When should you store derived attributes versus compute them on the fly?

When should you store derived attributes versus compute them on the fly?

• A. Store when computation is expensive or reused often; compute on the fly for frequently changing or rarely used attributes to avoid staleness
• B. Always store derived attributes
• C. Always compute on the fly
• D. Never store derived attributes

Answer: (A)

Balancing performance with data freshness guides how you handle derived attributes. Store derived attributes when the computation is expensive or the result is reused often, so you get fast access without repeatedly doing heavy work. This works well for metrics or summaries that are costly to compute and queried frequently, provided you have a strategy to keep them in sync with the source data. Compute on the fly when the attribute changes frequently or is rarely requested, to avoid serving stale values and to prevent unnecessary storage. If the source data updates often, cached or stored results would need frequent invalidation,which adds complexity. In short, store when the cost of computing is high or the value is reused often; compute on the fly when the value is volatile or seldom used. Always storing can waste space and risk staleness; always computing can hurt performance; never storing ignores caching benefits.

Balancing performance with data freshness guides how you handle derived attributes. Store derived attributes when the computation is expensive or the result is reused often, so you get fast access without repeatedly doing heavy work. This works well for metrics or summaries that are costly to compute and queried frequently, provided you have a strategy to keep them in sync with the source data.

Compute on the fly when the attribute changes frequently or is rarely requested, to avoid serving stale values and to prevent unnecessary storage. If the source data updates often, cached or stored results would need frequent invalidation,which adds complexity.

In short, store when the cost of computing is high or the value is reused often; compute on the fly when the value is volatile or seldom used. Always storing can waste space and risk staleness; always computing can hurt performance; never storing ignores caching benefits.