Recently, a Boston-based consulting firm - Cloud Spectator - published a report benchmarking Public Cloud providers. The results were very interesting, and highlight the importance of considering cloud performance when planning a move to the Cloud; as the report argues, if you don't consider performance from the start, you may end up overspending to maintain sufficient application performance.
What got me thinking about performance were these two statistics from the report based on their tests among various Cloud Service Providers (CSPs):
- Looking solely at performance, virtual machines (VMs) differed by more than 3.5x across top CSPs.
- For block storage IOPS, that difference spread by 10x across the same CSPs.
Think about that: Performance varied by 350% (for VMs) up to 1000% between clouds provided by the various top CSPs. So, performance in the public cloud isn't consistent. At least according to the Cloud Spectator data.
One other interesting point from the study, for the CSPs that had the most stable performance on block storage IOPS, the report concluded that this "is likely due to performance throttling, which controls performance for each VM to sustain a steady user experience."
In other words, some CSPs reduce overall performance - make it worse - to keep it consistent in their own cloud.
Now, we knew some of this before this report was published, which led us to work with Verizon on Multi-Cloud Control (Verizon taps VMTurbo to develop Intelligent Cloud Control - also covered here in this CloudPro article), but I wasn't aware that the performance delta between clouds could be as much as 1000%.
There's a great deal more interesting data in the report, which you can download here: Top 10 Best Cloud Providers 2016 | Cloud Spectator
Let us know what your experience is with Public Cloud performance - does it feel consistent to you? Or are you seeing what this study did?