Third time was the charm for SingleStore in the cloud, CEO says
Apache Iceberg support makes it a good option for a transactional layer over data lakes, he tells The Register
SingleStore, the database that promises analytics and transactions on a single system, took three attempts to get its technology working in the cloud, CEO Raj Verma admitted to The Register.
The vendor, which counts Uber, Kellogg's, and GE among its customers, "got it right the third time," the chief said.
SingleStore launched its cloud database, MemSQL Helios, in September 2019, but the company did not have a substantial cloud business until two-and-a-half years ago. Rival database vendors like MariaDB and DataStax, for example, have had database services on the market for around four years, while MongoDB launched https://www.theregister.com/2016/06/28/mongodb_launches_atlas_to_manage_deployments_taking_the_ops_out_of_devops/, Atlas eight years ago.
Part of the reason was that these companies enjoyed an open source code base and benefited from a "bottom-up" approach, Verma said. He added that putting a database in the cloud was not as simple as moving its infrastructure support.
"It is hard," he said. "You can move your technology to the cloud [but] the seminal challenge is how do you go from selling software to providing a service? How do you make money selling software versus making money running your software? When you're doing the latter, the experience that someone gets of your service has to be seamless. It has to be elastic. In the developer experience, you have to be able to incorporate a lot of things that cloud makes possible that it was just not possible to do [on-prem]. Those were the areas that we found very hard. The moving of the technology stack actually isn't that hard."
Verma added that high availability and disaster recovery were also challenging to recreate in the cloud according to customer requirements.
"They are an order of magnitude higher than what you would need on-prem. The fact that you need uptimes which are the best-in-class, or five nines. Those are just hard things to build," Verma said.
- Oracle's database family gets trendy AI makeover
- Google advances with vector search in MySQL, leapfrogging Oracle in LLM support
- CockroachDB tempts legacy databases to crawl into the cloud age
- Couchbase takes fight to MongoDB with columnar side store upgrade
The switch was enabled by a relatively low cash burn rate and revenue per employee close to $300,000 putting it on a par with Snowflake and Netflix. By next year the company expects 70 percent of its revenue to come from the cloud, while only 10 percent of its cloud revenue will come from customers moving from on-prem installations.
"We don't need to raise additional capital to run our business. We'll be self-sufficient in about 18 months," Verma predicted.
Last month, SingleStore added bi-directional integration with Apache Iceberg to help users access data more easily since the open table format now has broad support from vendors including Cloudera, Snowflake, AWS, and IBM. Databricks also bought Tabular, founded by Iceberg's original developers, in June for a tidy sum of $1 billion.
Verma said that users of data lakes and data warehouses were looking for an operational layer on top of their analytics systems. With its adoption of Iceberg, SingleStore would be able to fulfil that role, he added.
"Being the performance, operational layer on top of data repository like Iceberg is a competitive differentiator and a huge advantage for us, at least for the time being," he said. ®