On the latest episode of The EnterpriseReady podcast,Grant Miller sits down with Karthik Ranganathan, Founder and CTO of Yugabyte. A pioneer in the world of scalable databases, Karthik is one of the masterminds behind YugabyteDB, a high-performance distributed SQL database for global, internet-scale applications.
The discussion begins with Karthik’s time at Facebook, where he worked as part of the infrastructure team from 2007 to 2013. As a precursor to his work in building scalable databases, Karthik watched as the monthly active users of Facebook exploded from 58 million to over 1 billion in total.
At the time, open-source data infrastructure projects and public cloud were both in their infancy, so Karthik’s team went to work on figuring out how to scale their database strategy at Facebook to accommodate the massive influx of users. At the heart of this issue was the need for a scalable database for Facebook’s inbox search. Karthik explains, “What did this growth of users mean for the infrastructure? It meant we were getting 10-20 billion messages per day sent through chat. That meant we had a lot of messages flowing and needed to get store, as opposed to memory, so we really needed a scalable database.”
This scaling problem led directly to his work building the scalable OLTP (Online Transactional Processing) database Cassandra while still at Facebook. Cassandra has since gone on to become one of the most popular open-source databases available today, with implementations at Apple, eBay, Netflix, as well as many other industry leaders.
This experience at Facebook taught Karthik, and his fellow founders, invaluable lessons on what the world needed from the next cloud-native database. They decided that the most pressing features would have to be:
- Support for transactions and indexes
- High availability and resilience
- Horizontal scalability
- High performance
- Operational simplicity
- Geographic distribution of data
- All the features of PostgreSQL
After leaving Facebook, Karthik found himself at Nutanix, working on distributed storage and “learning the ropes of building an Enterprise company.” This is where Karthik says he received his core learnings of the contrast between a hyper-scale consumer company, like Facebook, and that of a fast-growing Enterprise company.
“At a hyper-scale consumer company, you could think of it as having 10-20 workloads, but at an ultra massive scale. So, each of these workloads would have billions of operations and petabytes of data, and crazy demand. In an Enterprise company, it’s a similar domain in the sense that it’s still distributed data, but the type of problems are so varied, even within a single customer. It’s not 10 or 20 workloads at billions of operations, it’s 10,000 workloads at 10’s of millions of operations.”
Still not content with the industry’s true lack of a “high-performance, distributed SQL database,” Karthik and his co-founders set out to create one themselves. They agreed on a customer-driven approach that would incorporate all of the features that they had dreamt of while still at Facebook. Just like that, YugabyteDB was born.
Created to stand on the 3 pillars of “high-performance,” “all of SQL,” and “scale and geodistribution,” Yugabyte was launched in 2016. The team quickly decided to launch as open-source software, citing the level of transparency that open source invites. Karthik also praises the benefits of open source as “accelerating your maturity level as a software, but also accelerating your feedback from your customers.”
The level of transparency that open-source software provides means that initial users can act as QA for software on a scope that an internal team would struggle to ever achieve. It also means that the level of feedback can exceed that of a closed company with hundreds of customers, as usage is high and open-source users and contributors actively seek out issues and report them accordingly.
Staying entirely open source with Yugabyte was out of the question, though. In order to achieve a sustainable business model, Yugabyte offers the use of YugabyteDB to anyone outside of a production environment but monetizes its use in large-scale environments. This means that they can still keep the high level of user feedback that open-source software encourages, but also ensure that Yugabyte can continue development with a long-term enterprise model.
Today, Yugabyte receives large-scale Enterprise adoption, with its largest customers each performing 10’s of billions of operations per day. Their open-source community boasts over 1,400 contributing members and shows no sign of slowing down.
Here at Replicated, we also strongly believe in the power of open-source software. Our core products KOTS, kurl, and Troubleshoot are all open source, and even EnterpriseReady, the site that we host this very podcast on, is open source as well.
To listen to this incredible episode in its entirety, please click here. For all episodes of The EnterpriseReady Podcast with Grant Miller, please head over to the EnterpriseReady Podcast page and start streaming today!