Overview
Cluster Compute Instances provide a high-performance network interconnect along with a high-performance CPU. For example, the cc2.8xlarge instance type has quad-core processors with hyper-threading enabled (8 physical cores with 16 threads). When these instances are grouped together and launched in the context of an EC2 Placement Group, the instances benefit from a high bandwidth, non-blocking interconnect system, which makes them ideal for applications that require lots of network I/O. (e.g. High Performance Compute (HPC) applications and other demanding network-bound applications)
Current Limitations
Be sure to reference Amazon's documentation for current limitations. See Cluster Compute Instance Concepts.