A tightly coupled simulation runs on supported EC2 instances. Workers exchange large volumes of data and require very low inter-instance network latency. This particular job can run within one Availability Zone. Which placement strategy is most appropriate?

1. A cluster placement group in one Availability Zone, using instance types appropriate for the network requirements.
2. A partition placement group solely to guarantee the lowest possible latency between every pair of workers.
3. A spread placement group distributed across Availability Zones to minimize every worker-to-worker latency.
4. One instance per AWS Region with a public internet connection between workers.