Performance tuning on Google Compute Engine

This page describes performance tuning for Aerospike Database on Google Compute Engine, including VM sizing, storage selection, network configuration, and benchmarking. Avoid applying CPU-pinning recipes without benchmarking them on the selected machine family and kernel.

VM sizing

Select a machine type with enough memory for the primary index, enough CPU for your target throughput, and enough network bandwidth for client and cluster traffic. See Capacity Planning for Google Compute Engine and capacity planning.

Storage

For persistent namespaces, validate whether the workload needs Local SSD, Persistent Disk, or shadow device configuration. See Recommendations for Google Compute Engine and namespace storage configuration.

Network

Use internal IP addresses for node-to-node traffic and restrict Aerospike ports to trusted clients, tools, and cluster nodes. See Network configuration.

Benchmark

Benchmark the cluster with the same object size, read/write mix, concurrency, and access patterns you expect in production. See Benchmarking.