What's the future of the database? Jeff Naughton, Senior Vice President and Fellow at Celonis, has a few ideas.
Naughton oversees the development of process mining capabilities, data management and the in-memory SaolaDB at Celonis. Naughton's job is to hone and improve the data engine of Celonis products including the Execution Management System. He joins Vaishnavi Sashikanth, Chief Engineering Officer at Celonis, and Ariel Bardin, Chief Product Officer, to build out Celonis' global engineering team.
I caught up with Naughton, a former professor at Princeton University and University of Wisconsin as well as a Google executive, to talk about the future of databases and how process mining and execution management requires a new approach. Here are some of the highlights of our talk.
Combining actions, insights and databases. Naughton said he joined Celonis for two reasons. First, Celonis had a compelling business case, with customers buying the product and executing on the insights it delivers. The second thing was the technical side. "As a technologist, you are looking for projects and companies that are doing something that is somehow new and somehow has some depth to it, where you can really advance the 'state of the art,'" said Naughton. "You can solve important problems that people haven't even thought of before."
Naughton also said the combination of process mining, execution management, analytics and traditional databases is promising. "The combination is very fresh and exciting with lots of room for technical innovation," he said.
Unique data management approaches. Naughton said analyzing processes and solving execution management issues require a new mindset and database system. The approach to data management must be different due to the unique mix of sequences of events as well as more traditional data.
"Do you need to build new data management technology for this? I think the answer is yes," said Naughton. He added:
"If you do try to do this using a traditional database system, you find that the queries are extremely difficult to express. They're hard for people to understand, they're hard for the system to understand, so the performance is going to be very slow. I think it's to the point where it's just not practical, you need a system built with a query language and evaluation system that has the right fundamental concepts baked in all the way down, so things are easier to express, easier to evaluate."
Fortunately, Naughton isn't building this data management system from scratch. Celonis has PQL, its process query language, as well as SaolaDB. "We can keep developing it, pushing it farther, increasing its capabilities, for sure," said Naughton.
PQL improvements ahead. PQL has been adopted quickly by Celonis customers and Naughton said the plan is to continually improve it. Naughton explained:
"Like any language, we need to improve the experience of writing PQL, debugging and maintaining PQL and improving the development environment so it’s easier to write queries. People tend to write complicated queries and we could be more helpful there. In terms of the engine itself, like almost any engine, we can work on improving the scalability and performance. It's already very fast, but our customers are pushing us to handle more and more data and answer questions quicker. We can keep working on that and improve."
The future of databases. In the long run, Naughton said databases should largely disappear for users. "Eventually, the database system would get to the point where they just work, and the people building higher level applications and running businesses, using them, will just take for granted that they work. The system will provide value to users without them having to worry about the name or idiosyncrasies of the database management system or data analysis platform underlying things, just as cars provide value to drivers without them knowing the name or idiosyncrasies of their engines" said Naughton.
Special purpose databases. The number of types of databases has proliferated as developers use specific tools for various workloads. Naughton said the debate is ongoing in the database community. Ultimately, databases are going to have to include multiple capabilities. In some cases, there may be a broad database at the core and others featuring multiple databases.
Naughton said the combination of different perspectives can lead to exciting advances. "One of the things that I'm really excited about is the combination of multidimensional data analysis and process mining, combining those two together, to get the kind of insights that you couldn't get from either one by itself," he said.
Building a team. Naughton is looking to build out the data management team with people who are builders and focused on reliability, scalability, and performance.
"We also need people who like thinking about what's possible in the future, who would look at what we're doing and say, 'this is really cool, but we can build something new that's going to be even more powerful in the future,'" said Naughton.Rising Technical Star