Inter-Datacenter Multicast with Store-and-Forward in Software-Defined Optical Networks.

Xiao Lin,Jiangnan Zou,Shengnan Yue,Weiqiang Sun,Weisheng Hu

DOI: https://doi.org/10.1109/ccnc49032.2021.9369482

2021-01-01

Abstract:Emerging online services have fueled unprecedented demands for multicast transfers across geo-distributed datacen-terse However, the time- and space-varying nature of background traffic in inter-datacenter networks (inter-DCNs) makes conventional multicast trees difficult to fully utilize the residual bandwidth in the inter-DCNs. In this paper, we introduce datacenter storage into multicast transfers, and propose Store-and-Forward-Tree (SnFTree) scheduling method for multicast transfers in the software-defined optical network. Compared to the conventional multicast scheduling methods, the advantages of SnFTree are as follows: (i) by enabling temporary storage at intermediate sites, SnFTree can segment the conventional multicast tree into independent subtrees, which maximizes the flexibility of provisioning; (ii) by introducing load-balancing routing into the tree computation, SnFTree can balance the traffic load across network efficiently; (iii) benefiting from optical multicast, storing, forwarding and reception can be simultaneously implemented on a site with negligible processing delay; (iv) by formulating the multicast scheduling problem into a routing problem, SnFTree greatly simplifies the problem for dynamic traffic. Simulations demonstrate that compared to conventional scheduling methods, SnFTree obtains lower blocking probability and requires fewer OpenFlow Agent Updates.