For its supporters, cloud is more than a tool to reduce the infrastructure costs of producing and consuming market data; it also provides an environment to experiment, drive innovation and break down barriers of entry, while offering a greater selection of sources, content and data that can be fed into the trading lifecycle.

With all these things considered, it’s easy to see why many banks have performed a U-turn on their attitudes toward the cloud. As recent as 2014, the consensus was an attitude of wariness, questioning whether the appropriate security controls were in place around public clouds, leaving firms reluctant to exploit it. 

But now the attitude is changing. As an executive from a global bank recently told an industry gathering, the nagging feeling that banks may be missing an opportunity by not exploiting the cloud far outweighs concerns over safety. Over the last three years, time and resources have been invested in exploring how to safely use the public cloud, which is now “no more risky than the environment we run ourselves,” the executive said, adding that his bank is now deploying wholesale banking applications in the public cloud.

“The truth is that many banks are experimenting with the cloud in the back-end. They’re not too public about it because they don’t want to get scrutiny over it, but once they figure it out, they’ll start getting very public about it,” says Stephane Dubois, chief executive of web services data provider Xignite.

Regulators such as the UK’s Financial Conduct Authority, and FINRA in the US, have adopted a progressive approach to the public cloud, using it to store trade and execution data. The FCA said in its July 2016 Finalized Guidance document FG16/5—Guidance for firms outsourcing to the ‘cloud’ and other third-party IT services  that “cloud can provide more flexibility to the service that firms receive, enabling innovation and bringing benefits to firms… and the wider market.” 

“The public cloud debate is over,” says Adam Honoré, chief executive of financial markets and technology advisory firm MarketsTech, who previously served as a managing director at Nasdaq, responsible for the exchange’s FinQloud financial markets-specific cloud offering. “Today there is nobody who says we can’t do cloud anymore…. The challenge is that it takes people a lot longer to adopt to cloud than they think it’s going to. The question isn’t whether or not they do it or what the value proposition is; it’s really ‘How do we do it?’”

Excitement Vs. ROI

A global market data manager at a large international bank says one reason why cloud is potentially exciting is that it “creates opportunities to access certain newer technologies for proof of concept or production without heavy investment,” citing Amazon Web Services’ upcoming AWS F1 cloud, which offers access to FPGA compute resources. In addition, cloud “helps in bringing the utilization level of current assets to a more reasonable level. There is lots of processing required at certain periods, like month-end or quarter-end. But having equipment sitting idle just to handle those usage surges does not provide a good return on investment,” the data professional adds.

If banks want to be at the cutting edge of technological advancements, they need to embrace cloud, says the head of market data at a North American bank. “The companies that are running the cloud… are adding new servers and functionalities way quicker than we could do internally. If you are trying to be at the forefront of new technology, they’re likely to deploy it quicker than we could get at it internally in the first place,” he says. “They’re leveraging their mass scale in a way that we could leverage for agility purposes. There is a definitely an innovation piece—not just a cost-saving piece—that you’re getting.” 

These arguments are the factors swaying firms’ decisions. According to Celent’s 2016 The Cloud Comes of Age in Capital Markets report, “firms of all sizes are looking for easy ways to access data and test theories in a low-cost, low-risk fashion.” Brad Bailey, research director for capital markets at Celent, and one of the paper’s authors, says that “for a large bank it could take six to nine months to get a new server up and running… whereas you can spin up an instance in the cloud in minutes, or even less. From an investment perspective… it gives them a lot of ability to test.”

The global market data manager says another interesting development is that as applications move to a portable cloud stack, “people are also finding that they can actually operate private clouds more cost effectively than using the public cloud. So, for the larger institutions, private cloud and the technology to develop for, manage, and operate the private cloud, will become very important. The public cloud will be used as an extension for the surge capacity requirements.”

However, he adds that one of the biggest challenges to cloud adoption that does not get enough attention is the “wholesale change in the way large companies operate their large processing tasks. This requires cultural change, a new set of tools, and a new way of developing applications. The applications need to become more portable and designed such that they can be deployed using a similar stack as the cloud vendors.”

The recent wave of banks setting up “Innovation Labs,” usually in more than one region, reflects how firms are actively experimenting with cloud resources, says the North American data head. “That is really about hiving off a bunch of employees who are trying to test and build things up in the cloud external to the bank to see what they can or can’t do” without having to work under the scrutiny of internal risk and information security restraints, he says. “They can just figure out what they could run up there and what they couldn’t. Once they get to the point of ‘This looks like a viable opportunity,’ then you would transfer that over to people who are in the bank…. We have several innovation labs and several other in the pipeline,” he adds. 

New Datasets: Test Fast, Fail Fast

Ollie Cadman, head of product and strategy for EMEA at low-latency data platform and feed handler provider Vela Trading Technologies, says one of the key benefits of cloud is that its capacity to allow firms to dial up compute resources quickly enables them to test fast—and, should an idea fail, to “fail fast” before it has consumed budget and resources. Vela, for example, is seeing most interest from firms looking at new markets, services or different ways that they can approach data. 

