The challenge of Cloud PC performance testing

I concur with your technical arguments that demonstrate measuring the performance of Microsoft Cloud PC is an imprecise undertaking, particularly when relying on outdated benchmarking techniques and tools. This is also applicable to other categories of virtualized instances provided by Azure and other public cloud vendors, including virtual machines and containers.

One approach to benchmarking the performance accuracy of public cloud virtualized instances is to conduct a UTC accuracy performance test. Such test can be done without involving other aspects of a general compute platform, including application, networking and storage. Furthermore, to enable effective performance evaluation, in addition to new testing methodologies and tools, public cloud virtualization platforms like Hyper-V need to have complementary public hooks, such as TSC instruction or live migration events, that are available to these tools for correlating between the two realms.

The claims made by vendors in their sales and marketing efforts that their public cloud environments are suitable for financial workloads that require UTC accuracy within specific precision ranges are not credible unless they can provide proof of time accuracy performance of the virtualized instances running customer apps.

Regards,

--svb (b-svoba)