For deadline driven delivery, I’m curious if others (perhaps Lightning) are using the elastic scaling features associated with cloud-based resources? @Andreas_Olsson Mentioned that splits would be useful in distributing load. I have experience with multi-node clusters in the clusterODM configuration but am wondering if any performance can be gained by turning on elastic scaling for an instance running WebODM. Would the application be distributed over the scaled cluster at all, e.g. by worker? Likely the answer is ‘no’ - which is why clusterODM exists.
I don’t track the cloud buzz words as closely as I should, so I don’t know if elastic == autoscale but if you’re using a cloud provider, there’s no reason to allocate static nodes for ClusterODM:
In other words, nodes are allocated on demand as needed. Combined with split merge, this is incredibly powerful and cost saving.
Thanks for the reminder about this. I do believe elastic = autoscale (up and down). I first ran into the term in using Spark/Databricks where more machines can be added into the cluster automatically as needed. Great to see the pointer on how to do this correctly with ClusterODM. In the past I used static nodes and manually assigned them.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.