available, which is typically a gigabit at a time. Interestingly some of the other players in the space who used to be able to provide the bursting model have stopped doing it.
“One of our main competitors moved over to a new common shared platform which didn’t really support bursting,” Walker explains.
Automation is also key in supporting transparency and self-governance, Young notes, with Colt’s PrizmNet clients accessing a portal to view their own network in order to self-provision.
“They can manage their own networks without the old model of calling an account manager within their telecom service provider, order a new line, and wait for 20 days or more to get the line installed,” he says. “Now once you are a member of the extranet with a portal, you can go on and price up your own services, once you are on the network you can actually provision new services to additional counterparties yourself, without any manual intervention. Issues or challenges on the network can be handled automatically through the portal.”
In many ways the industry is seeing the level of service provision expected at a consumer level extended to the enterprise level.
“Everyone individually expects to be able to order their normal things at home on the internet and the same is happening from the connectivity point of view,” Young observes. “With intelligent networking becoming more mature, the sort of speeds at which new services are spun out become much more accelerated. It’s a game changer really.”
Right type of road
Getting the right amount of speed and reliability is down to the capacity of a network to support low latency and fail over to new services. For firms playing in the high-frequency trading market – with high volume, low order size and high message cancellation – there is a certain latency performance that is needed as a given.
Keeping latency low and consistent does impact certain trading strategies and there has been a gradual migration from proprietary datacentres to commercial centres that have aggregated industry players and thereby created a lower latency cross-connect ecosystem.
“We have seen the desire to reduce the latency; it started off with buy-side firms operating in their own dedicated data centres outside of the core liquidity sets [of the major data centres] in London, New York and Tokyo,” Walker says. “Then they wanted fast dedicated connections into those [three data centres], and in the last couple of years we have seen more and more of those clients wanting to move into those centres.”
The result has been a migration over the last few years towards a more managed hosting offering where those clients, who have established trust with a network provider, decide they no longer want remote access, but actually want to be in data centre such as those run by Equinix.
Some customers are requiring very low latency, says Hafeez, requiring a specific set of services primarily targeted to those customers. Where customers want the absolutely lowest latency they need to be co-located where the FX venues are.
“For them we provide hosting capabilities so they can be next to the venue,” he says. “Because the FX community is so big, we also provide a managed cross-connect service in key financial services data centres around the world. In some data centres we can have customers who will cross-connect to around 70 other participants to allow them to trade FX. We will manage each of those connections to monitor their latency and performance.”
Routing messages effectively requires a solid understanding of the impact that different technical options can have on trading. Where networks can become a single pipe for all traffic that simplicity can cause problems with speed. Consequently traders may dedicate one network purely for market data and trading, with video via a separate network, so there is no risk of any contention occurring on the trading network. Multiprotocol Label Switching (MPLS) networks will route data via less congestive paths and have been around for a long time as an option, however as an option for traders concerned about consistent speed of delivery, they can create challenges.
“We use it within part of our network as well, although a lot of our network doesn’t use it,” explains Walker. “MPLS networks are often built to the benefit of the carrier or network provider, ensuring that less used routes are used. For real-time data and latency-sensitive data they aren’t ideal, because you want the traffic always going down the fastest route. Our method is not to prioritise traffic but to have a network that is scaled fully and where there is always enough bandwidth for everything to go that fast.”
Hafeez says, “We provide customers with the connectivity which allows customers to do whatever they want to. If they want to carry instant messenger traffic, that’s no problem. If you want to carry video or voice over it, that can be done. But we find – particularly for some of our latency sensitive traders – they want to keep their trading network separate from other networks.”
Walker concurs noting “We haven’t really seen any demand from our clients to converge networks and to bring voice over it. They tend to use us because of our reliability and the fact that they know the latency is going to be good; they know that in the event of an outage it’s going to re-route where it should.”
Keeping on the straight and narrow
Having certainty around service provision requires that trading firms, when selecting the strategy for their connectivity within the FX markets, make sure there is not only a certain latency performance but also determinist latency.
“That’s a key benchmark,” says Young. “Many of the applications in the space need consistent performance as well as low-latency performance, so ensuring that performance is going to be the same every day of the week, is vital.”
Clients want more reliability and more resilience. While the market has moved more to the datacentres in London, New York and Tokyo there is a school of thought that if everybody is there, firms do not need resilience beyond those locations because if something dreadful happens then everyone is out of the market. However, over the last couple of years the regulations have tightened up and increasingly clients think they need more resilience either in the region or out of the region.
Walker says, “That means they ask us to just host a small engine of theirs, or a pared down version in New York so they can dynamically fail over from London to New York in the case of something dreadful happening. It used to be that that was within the region, for example a failover from one data centre in London to another data centre in London in the event of a disaster. Now they want a wholesale move so if London shuts everything moves over to New York or Tokyo. They want that in a low-latency fashion as well and to move over as quickly as possible, in an automatic fail-over scenario or in a planned, elected scenario.”
Events such as Brexit have been test cases for capacity with networks providers reporting that clients wanted to have the ability to ensure that their messages and trades were going through as that activity was mounting.
“If you have a bottleneck on your connection which impacts the volume of trade or traffic going through, you are as good as out of the market,” Walker concludes.