Neo4j has secured another $325m in a funding round and said it was ready to demo a distributed graph database with a trillion relationships, sharded across 1,000 servers, returning queries in a matter of milliseconds.
New investors DTCP – formerly Deutsche Telekom Capital Partners – and Lightrock join existing investors One Peak Partners, Creandum, and Greenbridge Partners.
The latest injection of funds – which is said to value Neo4j at $2bn – adds to earlier rounds including $10.6m in 2011, $11m in 2012, $20m in 2015, $36m in 2016, and $80m in 2018.
At the same time, Neo4j is set to show off a distributed version of its graph which it said would answer questions raised over its scalability, as reported by The Register in May.
Talking to El Reg, CEO Emil Eifrem said the developer conference this week would see him demonstrate an application Neo4j wrote for a social network graph with more members than Facebook, around 3 billion people, exhibiting a trillion relationships sharded across over 1,000 servers. Servers would be "all executing as a single graph database running complex, low-latency queries that return in 10s of milliseconds."
Eifrem said: "This is just above and beyond what in the graph technology has shown running live in the cloud in a distributed fashion ever before."
The demo is of Neo4j's new iteration, 4.3, running in the company's DBaaS, Aura. The latest release will also be available as an on-prem or cloud-hosted paid-for enterprise edition, as well as a free community edition.
- Jaguar Land Rover reaches for graph database in search of supply chain knowledge during chip shortage
- EDB tries to crowbar graph, JSON, and time-series data models into PostgreSQL – but can they pull it off?
- Hitting the mother-node: TigerGraph devours $105m funding in Series C round
- Time to haul DBaaS: Neo4j joins the fully managed club as doubts linger about its cloudiness
Eifrem said the perception that Neo4j was not distributed was a myth built around the community edition which is "not built for massively sharded, highly distributed, massive-scale type of things. It gets you started, and you can run it in production, and we have thousands of applications and 100,000s of developers running that. But when it comes at that the scale is really the Enterprise Edition," he said.
Meanwhile, the graph firm was also today introducing a free-tier version of its Aura DBaaS for developers "all over the world, so they can just sign up, and get up and running for free."
"Once you get beyond a certain scale, then at that point you're gonna have to start paying for it," he said.
Carl Olofson, IDC research vice president, said Neo4j was bolstering its credibility in terms of scalability, but also in terms of flexibility in deployment.
"Neo4j was designed as a deep traversals graph database. So the emphasis at the time was on a scale-up architecture, because of the nature of complex graphs: they work best if all the nodes of the graph are in one system because if you have to jump to another system, you introduce latency," he said.
Meanwhile, Neo4j rival TigerGraph has made much of its cloud-native approach.
"TigerGraph can be deployed on-prem too, but they're saying they were developed from the start with the cloud in mind. The architecture is meant to be cloud-friendly, but I think over time, Neo4j have figured that out," Olofson said. ®