Nylas is a pioneer and leading provider of productivity infrastructure solutions for modern software. Over 50,000 developers worldwide use the Nylas platform to quickly and securely build productivity features into their applications. With Nylas, developers get unprecedented access to rich communications data from their end-users, pre-built workflows that automate everyday tasks, embeddable UI/UX components for fast front-end development, and comprehensive security features - all delivered via a suite of powerful APIs that make integration easy.
As a leading business automation infrastructure solution, improving API performance and providing the best quality of service is a crucial KPI for Nylas. Nylas handles billions of API requests per day on GCP and has a constantly growing environment. As Nylas’ user base is growing by 2-4x year over year, they needed a way to dramatically improve performance quickly with no code changes or R&D efforts.
Their API platform provides real-time bi-directional sync with user’s email , calendar and contact data, with an intelligent automation process built on top. Nylas’ Sync service is the key component of the platform. When an account is connected to Nylas, which happens thousands of times per second, the sync engine starts pulling in email messages for the account, pulling new message arrivals in parallel, and prioritizing the most recent ones to achieve a better user experience.
As a core piece of Nylas’ API puzzle, the Sync service is required to operate quickly and efficiently. That means that it is extremely sensitive to performance and requires an aggressive SLA of “time to sync” or latency.
The Sync service runs on GCP and includes ~200 running instances and over 4K cores.
By implementing Intel Granulate on a few Sync machines, Nylas dramatically improved the service performance, cutting 35% off the cost of their compute spend and increasing the number of sync processes executed per machine by a large margin. This was a convincing demonstration of the potential cost reduction that can be achieved using Intel Granulate’s application-driven resource management optimization.
After learning the Sync service resource usage patterns and data flow, Intel Granulate began making real-time decisions at the runtime level to prioritize resource allocation mechanisms and queues to achieve optimized performance. Immediately after activation, Nylas experienced a 58% reduction in latency, allowing the same cluster to handle 2X more syncing requests on the cluster level while reducing CPU utilization by more than 10%.
These results allowed Nylas’ cluster to handle the sync queue much faster, handling more requests with fewer machines and reducing CPU utilization per machine. This led to substantial cost reduction, alongside improved performance on the cluster level.
*Intel does not control or audit third-party data. You should consult other sources to evaluate accuracy
Nylas chose Intel Granulate's real-time continuous optimization solution to optimize GCP workload performance and costs. Within 5 days and without any code changes or R&D efforts, Nylas achieved over 35% cost reduction leveraging 58% faster processing time while handling 35% more throughput.
