You have a Microsoft Agent Framework workflow processor that handles customer support requests from a queue. Each request is evaluated by three independent Microsoft Foundry agents. You discover that the existing processor dequeues 30 requests at a time and starts every agent run immediately. During peak load, up to 90 agent runs execute simultaneously, and the downstream API receives partial participant messages. A single agent run completes in five seconds at the 95th percentile (95p), and the target throughput is 120 requests per minute. You need to change the orchestration so that it meets the throughput target, prevents more than 30 agent runs from executing simultaneously, and produces one consolidated downstream payload for each request. What should you do?
Running 10 three-participant ConcurrentBuilder workflows creates at most 30 simultaneous agent runs (10 × 3). Because the three independent agents run in parallel, each batch of 10 requests can complete in approximately five seconds at the 95th percentile, which equals 120 requests per minute. Concurrent orchestration aggregates participant results into a single terminal response, enabling one consolidated downstream payload rather than partial participant messages. A sequential workflow is the strongest alternative for enforcing a limit, but it makes the three five-second agent runs additive for each request. A 10-request batch would take about 15 seconds, yielding only about 40 requests per minute.