フリー問題

問題文

A team hosts a model and finds that peak traffic exceeds what one unit can serve. What is the documented way to increase throughput?

選択肢

1. Switch the endpoint to on-demand mode temporarily, and the peak traffic is then served out of the shared capacity pool.
2. Create a second endpoint for the same model, because each endpoint receives its own share of the capacity and adding endpoints therefore multiplies the throughput.
3. Change the unit shape of the existing cluster to a larger one, which the service applies in place without recreating the whole cluster.
4. Increase the number of model replicas on the hosting cluster, which requires a corresponding number of units.