Google Cloud Platform (GCP) is the most performant public cloud infrastructure-as-a-service (IaaS) provider for running online transactional processing (OLTP) workloads, but Amazon Web Services (AWS) remains the best value for the money.
That’s according to the 2021 Cloud Report from Cockroach Labs, the company behind the open source CockroachDB database, which recently raised a $160 million mega-round of funding.
“Declaring a winner was much harder to declare than in years past,” according to the authors of the report, as the gap on most metrics was razor thin.
The researchers benchmarked each of the major providers according to 54 machine configurations and nearly 1,000 benchmark test runs for CPU, network throughput, latency, storage read and write performance, and a homegrown derivative of the industry standard TPC-C benchmark. The full methodology is available on GitHub.
Taking the big three providers one by one, AWS offers the most cost efficiency and the best network latency, but lagged in storage I/O and single-core CPU performance. That being said, Amazon’s Graviton2 processor-based machines performed best for multi-core CPU performance. On cost, the c5a.4xlarge machine came in 12% lower cost than GCP alternatives and 35% lower than Microsoft Azure at $0.813 per tpm, according to the report.
Google Cloud had the best single-core CPU performance and delivered the most throughput of all providers, and was the second most cost-efficient option, even without an advanced disk option.
Microsoft Azure had comparable throughput to GCP and AWS, with the winning margin so thin as to be almost negligible for most workloads. Azure did pass AWS for storage I/O performance this year. At the top end, workloads running on Azure ultra disks led all cloud providers for I/O read and write IOPS and write latency. That performance comes at a cost, however, as Azure was the least cost efficient of all providers.