ScyllaDB raises $43M to boost next generation of NoSQL database scalability

ScyllaDB has made a name for itself in recent years as a high-performance database used for some of the most demanding organizations on Earth. Among ScyllaDB’s notable users are social networking service Discord, travel site Expedia and media giant Comcast.

Today, ScyllaDB announced that it has raised $43 million in a Series C3 round of funding. This brings total funding for the open-source database vendor up to $100 million to date. The new funding round was led by Eight Roads Ventures and AB Private Credit Investors and included the participation of TLV partners, Magma Ventures and Qualcomm Ventures.

ScyllaDB is an open-source NoSQL database that was originally designed to be a drop-in replacement for the open-source Apache Cassandra database, with the promise of providing more scale and performance. The technology has expanded in recent years to also be a competitive replacement for the Amazon DynamoDB database. With the new funding, ScyllaDB is looking to build out new capabilities and also take aim at the MongoDB database.

“We raised the money because we could and not necessarily because it was a must,” Dor Lior, CEO of ScyllaDB, told TechForgePulse. “We’re trying hard to go after MongoDB as it’s a great target for us, especially in this market atmosphere where people try to be more efficient and reduce costs.”

ScyllaDB set its sights on scaling NoSQL database 

There are multiple ways that ScyllaDB can be deployed, with open source, enterprise and cloud database-as-a-service options.

The current release branch is ScyllaDB 5 which first debuted in July 2022 and has been steadily iterated on in the year since then, with incremental improvements to help provide faster database operations. While the ScyllaDB 5 database is his company’s present, Lior is very focused on what’s coming next with ScyllaDB 6 which is currently in development.

The big innovation that ScyllaDB 6 will introduce is a concept the company calls – tablets. The basic idea behind tablets is to have a new, more scalable, faster way to grow a database cluster, than with existing NoSQL approaches. According to Lior, tablets are designed to make it trivial to scale out the database across additional servers.

Lior explained that a common approach to database scalability today is to share data, which is all about having data broken up into smaller pieces that are distributed across multiple database nodes. The idea with a tablet is to take the database sharding approach to the next level. Lior said that a tablet can be a 10-gigabyte chunk of data that can be load-balanced across available computing capacity. The tablet promises that it can enable faster elasticity as a method for organizations to rapidly add or remove capacity for a given workload.

The faster elasticity will also be enabled by what Lior referred to as Raft consistent metadata. Raft is an open-source consensus protocol that is supported by ScyllaDB to help enable consistency of data across distributed clusters. With the Raft consistent metadata updates, Lior said that multiple database schema operations can occur in parallel while maintaining consistency. It also will allow for multiple topology changing operations like adding nodes to happen simultaneously, rather than just one at a time as in the current release.

Vectors are not a (current) target for ScyllaDB

While much of the database industry is increasingly chasing generative AI workloads, typically by supporting vector embeddings, that’s not a direction that ScyllaDB is taking.

Datastax, which is among the leading contributors to the Apache Cassandra database, recently added vector support to its commercial database and contributed code to enable vectors in the open-source project. MongoDB is now a competitive target for ScyllaDB and also supports vectors as it aims to support generative AI.

“Currently, we see lots of more traditional, non-generative AI usage with ScyllaDB, and there are plenty of them,” Lior said. “There’s just a ton of use cases without vector-based generative AI.”

Lior explained that with the ScyllaDB database architecture, one of the downsides is that implementing a vector search is harder than a traditional architecture for various performance reasons. He did note however that supporting vectors is on the ScyllaDB roadmap, though it is not a feature that will be released in the near term.

“If you’re not in the very large-scale database business, then vector search is something you add but it’s not a unique difference because everybody will have it,” Lior said. “We’re in a different ballgame and sometimes it’s good for us, sometimes it makes it very, very hard to implement things because we try to keep the level of consistency as high as we can.”