Like strata of wildlife on a hot savannah, traders seek out liquidity pools across the FX landscape in different ways and with different motives. Consequently, they are evolving into highly specialised beasts. For some, the emphasis is on finding liquidity reliably, with broader parameters on spreads and speed as a result. Others need to monitor markets every millisecond so sub-millisecond opportunities can be pounced upon.
Jay Hibbin, regional sales director for Financial Services in EMEA at CenturyLink says, “There is commonality around access to liquidity. I think for traders, picking a network provider who has got reach to a number of platforms and venues, multi-broker platforms, but also single-source platforms as well, and the big banks is important. So the broader the reach of the network provider the better, and I think that’s across all trading firms.”
However at a more nuanced level there are clear specialisations which affect firms’ infrastructure use. For retail FX brokers the development of a connectivity network in-house would be like putting a camel’s hump on a hippopotamus, unnecessary and inconvenient.
Yousaf Hafeez, director for Capital Markets Development at BT says, “A lot of these brokers will be looking at a managed service, simply because it’s significantly cheaper than going away and doing it themselves. They like the idea that they can pass all these complicated technology requirements onto a company which specialises in that field to guarantee that performance.”
While there is considerable variation In terms of cost, he says connectivity alone could mean a far higher rate compared with a managed service, depending on the route. However a managed service can be about a far greater set of services than connectivity. The whole technology environment needed to run a business may be included.
“One of the things we have in our portfolio is the ability to manage their hardware, their software up to the operating system (OS) level and including the OS level and also to look at their current inventory and rationalise some of the infrastructure for them, by using the cloud or by using virtualisation of the existing servers,” says Hafeez. For a large number of retail FX customers the internet is adequate for posting a few trades per day, or even per week. However some key customers may say they want a more robust, solid service as their levels of activity begin to reach the limits of typical retail activity.
“At that point retail FX brokers may ask for a higher quality network, and what they find is by offering the capacity to handle higher volume retail users on their network, they receive an increase in order flows,” Hafeez explains. “So as an example, over the internet they may have been getting 50 orders a day, and what they see by using a more robust network may double that.”
Even where they increase the robustness of their infrastructure, retail brokers will typically not require the same levels of low-latency performance that is needed for firms that are running highfrequency, proprietary-trading strategies.
Ralph Achkar, Director for strategic alliances at Colt Capital Markets says, “Hosted services may not have the sort of low latency that one might need when determining high-frequency trading (HFT) for example. However if you are offering a retail FX trading service that’s not really going be so much of an issue. Then of course there are other considerations which might be around security, or around data transfer.”
THE IMPORTANCE OF LIQUIDITY IN LARGE SIZE
Institutional traders and brokers typically trade in larger size, which can risk market impact where liquidity is poor, or at small size but higher volume, which creates a high level of order cancellation. A more nuanced view of liquidity is therefore very important for these traders. In the FX space liquidity is increasingly becoming an issue as traditional sell-side firms engage less in the market.
In its 2016 Liquidity Survey, State Street found that 49% of respondents expect the importance of non-bank liquidity providers to grow. The risks that poor liquidity creates in the market was highlighted on 7 October when Sterling plunged against the dollar and bounced back in a ‘flash crash’. While the circumstances are still being investigated by the Bank of International Settlements, previous flash crashes in other instruments – including US equities and US treasuries – were attributed to a combination of dealer withdrawal, highly automated trading and illusory or phantom liquidity.
According to a Bank of America Merrill Lynch report issued on 5 October called ‘The illusion of liquidity’, these were the same circumstances that the FX market was facing at the point of the flash crash. For firms that both need to engage with HFT liquidity and distinguish between phantom and real liquidity, having a highly connected set of market places to support price formation and liquidity access is imperative.
“I think for some sub-segments like the HFT firms, obviously there is a much narrower focus around latency and where the network is interconnected to, for example particular data centres,” says Hibbin. “That discussion can get extremely detailed around how connectivity is put together, and what is the optimum in terms of latency between key liquidity pools.”
For more institutional traders there could be a range of elements that are important to them, including visualisation, management and reporting as best execution creeps into the picture for FX spot and derivatives, for example under the MiFID II rules in Europe. The demand for greater transparency for both regulators and clients can increase the importance of integrated systems for firms, as this supports ease of reporting. “Potentially they may be looking at integration with the application platform, so if it is possible to get a hosted application platform together with a managed network then there are the distinct benefits from having a single solution service level agreement (SLA) that covers both the publishing of rates, capture of liquidity and the uptime of the platform itself,” Hibbin says.
In addition to the withdrawal of dealer liquidity, overall volumes have fallen from non-bank liquidity providers according to the BIS Triennial Survey, which in 2016 found the first decline in spot trading volume since 2001, falling by 19% from April 2013 to hit US$1.7 trillion per day in April 2016. For trades where hedge funds and HFT firms were a counterparty daily turnover fell 29% in April 2016 from April 2013, hitting US$200 billion. That reflects not only the market decline but also the limits that a crowded market creates for HFT firms.
Sylvain Thieullent, CEO of trading technology specialist Horizon, says, “For HFT firms the costs are so high that the game is not sustainable for new entrants, but also for firms that have been in the business for a long time. For example, we saw Teza recently withdraw from the proprietary trading business. HFT firms need to adjust their strategy.”
The extent to which they will be able to engage with operators who provide Managed Services to multiple FX venues will require them to carefully balance the issues of cost with the capacity to offer the right levels of performance. For HFT firm this may mean hosted or Managed Services lack the reliability needed.
