Benchmark
Petabyte-scale performance on 20 nodes in the cloud
What you'll learn
Managing petabytes of operational data in real time demands more than raw capacity. IT leaders face relentless pressure to meet sub-millisecond SLAs, control TCO, and maintain five-nines availability while server footprints and costs spiral out of control. Most NoSQL and relational platforms buckle at petabyte scale, forcing organizations into sprawling infrastructures that drain engineering resources and block innovation.
Developed in collaboration with Intel and AWS, this benchmark shows Aerospike delivering 4 to 5 million transactions per second with sub-millisecond latencies on just 20 nodes managing 1 PB of unique user data. A comparable Cassandra deployment would require up to 792 nodes and $15.29 million over three years, versus Aerospike's $4.17 million.
Key highlights
Aerospike delivered 5 million TPS for read-only workloads and 3.7 million TPS for 80/20 mixed workloads, with sub-millisecond server-side latencies in nearly all cases.
A 20-node Aerospike cluster manages 1 PB of unique user data, compared to 514 nodes required by Apache Cassandra for an equivalent petabyte-scale workload.
Three-year TCO for Aerospike is $4.17 million versus $15.29 million for Cassandra, saving organizations more than $11 million on the same petabyte-scale production workload.
Reduce server footprint up to 80% and save $1 to $10 million per application while acting instantly across billions of transactions with predictable, sub-millisecond performance.
Deploy on premises, in multi-cloud, or hybrid configurations with five-nines uptime, strong data consistency, and self-healing cluster management requiring no downtime.