A group of former Facebook and Dropbox engineers is developing a service for debugging complex systems and answering ad hoc questions in real time.
Honeycomb, currently in an open beta cycle, is a SaaS platform that reduces MTTR (mean time to repair) for outages and degraded services, identifies bugs and performance regressions, isolates contributing factors to failures, and reproduces user bug reports.
The collective debugging skills of teams would be captured and preserved, according to the project website. Rather than relying on a dashboard, Honeycomb is for interactive debugging.
"Honeycomb is event-driven, not time series-based," said Charity Majors, CEO, co-founder, and engineer at Honeycomb. "This means it doesn't roll up values into time 'buckets' that get less and less accurate or granular over time. You always have access to the raw data, and you can issue arbitrarily complex queries against your data in real time, refining them as you go."
Looking to bring the business intelligence model to systems and software, the project moves away from a model where developers try to predict ways systems can fail. "Newer systems are too complex and fluid for you to predict all the ways they can go wrong anyway," Majors said.
Honeycomb will be a fee-based service, with connectors released via open source for security and transparency. Parts of the project are based on founders' experiences at Facebook. "We developed our own columnar store, inspired by Facebook's Scuba, because there isn't any other database that provides the kind of flexible schemas and read time aggregation that we depend on," said Majors.