Businesses are gathering more and more information than they ever have before, whether it’s a manufacturer or a bank, and to Timescale CEO Ajay Kulkarni the existing database software soon won’t provide the horsepower they need.
That’s why he and CTO Mike Freedman started Timescale, a new open source time series database software. The idea is that businesses increasingly need ways to divine insights from the data they gather, whether that’s from sensors in a power plant to financial transactions — and they need to do it in real time without room for error. Timescale hopes to build out a new kind of database software that will feel familiar to developers but offer that kind of capability, and then look to build a business on top of that. To do that, Timescale has raised $12.4 million in a new round of financing led by Benchmark Capital, with NEA and Two Sigma Ventures participating, with Benchmark’s Peter Fenton joining the startup’s board of directors.
“Developers had two choices: SQL and relational databases, which are easy to use and reliable but hard to scale,” Kulkarni said. “They also had NoSQL, which scaled but [wasn’t so reliable]. When we looked at the nature of machine data we realized there’s a way to offer the best of both worlds. But also realized that machine data was largely time-series in nature. Every machine record is a measurement with a timestamp and some metadata. What we did was a little different and heretical, we built it on top of Postgres, and by doing that we inherited all the operational tools of Postgres, but we ended up making it scale for time-series data.”
Timescale sits on top of Postgres, one of the most popular open source database tools used by developers today. Kulkarni said Timescale is riding a resurgence in the usage of Postgres as businesses begin to collect more and more data, and that’s helped it attract a wide swath of developers looking to find a more efficient real-time way to access that data. Developers have downloaded Timescale more than 100,000 times, Kulkarni said.
Timescale is riding the idea that companies are doing an increasing amount of inserting information into their databases, which typically happen in order — hence making sense to order it in some kind of time series. The idea is that once the data is there, it’s rarely updated unless there’s an error, and it’s already naturally ordered, so it makes sense to partition the data in that way beyond just normal partitioning. When businesses query data, they might query it within the same time interval, and there are advantages of figuring out how to chunk that data and make it more readily accessible (like operating between memory and disk).
As developers add that information to their database, it’s getting automatically partitioned by time, and they access it like a normal table through a layer of abstraction that the startup is calling a “hypertable.” So the developer, for the most part, sees the kind of interface they would normally see, with Timescale operating behind the scenes to make things more efficient and faster.
“Even on disk, you’re typically querying data from the same time interval — maybe yesterday’s data,” Kulkarni said. “We make it easy to load this chunk in memory. We’ve designed the database in a way that’s optimized for this kind of data. You may end up with tens of thousands of partitions, and we’ll hide them from the user behind [the hypertable]. It gives the illusion of a continuous table across time and space. We optimize the query and surface that data. If you’re inserting, you just insert into the hypertable and we create new chunks.”
Like many startups spinning up on top of open source software, Kulkarni expects the startup to find some niches in last-mile tools that will help it turn into a fully functional business. One of the biggest challenges companies like these face is finding ways to capitalize on their expertise of the software to take over some of the most complex operations for businesses that need a more specialized kind of deployment, but want to hand off that responsibility to someone else with core knowledge of the tools.
That, of course, comes with its own challenges as many startups have seen. MongoDB went public last year on the strength of a business built on top of open source software, and since going public has had a very rocky start and is down 10% since going public. Timescale is also not the only one to tap this trend, with startups like InfluxData also raising capital. And, of course, Timescale has to convince developers and businesses that the tools are better and faster. But the proof is in the usage, Kulkarni said, as it starts to attract new businesses that already sought to deploy some kind of technology that’s faster than what’s currently available.
“Time series has become the fastest-growing category,” Kulkarni said. “When we started this company, we were scratching our own itch. But looking back in the last year we happen to be at the confluence of these two major trends — the resurgence of Postgres and emergence of time series data. For us, year one was really buidling the community and getting it to hold with some customers. Year two and year three are really scaling the business model and scaling your enterprise deployments.”