"Cloud has become the new normal." This was a fitting remark from Andy Jassy, Senior Vice President of Amazon Web Services, during the opening keynote at the 2015 AWS re:Invent conference. His statement captures both the evolutionary and the revolutionary trends behind the growth of cloud computing over the last few years.
On the one hand, it summarises well the evolutionary progress of businesses completely ditching privately run data centres and moving 'all-in' to public cloud computing providers. Or as Charles E. Phillips, CEO of Infor, says: "Friends don't let friends build data centers any more". At some point we technologists have to look at what we can do with core infrastructure; if there is little value that we can add to our current technology layer, then it is time to move up the value chain.
More importantly for the trading community, however, this rapid
growth of public cloud services such as AWS has also kicked off
a small revolution. The basic infrastructure and core technologies around computing, networking, storage and databases have been commoditised on a massive scale. This has been followed by an exponentially growing democratisation of trading and asset management, across all business functions.
Traditional hurdles such as research, data, execution or completely starting and running a fund have been eroding at an ever-faster rate. Technology has traditionally been a substantial spend, yet with more innovative implementations, this has now been brought down to a fraction of what it was several years ago - much cheaper than doing it in-house or with traditional outsourcing providers.
The notion of large upfront capital expenditure, be it hardware or software, is becoming a legacy concept, and rightly so. Whole ecosystems of service providers are appearing around this technology, which can provide specific services for any stage of the trading life cycle, and they are providing these services on-demand and cost efficiently.
Perceptions and Buzzwords
It would help to step back for a moment and demystify some terminology. The term 'Cloud' suffers from a semantic overload, ruined by years of sales and marketing attempts to jump onto the latest bandwagon, creating buzzwords and headlines.
If we peel away the hype, the actual premise of cloud computing is not new at all - in fact it is over 60 years old. The mainframes of the 1950s were the early incarnation of shared, centralised computing resources. At the end of the day, it is virtualised computing, run by someone else.
There has been an often raging debate between 'private' and 'public' clouds. However, the debate has mostly been based on perceptions as opposed to facts, and requires careful picking apart. For example, private clouds for the hedge fund industry, championed by some providers, are as much open to the same security problems as the public ones, if not more so. But they also lack the features, flexibility, innovation and cost efficiencies of the public clouds.
The Challenges and Opportunities
In just a few recent years, the public cloud has gone from an interesting idea to mainstream acceptance, and a de facto standard for the vast majority of businesses today. It represents the largest disruption in enterprise computing since the advent of the PC, and it is moving rapidly into the asset management community. As with any technological disruption, there are challenges and opportunities.
Born in the Cloud or Moving to the Cloud
New firms starting operations these days are naturally inclined to go straight to the public cloud. Starting from a clean slate, they are at an advantage, as they have no baggage of legacy technologies to worry about, and can quickly adopt the flexibility and agility of the cloud.
Existing firms with in-house or private cloud deployments, legacy kit and sunk costs face the bigger challenges. However, as established businesses they also have the bigger opportunity prizes, such as transforming their business model, cutting costs, fending off competitive challengers, or broadening their services.
There are methodologies to help migrations of existing environments, such as the AWS Cloud Adoption Framework. Each business will have variations on these approaches, and overall they need to be applied in the context of asset management and trading systems.
For the vast majority of applications there are few technical challenges to prevent the cloud adoption in trading environments. In fact, there are some notable hedge funds and prop funds operating close to or at 100% using only infrastructure from cloud providers such as AWS, Azure and Google.
Security, Compliance and Regulations
Security is at the forefront of daily operations for many funds, and rightly so. In the past, the security of the cloud has been negatively perceived, but as the cloud providers have matured, this has slowly shifted to general acceptance. In reality, the public cloud providers have much higher stakes protecting organisations with far greater security requirements such as government agencies.
The bonus of hosting such high security businesses is the fact that what is implemented for these entities is commonly available for all users of the public cloud. As a smaller business, it would traditionally be impossible to afford the level of security of bigger organisations. With the cloud everyone can benefit from the same security as anyone else, no matter the size of the business.
A recent quote from Rob Alexander, CIO of CapitalOne, highlights this tidal shift in perceptions. His belief is that the public cloud enables CapitalOne "to operate more securely in the public cloud than we can even in our data centres".
With regards to the regulators, the FCA's recent guidance consultation is a welcome step forward for the UK regulator and has generated a strong response from the financial community. The FCA recognise the ever growing important role of public cloud infrastructure, and they aim to adapt and update their guidance to keep up with technology.
Similarly, other regulators around the world accept the public cloud, such as the regulators in the Netherlands and Singapore, as well as FINRA in the United States. Anecdotally, the regulators themselves are big users of the public cloud.
MiFID 2 is looming; not necessarily as a threat, but more as an uncertainty. Some of the proposed changes, such as the timestamping requirements, could be an interesting challenge to implement reliably. However, the public cloud providers are aware of the value of providing solutions to the financial services industry. After all, hedge funds spend more than 2 billion USD on technology, something that cloud providers should be keen to capture.