“There is an upside in that you tend to get everything under one roof, you get potentially a single contract, a single SLA, faster time to market, and you can completely outsource it – the project management side of it and everything,” says Hibbin. “But the downside is that often you are dependent on that shared infrastructure, so things like change control that maybe being put in place for somebody else could potentially have an impact on you. There is a commonality to the infrastructure that you need to take into account, which can have an impact on uptime and on change if it’s not managed properly.”
Purchasing the firm’s own circuits to each venue privately and individually and then pulling that altogether and managing the infrastructure it needs to connect gives that level of control.
“You are buying your own capacity and you can dictate the routes, you can dictate the bandwidth you need, and you are not sharing that access with other people. Depending on how far you take that you can get to the point where if you buy enough capacity and you specify enough dedication, then you can avoid a lot of the pitfalls of potentially a poorly run managed service,” Hibbin adds.
DEPLOYED FOR PERFORMANCE
Hosted applications are not only of value as an adjunct to a managed network. Changing market structure is driving demand for hosted apps. Retail FX brokers do not have the same data-centrehugging requirements that the institutional firms have, and therefore have more liberty to use cloud hosting in that sense. Yet they will usually have some sort of matching engine underpinning trading which tends to be proprietary, despite the commercially available packages.
“For the most part there is normally an element of customisation, or completely proprietary back end matching services,” says Hibbin. “After that the difference is in the presentation layer, so with the retail clients you are presenting normally through some sort of web based front end, or proprietary front-end system, whereas on the institutional side it’s normally an application programming interface throwing prices out to a multivenue broker.”
Achkar says, “On the institutional side historically there used to be these three centres Tokyo, London and New York, with the majority of the matching engines being in either New York or London. About two years ago we started seeing some of the matching engines of the multibank platforms, asking why they have to have the central liquidity pool for Asian client just in North America, when they could start putting some of the matching engines in Tokyo.”
That creates more distributed hosting needs across the globe, not only in New York and London, assuming they follow the normal hosting models where they buy a group of servers, put them in a physical location and then expect interested clients to aggregate around the matching engine.
“This doesn’t have to be the model for the future,” notes Achkar. “There is a lot more confidence in cloud deployment by capital market participants. Whereas historically they would shy away from anything that is cloud-related because they saw it as either non-performant or lacking enough security, there is more acceptance amongst market participants and consequently there is nothing that would stop FX hubs from getting deployed in a cloud environment.”
Depending upon a firm’s capacity to support performance there may also be no reason that the deployment would have to be in a specific data centre in Tokyo, New York or London. If the infrastructure requirement can be met in terms of performance, security, scalability can be met in a cloud environment there is nothing that would stop a new FX venue from moving there.
Operationally supporting retail or wholesale operations is very similar says Hafeez, it is primarily the scale of the demand from the client that is different.
“For an FX retail broker you are maybe talking about a relatively low number of servers, a relatively small number of racks, whereas if you talk to a bigger larger venue or broker, then obviously that scales up quite radically,” he observes. “But in terms of the actual services they are both very similar and it’s more about the scale of the operation and the requirements. [Servicing] is about clarity around what they both client groups want; if they want a managed solution which takes care of the network; or power servers and the management of those servers and the applications up to OS level on those servers.”
Applications have started moving into the cloud environment. Typically those that are not the most time or latency sensitive are the first to migrate. Yet increasingly front office applications such as testing, modelling and checking of algorithms are also moving across moving into the cloud environment.
THE KILLER QUESTION
Where performance of the cloud environment is able to handle quick response time and scalability, there have long been questions in terms of the security of the cloud environment.
Achkar says, “There have been a lot of technological advances in terms of encryption and how secure hosting and operating in the cloud is. We have seen very few breaches happening in the cloud. We have seen breaches happening in other areas that are not cloud located.” Security is not only an issue for firms that are hosting, there are also security and performance issues around getting into the cloud. Where historically it has been driven by internet connectivity, there has been a push starting 18 months ago by cloud providers to put more secure and reliable connections into the cloud.
“We are seeing a lot of pointto-point connections, we are also seeing some of the eco-systems starting to add these clouds as destinations or services,” Achkar explains. “That means that the performance and the security is extended not just to the cloud environment but also right to the office of the user.”
For some firms looking at the possible use of a hosting solution, the issue of security can be a deal breaker, making it the killer question for providers to answer well.
Hibbin says, “It depends on their internal security policy and how they view off premises client data really, or off premises systems. There is normally a very stringent set of controls around that, and Managed Service providers have to be able to comply with those controls, which in turn are driven by regulators in the market. So they could be everything from physical controls around a data centre to encryption controls around databases, to the security aspects of presentation of data and how that is secured, threat protection, DDOS mitigation, and then ultimately the ability to audit all of that estate and have that presented to the regulator and to the client.”
Although HFT clients may not have a choice in the model they request due to limited capacity for trade-offs, for most other potential clients in the marketplace there are very real trade-offs in assessing the maximum amount of performance and availability, prevention of all security threats and keeping costs under control.
“Almost any one of those requirements could drive costs higher and I think that’s a constant challenge that firms are facing,” says Hibbin. “There has been a push by regulators for firms to prove that their disaster recovery provisions are useable in production.”
All of that is putting cost pressures on firms which means there is an extra additional focus on how can they make best use of cloud technology to bring the cost down, so that they have got the budget to meet their other requirements.
“A lot of effort goes into ensuring that there is compliance, auditability and transparency at every layer to ensure that Managed Service providers who provide that ability can demonstrate to clients and meet the needs of clients both internal and external policies,” concludes Hibbin.