When you are creating a cloud storage solution such as on AWS, there are multiple parameters to consider. Not only do you need to know how much storage you need, but you need to know how the IOPs requirements and how bursty the workload is.
Before you can even begin to architect a cloud storage solution, you need to find out what your application is really doing. This includes understanding application dependencies and I/O patterns. The way an application has been deployed can dramatically change the I/O patterns so just because an application and its workflow have been characterised in one environment, that doesn’t mean that they won’t look different in another.
Breeze in action
Gaurav Kaul, senior consultant HPC at AWS Professional Services, has been using Breeze to profile I/O and architect the cloud solution for OpenFoam for a UK sporting body.
OpenFoam is widely used in the HPC industry, but is known to have problems with speed. Read our blog on how we profiled OpenFoam to see how performance could be improved.
After Gaurav traced the application using Breeze, this is what he discovered:
“For larger cell models, the small I/O and internode communication increases by quite a lot. This requires keeping track of various metrics such as IOPs, node to node latency and communication size etc. All this requires different tools which can be a laborious process.
“Using Breeze, we can get all this information easily without much overhead and explained in an intuitive manner.”
For more on how Breeze can help you to develop your cloud strategy, get in touch.