Keynote speech

Not only are data centers increasing in sizes and numbers, but also is there a growing need for cloud computing over multiple data centers to improve scalability, resilience, and response time etc.. Meanwhile, there are also many challenges and opportunities to make such distributed processing over multiple data centers feasible. More specifically, while it is certainly feasible to connect multiple data centers with an optical wavelength division multiplexed (WDM) network providing 10 Gbps or higher bandwidth per lightpath, network resources in terms of lightpaths for data transfers between data centers are more scarce (and costly) in that at any given time, only a few lightpaths for data transfers between data centers can be established. Accordingly, we must put more emphasis on network resource allocation and scheduling in a distributed multi-datacenter system. In this talk, I will focus on new approaches that jointly allocate and schedule computing and networking resources for optical cloud computing, a new paradigm that supports distributed processing over multiple data centers connected via WDM optical networks.