Hi Kevin / Andrew / (& all),
It might help if you can share some use cases of the sorts of scenarios where MAT is used in your environments, and environments you have seen other users using. (All feedback is welcome!). I think it would be a valuable part of the discussion, so that I can understand how unique our requirements are and what considerations I should be putting in for any potential future work.
The specific use case and goal I have may be unique to our circumstances. The bottom line for us here is that CPU and RAM is relatively plentiful, whilst engineer time is much more valuable.
- It is no secret at Netflix that we run on AWS and have a significant Java deployment for applications. We have large Cassandra and ElasticSearch deployments.
- Many of our applications run on instance sizes and Java heaps that dwarf what would be possible to reasonably download and process on developer machines. It is very common for application heaps to be in the 50-100GB range.
- This means that teams have workflows involving run Eclipse analysis of large heaps on cloud machines anyway using remote desktops. In this scenario, memory is not as constrained.
- Large memory machines come with lots of cores. Whilst we have these cores allocated, and engineers waiting on results, we would like to find a way to use them.
Let me know your thoughts and observations.
Thanks
Jason