“Datasets are evolving. We’re seeing new datasets created around sentiment, and from new sources like GPS data. I think cloud-based delivery really suits those types of services,” Cadman says. “People will very quickly be able to assimilate that data and understand whether it’s useful to them, and give them an advantage or a use. In that sense, it’s really going to enable the key integration of new datasets in a very efficient and scalable way.”

The cloud makes it easier to access and exploit emerging, alternative datasets, offering the potential to “flip the market data industry on its head,” Dubois agrees. “If you have an alternative dataset, anybody could consume it through the cloud…. It also makes it easier to sell that alternative dataset,” he says, adding that the vendor is seeing an increase in the number of cloud solutions for market data ingestion, including making Level 1 and Level 2 exchange data streams available via public cloud solutions for storage and analysis purposes. 

Cloud resources also present a potential solution to the storage headache created by the explosion of data volumes over recent years, says the North American data head. “If you think of all the market data that exists in the world from all the exchanges, every bank would then store that data 10 times over internally, so you’ve got huge data repositories occurring. Do they all need to exist locally in each bank?” he says. “If you could move all that data centrally offsite—because it’s generic data that’s coming from the street in the first place—why wouldn’t you want to do that? Because there would be cost savings to be had by having data repositories that live outside of the bank that you just dip into as needed.”

Some firms have already made this a reality. According to the Celent report, “accessing commoditized data such as reference and market data through the cloud is already taking place,” with firms being able to run and back-test algorithms, or perform risk analytics using data stored in the cloud. 

Limitations

However, due to the capacity limits of cloud development at present, the latency of such data—for the moment at least—is measured in seconds. Currently, most of the tasks that use cloud data are those that include end-of-day or next-day tasks that involve historical analytics, Bailey says.

Jeremy Hansford, head of cloud and platform services at UK-based data consultancy and systems support provider CJC, says customers are asking how the cloud can help them simplify all the different market data feeds, delivery mechanisms and connection types that they currently use, and “enable a subscription-oriented model around the streaming of data that goes to the different users to meet their requirements.”

However, the industry is still a few years away from that reality, Cadman says, despite Vela seeing interest from smaller, regional banks and smaller sell-side firms in wholesale consumption of market data through the cloud, as well as from tier-one banks around being able to provide some of its services via REST-based APIs.

More important, he says, is how participants define wholesale market data consumption through the cloud. “If we’re talking about a pipe that you can effectively just take, I think we’re still two or three years away.” While a number of vendors offer API-based market data calls, these remain individual calls, so are unsuitable for wholesale enterprise data consumption. “I think being able to take a real exhaust of market data across a number of key markets, similar to how you would take it today over a cross-connect, is still a few years away. Today, some firms would still struggle to take that over a VPN. Where we’re seeing more of the use cases initially is around the request-response and being able to scale up to that and maybe some incremental streaming. It’s going to be a while before you can get a performance Level 1 US feed streaming over the internet—and there are some technical challenges involved in that as well, not just bandwidth.”

Missing Link

Technical barriers aside, for the cloud ecosystem to fully develop, exchanges must be a part of it, and would bring a whole new level of efficiency to the proposition. But exchanges’ attitudes to the cloud vary, with some being more aggressive than others, Dubois says. “Many of them are like banks: they’re experimenting, they’re looking. They have big infrastructure they care about… and at some point in time they’ll say ‘We have a new engine,’ and they’ll start migrating to the cloud… but it’s not going to happen overnight.” 

Instead, Dubois says he envisages that exchanges will gradually migrate their own data ecosystems piece by piece until they reach a tipping point. “They’ll start migrating some new applications there, putting some data there, feeds, putting PoPs in the cloud… and if you’ve got PoPs in the cloud, that means anybody in the cloud can access your data. They don’t have to go co-locate in Sydney to get your feed; they can pick it up from Amazon or Google. That saves a lot of money and effort,” he adds. 

The North America data head says exchanges are looking at ways to store and make their data available in the cloud: “Somewhere that you can then access, and if you’ve got a license in place with them, you can go and access that data, and you wouldn’t need to drag it down and store it internally…. Why would I keep all the data on-site locally as long as I can access it within the cloud somewhere from the provider who gave it to me in the first place?” he says. And though he believes exchanges are close to achieving this, he notes that this would not yet cover real-time price data because of the latency constraints. “If you’re latency-sensitive, you’re going to be getting your data in the co-lo. This is more about if I need data from yesterday, where would I go to get that, and why would I keep it on-site? Why wouldn’t I just go and leverage it off Nasdaq, for instance? They… all have to keep the data as well as give it out to us, so why wouldn’t they just open up where they keep it and allow us through a subscription service to access it?”

In addition to making data access more efficient for end users, exchanges migrating their data infrastructures to the cloud would likely drive consumption of their data, Dubois says. “One of the big issues for firms, if you want to access US data in Asia, for example, is the cost of the connectivity,” he says, adding that he expects to see the first major exchange move its matching engine to Amazon Web Services’ cloud within 18 months’ to two years’ time.