Born in the cloud
As a technology start-up, Accelex was “born in the cloud”. From inception, our engineering, data science and infrastructure teams have focused on building and developing a modern, state-of-the-art, cloud-native technology stack.
Our data acquisition and analytics platform has been built using a sophisticated, containerized microservices architecture deployed in the cloud using Kubernetes. Kubernetes has become the leading container orchestration tool, giving incredible flexibility and control to developers. It is one of the fastest growing projects in the history of open-source software, successfully abstracting infrastructure and environment configurations from developers.
Kubernetes has allowed Accelex to develop services that elastically scale up and down on demand, can perform rolling upgrades with zero downtime for clients and which are seamlessly portable across different cloud providers. Such capabilities previously required large infrastructure, operations and sys-admin teams, inaccessible to the agile start-up.
Challenges of scale
Through 2021, the demands on Accelex’s infrastructure have significantly increased. Accelex has added clients around the world: in Europe, North America and our first client in Asia. Notably, those clients include large asset servicers, processing hundreds of thousands of documents through the Accelex platform on behalf of asset owners. Our platform regularly handles thousands of simultaneous document processing requests across multiple geographical regions and environments.
In addition to this, Accelex has added support for additional document types: capital account statements and cashflow notices can now be processed in the Accelex platform. Therefore, in addition to adding new clients, existing clients are processing more documents through Accelex.
Finally, the sophistication of Accelex’s engineering and data science stack continues to increase, placing additional demands on our underlying infrastructure.
Although Kubernetes provides us with tools to address these challenges, scaling a Kubernetes cluster up or down can take minutes to complete. As the cluster scales, Kubernetes re-schedules services across hardware nodes, which can delay ongoing processing requests. Maintaining a large steady-state cluster to give optimum performance is expensive, so inevitably compromises need to be made between cluster size, scaling speed and throughput.
These considerations led us to explore alternative deployment models for our platform.
Serverless architecture is a way to develop applications without having to manage the underlying infrastructure. In this model, the cloud provider allocates compute resources on demand. This allows services to scale from zero to thousands of replicas in seconds, enabling enormous throughput at low cost. For an event-driven workload, such as Accelex’s data acquisition pipeline, this has significant benefits.
Recent advances in serverless offerings support containerized runtimes, with generous CPU and memory limits able to support sophisticated machine learning models, such as those which power Accelex’s data acquisition pipeline. Serverless support for containerized runtimes has also allowed Accelex to maintain backwards-compatibility with the Kubernetes deployment model, giving us the flexibility for multi-cloud or hybrid public-private cloud deployments in the future.
Benefits for clients
Accelex’s transition to a serverless architecture means clients should enjoy significantly improved document processing times. This will be particularly apparent when processing long documents or large batches of documents in parallel. The power of near-infinitely scalable cloud computing combined with our industry-leading workflow tools, ensure the Accelex platform lets you take control of your alternative investments data.
Written by Chief Scientist Phillip Mortimer. Connect via LinkedIn.