Datacast: A Scalable and Efficient Group Data Delivery Service for Data Centers
Chuanxiong Guo · 2011
Reliable group data delivery (RGDD) is a pervasive traffic pattern in data centers. In an RGDD group, a sender needs to reliably deliver a copy of data to all the receivers. Exist-ing solutions either do not scale due to the large number of RGDD groups (e.g., IP multicast) or cannot efficiently use network bandwidth (e.g., end-host overlays). We designDatacast for RGDD. Datacast turns group state management to in-network packet caching by leveraging the content centric networking (CCN) concept. It uses centrally calculated multiple edge-disjoint trees for efficient and fast data delivery. By introducing a robust rate-based congestion control, which interprets duplicated CCN interest packets as congestion signal, Datacast adapts well to varying network conditions. We have implemented Datacast using the ServerSwitch [13] platform. Our implementation shows that all the features needed in Datacast, including flexible packet filtering, source routing, and in-network packet caching, can be implemented using commodity devices. Our implementation can offload packet forwarding to switching chip for 50 % nodes in a typi-cal group with 280members and over 500 intermediate nodes. Our testbed experiments demonstrate that Datacast achieves 1.74Gb/s goodput using two 1GbE Steiner trees, it uses net-work bandwidth efficiently, adapts to changing traffic condi-tions and reacts to network failures quickly. 1.