Dan Kleiman, Klaviyo
Last year we launched a Query Service to provide real-time analytics for a mix of workloads that includes public APIs, dashboards in our app, and report generation.
Consolidating these use cases to a single service was a huge infrastructure cost and complexity win, but we soon started experiencing intermittent waves of timeouts, impacting all our callers at once.
We thought we had provisioned the service with enough capacity, so why were we hitting congestion?
Inspired by Jon Moore's 2017 Strangeloop talk "Stop Rate Limiting! Capacity Management Done Right", Netflix's blog post Performance Under Load, and their Concurrency Limits library, this talk will share how we iteratively applied the principles of concurrency control to improve the user experience of our Query Service.
Dan Kleiman, Klaviyo
Dan is a software engineer at Klaviyo—a marketing tech company that powers email, sms, and ecommerce integrations—where he works on query services for Klaviyo's data platform.
Open Access Media
USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and open to everyone. Support USENIX and our commitment to Open Access.
author = {Dan Kleiman},
title = {Adaptive Concurrency Control for Mixed Analytical Workloads},
year = {2023},
address = {Santa Clara, CA},
publisher = {USENIX Association},
month = mar
}