How do you think about ES hot/warm architecture

Generally, I think warm node can act like hot node, but slowly. And machine cost a lot saved.

But ES runs on Memory and require sufficient memory to process request regardless of how it fast.

Memory is the most expensive part in machine.
Imagine 10TB data, perhaps about 1TB memory will be required. we can not supply cheap memory if same query need to run on warm node. In that situation cheap disk or less cpu core can not save the total cost a lot.

Finally I think hor warm architecture will not save money than I expected.

How do you operate hot/warm node hardware?

A hot/warm architecture is very useful for a lot of use-cases entered around time-based data and indices. It is however not necessarily suitable for all types of use cases. If you can provide some details around your use case, we may be able to provide more appropriate guidance.

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.