Cloud performance

At times during testing of our ACI Windows prototypes, we have experienced some less that ideal performance results. This has included the time to complete tasks such as log-in, application launch, and data upload/download. In and of itself, this is not a concern and may only require we utilize larger server instances with more available CPU and memory. However what was unexpected was the inconsistency of the results. Two identical instances (cpu, memory, storage, OS, patch level, applications, policies, etc) should provide very similar (good or bad) test results.

Webmetrics financed Bitcurrent to research a month long study of the performance of five of the major public cloud providers. They recently published their results.

While it doesn't explain the inconsistent results, it does help to understand the strengths and weaknesses of each vendor's approach and architecture. This kind of information, when considered with cost and management capabilities, will help to determine the right vendor for a given use case. As to the consistency concern, we'll need to begin to chart activity to assess what is the most likely source of the issue to determine the best approach for a solution.