Cluster computing systems are becoming increasingly popular for  providing cost-effective and affordable computing environments for a  range of applications (such as databases, data mining, visualization,  collaborative computing, and high performance computing).  These  applications require four fundamental communication services from the  underlying communication subsystem: point-to-point communication,  collective communication, synchronization, and Quality of Service  (QoS).  Modern Network Interface Cards (NICs) for clusters incorporate  one or more programmable processors, DMA channels, and large amount of  memory. The NICs are becoming faster, smarter, and intelligent and are  being designated as Active Network Interfaces.  Research is proposed  along the following directions to take advantage of such active  network interfaces for implementing and delivering scalable  communication services for next generation clusters: 1) Using  multi-CPU NICs and their architectural features to provide fast  point-to-point communication, 2) Designing NIC-level support for  high-performance and scalable collective communication and  synchronization, 3) Taking advantage of programmable NIC-level  features to provide QoS support, and 4) Experimentally evaluating the  benefits of the new/enhanced communication services for a wide range  of applications using message passing programming model.  Through  extensive theoretical, experimental, and simulation studies, NIC-based  solutions are planned to be designed, developed, implemented, and  evaluated for clusters with Myrinet and Gigabit Ethernet  interconnects.