Costs, Innovation and Business Transformation
Outsourcing is increasing, and costs have been identified by Ernst & Young as the primary driver. Funds are facing these rising technology costs against the backdrop of shrinking operating margins. The public cloud offers a very compelling way out of this squeeze.
The range of savings vary from the large entities such as FINRA's estimated 20 million USD of annual savings, to the very small funds becoming able to run secure, resilient trading operations on effectively a shoe string budget. In the middle of the pack are the companies that are able to make 20-50% savings, which Amazon Web Services (AWS) traditionally sees with their customers.
These cost savings are not purely based on technological advancements. With a move to more public cloud and SaaS adoption, there is a ripple effect across the business as more functions and teams are exposed to efficiencies, agility, automation and self-service. This drives a significant share of the Total Cost of Ownership (TCO) calculations in favour of public cloud.
The adoption of the public cloud and the ecosystem of features and solutions also provides a driver for innovation and business transformation. Seized correctly, there are far bigger opportunities and cost savings than simply the cost of replacing entire data centres. Firms can adapt to threats far quicker and move in new directions to grasp business opportunities faster and with less risk.
Clusters and Technology Choices
Public cloud computing is truly at its best for rapid elasticity and handling big data. It has an unparalleled ability to provide seemingly endless computing capacity that can rapidly scale up and down, with cost efficiencies unmatched by any in-house or private cloud. Risk clusters, research clusters and reporting jobs, which can be farmed out in parallel or in bursts, are best suited for these tasks. These have provided the most common use-case for trading-related migrations into the public cloud.
There is a wide range of configurations for these grid computing or high-performance computing (HPC) clusters. Amazon Web Services, Microsoft Azure and Google Compute Engine all have their specific offerings. Just taking the example of AWS, there are a number of possible approaches to setting up clusters, from regular CPU or even GPU clusters which have been in AWS since 2010, to native services such as the Elastic MapReduce (EMR) and an overabundance of options such as Hadoop, Spark and Presto amongst others.
The very recent acquisition of NICE by AWS seems to point to AWS expanding its offerings, or at least upgrading its capabilities, which will certainly be interesting to watch.
Overwhelmed by all these choices? There is certainly a vast array of potential tools that could be used and it is often difficult to pick the right tools for the right job without prior experience. The beauty of the public cloud is the opportunity to experiment for very little cost, other than one's own effort. Gone are the days of having to spend considerable fortunes on licenses, consultants and hardware for a technology that may end up not being the right choice. Fail fast and fail cheaply is an easy principle to follow with the public cloud.
Some Technical Challenges
Networking features high on the list of potential bottlenecks. These include how the data is moved in and out of the cloud, what the required throughputs are and the potentially huge variations in costs. Additionally, once the data is inside the cloud and needs to be moved between services or regions, there are some considerations to be made around internal network speeds and overall architecture.
Looking at networking speeds only, there have recently been several benchmark studies which have highlighted different providers as being more suitable for different workloads and use cases. Zach Bjornson's analysis (see link on page opposite) found the lowest storage networking latency with Amazon and Azure, and the highest throughput on Google. Workloads with smaller files worked best on AWS and Azure, and those with larger files on Google.
Similarly, when it comes to computing, networking, databases and all the other services, the seemingly endless architectural permutations create the risk of either under-exploiting the resources, or hitting underlying hardware bottlenecks.
All of these aspects highlight the importance of understanding the different approaches and of benchmarking with your data and use cases. Clearly, understanding the workloads and the target cloud architectures helps to avoid many pitfalls of misapplied cloud technologies. This reiterates some of the main principles of the AWS Cloud Adoption Framework, around assessing correctly, starting small and growing methodologically.
Undeniably, public cloud usage is exploding. It has been growing for a long while in mainstream business, but now this is also happening in trading environments. Firms are either moving select workloads into the public cloud, or they are moving 'all-in' including execution. This is not public cloud moving into the limelight, we are actually past that point.
However, the more significant development is the rapid build up of an ecosystem of services. There is a vast choice of powerful and agile technology solutions and resources at commodity prices, which in turn drives further business innovation and agility.
The smaller, nimbler trading firms are the quickest adopters and this is presenting some real threats to the larger firms. Nowadays, you can see and work with brokers born in the cloud, order management delivered as SaaS, cloud hosted algos and backtesting, and entire middle/back office 'business as a service' running from AWS and Azure.
The next much bigger and complex challenge facing the regulators - and funds, investors, auditors etc. - is keeping up with these services moving up the value chain. The vast majority of funds will move to the public cloud over the next few years. But what are the interdependencies of all the public cloud solutions? Where are the single points of failure and possible contagion scenarios across all in-house and cloud deployments, SaaS providers and other outsourcing partners?
The cloud can be navigated with the right strategy and oversight, and is increasingly going to come up in due diligence from investors, auditors and regulators.
AWS Cloud Adoption Framework
Rob Alexander, CIO CapitalOne Interview
Citigroup Survey on IT Budgets in Hedge Funds
E&Y Hedge Fund and Investor Survey 2015
FINRA Case Study on Cost Savings
Amazon Acquires NICE
Cloud Storage Comparison by Zach Bjornson