Work preference

[killme2008]

What problem does the new feature solve?

By default, GreptimeDB's queries and writes fully utilize resources like CPU.
However, in certain specific scenarios, you might prefer to prioritize real-time writes at the expense of query performance, or vice versa. Therefore, I think we can provide a workPreference option to let users decide:

• The default value would be balance,
• with options to choose read or write to indicate specific work preferences.

Once a specific preference is selected, configurations such as concurrency limits and task priorities can be optimized accordingly.

FYI. This feature is useful for standalone or edge scenarios. In cluster versions, workload separation can always be achieved using different node pools.

What does the feature do?

Let user choose the workload preference.

Implementation challenges

No response