As our Chief Technology Officer, Peter Williams, explained in a recent presentation, the answer lies in tooling.
At a recent FISD Technology Forum in London, Peter Williams recounted the journey CJC have been on to address this streaming market data in the cloud question.
For clarity at this point, it’s important to note that by ‘market data’, we mean real time, streaming data for price discovery – a common enough use case in financial institutions across the globe.
Two years ago the original question had effectively morphed into a challenge, put to our engineering and innovation teams, to create a deployment and management tool which would allow us to define a system, as well as its ‘Day 2’ management and operation tasks, and deploy it to the public cloud automatically.
Peter went on to explain; “We wanted to start on familiar ground, and since we’ve been providing Thomson Reuters/Refinitiv with a managed service and extensive consultancy around TREP for the last fifteen years, it was logical that our first deployments were based on TREP components. We defined and deployed a DACS (Data Access Control System) with an ADS (Advanced Distribution Server) POP for streaming data access with permissioning.”
The 2 minute video below demonstrates how this was achieved (in about 12 minutes in elapsed time) using our Cloud Tools.
NB: Just as an aside, this release around our DACS in the Cloud solution may be of interest.
But the journey didn’t end there. We’ve now taken things a stage further by adding a range of other platforms and technologies that can be deployed, including BCC Group’s ONE Solution and Push Technology’s Diffusion Real-Time API Management platform. And as Peter elaborated, “it’s important to highlight that our Cloud Tools make it simple for us to link the systems we deploy. We can deploy multiple technologies and bring them together to create a more complex, but effective, market data delivery system.”
And still the journey goes on. We’ve been speaking to a number of capital markets participants, including a tier one investment bank, about how we would go about migrating TREP to the cloud at scale. What we’d like to achieve is an entire market data delivery stack, deployed and managed within the Google Cloud Platform, at the kind of scale suitable for a tier one bank.
So we engineered a horizontally scaling, streaming market data delivery system using TREP. And with our technology partners at Google Cloud, we have been carrying out initial tests on what we can achieve in terms of performance.
Peter explained that “our initial testing was to determine what could be achieved on a minimal cloud infrastructure footprint – since anyone could throw resource at a discrete problem. But we wanted to achieve maximum value and deliver repeatable real world scenarios – not to use impractical designs merely to secure impressive statistics.”
However, as the slide below indicates, we have witnessed some pretty good statistics anyway!
Our first step in the programme was to model our test on bare metal to establish a benchmark.
The headline is that, based on the results, there is very little difference between our cloud deployment and bare metal. It should also be borne in mind that whether an organization consumes two/three/four million records, this test relates to a single “slice” of a horizontally scaling infrastructure – so it can be deployed to scale up to meet specific requirements.
So in conclusion, if we consider the original question / challenge around how to simplify streaming market data in the cloud, our Cloud Tools deliver:
Fully automated deployments with ‘day 2’ management and administration
Enterprise performance while leveraging cloud benefits
Security by activity and by role – all fully audited
We’d be delighted discuss in more detail the work we’ve done to date, as well as how Cloud Tools may be able to help you to accelerate the deployment of complex market data workloads into the cloud. We’d be happy to model any specific use cases and can add new systems for deployment by the tool in as little as a day, depending on complexity.