This is guest post by Baron Schwartz, Founder & CEO of VividCortex, the first unified suite of performance management tools specifically designed for today's large-scale, polyglot persistence tier.
VividCortex is a cloud-hosted SaaS platform for database performance management. Our customers install agents that measure the work their servers perform (queries, processes, etc) and generate metrics and events from that at high frequency. The agents send the resulting data to our APIs, where we host our analysis backend. The backend system is a collection of databases, internal services (quasi-microservices), and web-facing APIs. These APIs also power our AngularJS frontend application.
We deal with a lot of data. We ingest metrics and events at high speed. We also perform analytics that touch large amounts of data interactively. We are not unique and I don't want to imply we are somehow impressive in the scheme of things. We don't yet operate at "web scale." Nevertheless, our workload has some relatively unusual characteristics, and we've been able to scale as far as we have, while remaining pretty efficient in terms of cost and infrastructure. And my career in consulting has taught me that building systems like this is usually a challenge for a company (as it has been for us). Our story might be useful to others. For that reason I will go into unnecessary detail on specific parts of our workload and the challenges it brings.