The new distributed graph architecture promises unified transactional and analytical processing, enabling enterprises to scale real-time decision-making for autonomous workflows.
Graph database provider Neo4j has launched a new distributed graph architecture, Infinigraph, that will combine both operational (OLTP) and analytical (OLAP) workloads across its databases to help enterprises adopt agent-based automation for analytics.
The new architecture, which is currently available as part of Neo4j’s Enterprise Edition and will soon be available in AuraDB, uses sharding that distributes the graph’s property data across different members of a cluster.
“Enterprises are increasingly moving toward HTAP (Hybrid Transactional and Analytical Processing) to unify OLTP (operational) and OLAP (analytical) data. This convergence is becoming critical for enabling agentic AI, which depends on real-time decision-making,” said Devin Pratt, research director at IDC.
“Recent moves in the industry, such as Databricks’ acquisition of Neon and Snowflake’s acquisition of Crunchy Data, highlight how most vendors are aligning with this broader trend,” Pratt added.
Enterprises are progressively looking to adopt agentic AI or workflows — systems that can operate autonomously with minimal human intervention — for strategic reasons, such as resource optimization, operation efficiency, and scalability.
When it comes to analytical systems, enterprises that choose not to unify OLAP and OLTP are adding to unnecessary expenditure as they have to maintain more than one system and develop complex ETL pipelines to do analysis on all types of data, said Robert Kramer, principal analyst at Moor Insights and Strategy.
“Unifying workloads offers a single, reliable source of truth, reduces infrastructure overhead, and makes it easier to tackle complex tasks like fraud detection or customer recommendations,” Kramer added.
Sharding is key for Infinigraph. But can it sustain performance?
Neo4j’s use of sharding, or splitting the data among multiple nodes, has been a common technique to achieve scalability in relational databases.
However, applying sharding to graph databases is challenging because it can split related data across nodes, hurting performance, ISG Software Research’s director David Menninger pointed out.
“As a result, graph databases have not been as scalable as relational databases,” Menninger added.
On the contrary, Neo4j claims that its Infinigraph-infused databases will offer high performance over workloads and scale to over 100TB horizontally with zero rewrites.
However, Moor Insights and Strategy’s Kramer did not sound too confident of Neo4j’s Infinigraph-infused database offerings, at least in terms of performance.
“The key questions are whether Infinigraph can sustain performance under heavy mixed workloads and how well it integrates with existing enterprise systems. Customers will need to validate responsiveness and scalability in practice,” Kramer said.
Further, IDC’s Pratt pointed out that Neo4j has historically faced scrutiny over horizontal scalability, while competitors like TigerGraph demonstrated stronger scale-out performance.
Neo4j has been facing increasing competition, ISG’s Menninger pointed out.
It competes with Amazon Neptune, Azure CosmosDB, TigerGraph, Aerospike, SpannerGraph, and OrientDB, among others, in the graph database space, and nearly all data platform providers, as each of them offers graph database functionality, Menninger said.
“The real question is whether graph database specialists can offer a product that is differentiated enough to justify purchasing an additional data platform,” Menninger added.
