Hasnain says:

“After rolling out this change, we saw a significant improvement in query runtimes for bigger, more complex environments, including ones with more than 2,000 services and on the order of 100,000 events per second flowing into our system. Median query duration in situations like these started at around 20s (a customer experience we were certainly not proud of!) and decreased to around 0.2 seconds. That’s two orders of magnitude and well within our internal query performance SLOs.

Additionally, this layer of abstraction affords us a lot more freedom to add more features to Honeycomb without worrying as much about the implications of our data model on query performance, since the two concepts are now (mostly) independent concerns. And finally, we’re brimming with ideas about how to leverage this new model to continue to improve query performance!”

Posted on 2024-06-17T00:42:26+0000