Speedscale is a Y-Combinator backed startup that helps Kubernetes engineering teams build more resilient and performant containerized apps.
Using their tool, engineers can simulate production conditions, generate load or mock 3rd party backends modeled after real traffic patterns. Speedscale’s differentiator vs other tools is that it uses agents/sidecars to record and playback sanitized traffic you see in prod.
The Challenge: Unpredictable costs, non-intuitive queries
LeRay’s team already uses a major monitoring solution, but found they had to make major compromises to get their full range of data loaded. Even then they found running queries on older data was both expensive and tricky to work with on more longitudinal data.
Even more common queries often had too much of a learning curve. Engineers usually interact with the monitoring system in a time-critical crisis. They found their system’s query language unintuitive. Their specific needs were too unique to find a copy-and-paste solution on Stack Overflow or Reddit. Racing the clock to figure out a domain-specific language frustrated even senior engineers.
Surprise costs were also a problem. More than once an engineer just trying to do their work ingested a large volume of log data that created a monthly billing spike. It might be 10 times the usual. Sometimes a vendor will forgive such a spike, but sometimes not. LeRay, with long experience in DevOps and observability, didn’t want the team to be constantly pre-calculating the cost of their every move.
The Solution: Axiom pricing, plus its search box
Axiom’s ingest-based pricing proved to be “stupendously cheap” compared to alternatives, removing the fear of surprise bills that hung over monitoring and troubleshooting. That alone was reason enough to give it a try. But any worries that Axiom is a cut-rate Monitor Lite have proven 180 degrees wrong. LeRay especially praises its search box, where the Axiom Processing Language (APL) proved more intuitive than alternatives, and regex support lets even new hires figure out how to run queries quickly without training or research. With older logs, Axiom has proven both affordable to ingest them and adept at searching the longitudinal data on which engineers had struggled with other tools.
Beyond its operational value, for analytics Axiom is now the only tool they use. “We actually use Axiom as a business intelligence tool to make product management decisions,” LeRay says. Speedscale’s marketing team has created business analytics dashboards in the same tool the engineers use to look at operational data. There’s no need to ingest data into two separate systems, or try to guess which team will want which data. Engineering and marketing are familiar with each other’s tools and work from the same data set.
That’s the leap forward that Axiom’s founders had in mind: Power, ease of use, and cost don’t need to be a zero-sum equation. With a fresh start and new ideas, you can improve all three.