OtterTune raises $12M to automate database maintenance

The workload for database administrators may be getting easier – at least, if OtterTune has its way. This week, the company announced a new $12 million series A funding round to continue to expand and build out its tool that automates much of the work of tuning MySQL or PostgreSQL databases. 

OtterTune likes to say that its software will adjust more than 100 different “knobs” or parameters for these popular open-source databases. It uses artificial intelligence algorithms to watch the performance of the database and then suggest better values for the parameters or “knobs” to speed up performance. In one set of benchmarks, the company suggests that the database may run more than four times faster than the stock settings. 

The company is a spinoff from Carnegie Mellon’s School of Computer Science, where Andy Pavlo, the cofounder and CEO of OtterTune, is a faculty member. After he and a student tested the approach in the lab, their write-up caught the attention of Amazon, which used it in a blog post to illustrate the power of artificial intelligence tools. 

“We just started getting flooded with emails from people saying, ‘We have the exact problem. We’ll give you money to fly a student out and set it up for us in our databases,’” said Pavlo. “It’s obviously happened often enough and we said, ‘Okay, we clearly have a signal here.’”

OtterTune’s devops benefits

OtterTune adjusts control parameters that govern how the database schedules some of the maintenance tasks it does to rebalance indices and maintain the cache. In the past, the developers or database administrators would have the responsibility to adjust configuration parameters like the buffer pool size, the effective cache size, the background writer delay or the checkpoint completion target. Now, OtterTune watches the database and then generates a new set of more than 100 parameters. The devops team can either install them manually, tweak them further, or let OtterTune configure the database automatically. 

One of the main targets will be smaller companies that can’t afford a dedicated database administrator. OtterTune will charge by the database that they monitor. New customers can try a free sample.

“When I do these calls, customers actually apologize for not knowing what is inside the databases.” explained Pavlo. “And it’s because they’re a devop person and they got into this role of management because they drew the short straw.”

The artificial intelligence routines from OtterTune run independently and tap into the metadata that each database uses to track performance. The system does not have access to the actual data stored inside, something that allows them to avoid any personal information. 

“To be honest, the hard part is actually not the machine learning, right?” said Pavlo. “It’s really the infrastructure around working with the databases, collecting the data and then deriving the signal from live databases to determine whether you’re making things better. That’s the hardest part we’ve been facing.” 

The database competition

Other database companies are also chasing performance but in different ways. Oracle, for instance, likes to sell a new version of its flagship database as “autonomous.” That is, it also has similar internal mechanisms for adjusting behavior to optimize performance. 

Oracle also owns the original branch of MySQL and has been adding proprietary features that speed up some tasks. Their HeatWave version has a more sophisticated “in memory queue accelerator” and Oracle estimates that its hosted versions can run 6.5 times faster than Amazon Redshift and in some cases as much as 1400 times faster than Amazon Aurora.  

A number of startups are also building out products on the backs of MySQL and PostgreSQL and designing them to scale quickly and absorb extremely large data volumes. Companies like Fly.io, Google, Yugabyte and PlanetScale are just some that want to tap the popularity of the open-source standards while simplifying the path to working with big datasets. 

Some of these companies are rewriting the guts of the database and adding a new storage engine. Google’s new AlloyDB, for example, promises much faster queries for analytical workloads thanks to a new column store. 

OtterTune is not following that path. It works immediately with stock versions running on services like AWS’s Aurora or RDS. While Amazon tackles many of the challenges of launching and configuring a server, their services don’t provide the low-level tweaking that OtterTune is providing.

Tools for the future 

In the future, OtterTune wants to expand its offering by tackling more than just speeding up the query processing. It aims to watch other important factors like index creation and updating. The company also want to move into other clouds and also support database users who are running on their own hardware. 

“It’s not just going to be all about pure performance or pure cost reduction because that’s a moving target.” explained Pavlo. “We want to start adding features so we’re selling peace of mind. We can tell you that your database is running with all the best practices.”

“We’ve been running OtterTune for several months and it’s been a game changer for us,” said Daniel Rodgers-Pryor, CTO of educational technology platform developer Stile Education. “OtterTune keeps our databases running smoothly and it has removed the risk and research time associated with manual tuning. With OtterTune in place, our team can focus more on new development and be much less distracted by database administration.”

“OtterTune’s founders are multidisciplinary experts at the intersection of the database field and machine learning, as evidenced by their cutting-edge research while at Carnegie Mellon University,” said Nick Washburn, senior managing director at Intel Capital, who will also be joining the company’s board. “Databases are the bedrock of all applications and OtterTune is accelerating the journey for companies of all sizes to autonomously optimize this critical component of their tech stack, driving performance, managing costs and ultimately ensuring reliability.”

The new funding round is led by Intel Capital and Race Capital, with participation from Accel.