Goal is to monitor the real time utilization and dynamically/automatically up- or downscale the VM resources for Dataminer, Elasticsearch and Cassandra using IBM Turbonomic.
Looking for recommendations for such dynamic allocation changes.
Currently, there is no recommendation on configuring the virtualization layer, which virtualization SW, or how to scale up/down. Every machine (virtualized or physical) is expected to meet at least the minimum requirements. When working with virtualized environments that share resources it is not easy to verify these requirements (running a benchmark on one machine could be fine, but running the benchmark on all of them simultaneously could be problematic). We have seen that configuration on the virtualization layer can make enormous improvements (e.g. some virtualization layers disable CPU caching).