Highlights:

  • Aerospike’s latest solution is built on the Apache TinkerPop open-source graph framework, which also serves as the foundation for Amazon Web Services Inc.’s Neptune graph database.
  • Aerospike claims that the underlying architecture of its new graph solution enables it to outperform other offerings in terms of both performance and scalability, allowing it to execute multihop graph queries across trillions of vertices and edges in just milliseconds.

Aerospike Inc., the producer of a highly scalable NoSQL database, is venturing into the graph database market with a new solution that it asserts can surpass the capabilities of leading market players like Neo4j Inc. and TigerGraph Inc. regarding performance and scalability.

Aerospike’s latest solution is built on the Apache TinkerPop open-source graph framework, which also serves as the foundation for Amazon Web Services Inc.’s Neptune graph database. Unlike Neo4J’s Cypher or TigerGraph’s Graph SQL, the solution supports the Gremlin query language. The company has plans to add support for the open-source implementation of Cypher in the first quarter of next year. In addition, the solution utilizes Aerospike’s key-value storage as the foundation for a graph data model instead of employing native graph storage.

While Gremlin is renowned for its flexibility, portability, and scalability, it has a steep learning curve and a limited range of third-party tools. Nevertheless, Aerospike claims that the underlying architecture of its new graph solution enables it to outperform other offerings in terms of both performance and scalability, enabling it to execute multihop graph queries across trillions of vertices and edges in just milliseconds.

According to the company, benchmarks demonstrate that the solution can achieve a throughput of over 100,000 queries per second, with a latency of under-five milliseconds, even when operating on just one-tenth of the hardware infrastructure required by its competitors.

Hybrid Memory Architecture

Aerospike’s distinguishing feature is its “hybrid memory architecture,” which allows each node or server to consider a set of solid-state storage devices as an extension of memory. The solution automatically stores frequently accessed data in memory and transfers data between memory and disk without needing access to the underlying file or operating systems.

Instead of caching, the approach relies on precisely calibrated clustering algorithms and network pathways.

The process of roster-based clustering involves grouping data points based on the idea of rosters. This technique is commonly utilized in data mining, pattern recognition, and machine learning to identify inherent clusters or groupings within a dataset.

Horizontal Scale

Aerospike is engineered to scale horizontally across numerous nodes, with the ability to manage data distributed across a group of machines for optimal throughput and minimal latency. The solution also offers configurable durability options that allow users to regulate the degree of data persistence and replication.

Initially, the company intends to concentrate on its core markets, which are advertising technology and financial services, where it has a strong foothold.

Aerospike claims that its graph product can independently expand computing and storage, allowing users to pay for only the required resources. Although the company has not disclosed pricing details, Hensarling stated that the storage volume and the number of virtual CPUs utilized would determine it. Additionally, the company will provide monthly pricing options to accommodate customers and businesses, such as retail and media, that encounter surges in demand.