Report finds SQL Server on Azure is Faster Than AWS

A new research report has shown that SQL Server runs faster in Micrososft’s Azure cloud than it does on Amazon Web Services (AWS).
GigaOm, a research firm sponsored by Microsoft, compared throughput performance of SQL Server on Azure Virtual machines and SQL Server in AWS EC2. The results showed that the former performed significantly better.
“Azure emerged the clear leader across both Windows & Linux for mission-critical workloads. It was up to 3.4x faster and up to 87 per cent less expensive than AWS EC2,” commented Microsoft in a blog post Dec. 2. Microsoft chose the competitors, the test and the Microsoft configuration for this sponsored report. The GigaOm Transactional Field Test was based on the TPC Benchmark E (TPC–E) industry standard.
After conducting a number test, GigaOm concluded that Microsoft SQL Server on Microsoft Azure Virtual Machines showed 3.4x higher performance on Windows than Microsoft SQL Server (AWS) Elastic Compute (EC2). When tested on Linux Server OS, Microsoft SQL Server on Microsoft Azure virtual Machines performed 3x better than AWS. SQL Server on Microsoft Azure Virtual Machines (VM), had up to 86.8 per cent better price-performance compared to AWS License Mobility for three year compute commitment and up to 32.2 per cent better price-performance compared to the high-speed disks AWS io1 or Azure Ultra Disk.
[Click on the image to see a larger view.] These images are GigaOm comparisons of performance and price-performance. Performance is measured in throughput (transactions per seconds, tps); higher performance means better. The price-performance metric measures three-year pricing divided with throughput (transactions/second, tps); lower price-performance is better. (source: Microsoft). Microsoft commented on the tests and stated that Azure BlobCache, which offers free reads, is a key reason why Azure’s price-performance is better than AWS. This allows customers to save significant amounts of money, as most online transaction processing (OLTP), workloads come with a ten percent read-to–write ratio.