Why would a system designer use per-feature consistency trade-offs?

Why would a system designer use per-feature consistency trade-offs?

• A. To simplify design by using a single model for all data.
• B. Because some data can tolerate temporary inconsistency, enabling higher availability and performance.
• C. To guarantee global consistency for all data.
• D. To eliminate replication.

Answer: (B)

Per-feature consistency trade-offs mean applying different levels of data consistency to different features of a system. Some parts of a system don’t need perfectly synchronized data all the time; allowing those parts to be temporarily inconsistent can keep the system available and fast, even during network delays or partial failures. This is the practical reason for choosing weaker consistency for certain data: you gain higher availability and lower latency by avoiding heavy coordination everywhere. That’s why the best answer is the option that says some data can tolerate temporary inconsistency, enabling higher availability and performance. It captures the core benefit of tailoring consistency levels to what each feature needs. The other ideas aren’t aligned with this approach: aiming to simplify by using a single model for all data ignores the benefits of selective consistency; insisting on global consistency for all data contradicts the notion of per-feature trade-offs; and eliminating replication isn’t the goal of choosing different consistency levels for features.

Per-feature consistency trade-offs mean applying different levels of data consistency to different features of a system. Some parts of a system don’t need perfectly synchronized data all the time; allowing those parts to be temporarily inconsistent can keep the system available and fast, even during network delays or partial failures. This is the practical reason for choosing weaker consistency for certain data: you gain higher availability and lower latency by avoiding heavy coordination everywhere.

That’s why the best answer is the option that says some data can tolerate temporary inconsistency, enabling higher availability and performance. It captures the core benefit of tailoring consistency levels to what each feature needs.

The other ideas aren’t aligned with this approach: aiming to simplify by using a single model for all data ignores the benefits of selective consistency; insisting on global consistency for all data contradicts the notion of per-feature trade-offs; and eliminating replication isn’t the goal of choosing different consistency levels for features.