Bylined article by Stephane Dubois, Xignite CEO and Founder
Published in TABB Forum (requires free subscription to access)
Excerpt of article:
To deploy market data in the cloud, the best approach is a hybrid one. Rather than building a whole new market data technology stack in the cloud, as some very large firms have chosen to do, which would require significant maintenance of that technology stack, firms can choose to deploy battle-tested microservices. Given the nature of market data distribution, microservices have the potential to turn our industry on its head.
The primary reason why microservices are going to be so impactful on the world of market data is the low-cost scalability they afford. Capital markets tend to be peaks and lows operations. Even for global firms operating around the clock, the market open and market close times in New York, London or Tokyo generate huge peaks in inbound messages from the markets and demand for data from users and systems (terminals, trading systems, investors, etc). With a monolithic architecture, there is zero elasticity to the infrastructure, so you must provision it for the peaks. If not, delays occur – and market data delays can cost millions of dollars. Market data volumes during peak activity can be massive. Provisioning and maintaining an on-premise infrastructure for peak activities therefore is very expensive and it is not uncommon for the peaks to push the limits of what that legacy technology was designed for. In other words, legacy market data technology is not very horizontally scalable. It cannot be easily distributed and it can only scale slowly.
In contrast, microservices are designed for infinite horizontal scalability and quick elasticity. You can more effectively scale for growth and peak in users and transaction volumes, since the infrastructure can be easily horizontally replicated. This means that if you design your services properly (for example using container technology allowing for easy horizontal scaling coupled with appropriate orchestration management tools), you can bring up the capacity you need when you need it, and you can shut it down when you don’t. As you result, you only pay for the capacity you use, not the one you might need. If a pandemic is bringing huge volumes to the market, you just crank that capacity a bit more. Additionally, if your microservice is granular enough, you can truly fine-tune the amount of capacity you bring online, further lowering your costs.