City Power & Light is choosing compute for a multi-agent outage-response system. Requirements: burst scale during storms, isolate long-running planner jobs from low-latency chat agents, maintain reliability with health probes, and control cost during calm weather. Which compute specification is most appropriate?
Select an answer to reveal the explanation.
Short Explanation
Split hot-path chat from long-running planners, autoscale the chat tier, put batch/planning on worker/job compute, and scale down when storms end. That is option A — scalability, reliability isolation, and cost efficiency together. One giant shared VM (B) couples failures and wastes money. Browser-only agents (C) cannot meet outage reliability. Fixed average capacity in one zone (D) fails storms and availability design.
Full Explanation
Architects must specify compute that meets scalability, reliability, operational efficiency, security, and cost requirements. Option A separates latency-sensitive and long-running workloads, enables elastic scale for storm bursts, and reduces cost at idle.
Option B creates a single point of failure, noisy-neighbor risk, and poor cost elasticity.
Option C is not a production control plane for critical outage response and fails offline/unmanaged device realities.
Option D ignores burst demand and multi-AZ/region reliability patterns.
Exam tip: Heterogeneous agent workloads usually need heterogeneous compute profiles, not one VM for everything.