Data center networks (DCN) have become the nexus of Cloud services, providing an infrastructure for effective data sharing among different applications and collaborators. DCNs use the innovative paradigm of software-defined networks (SDN) that advocates decoupling the control plane and the data plane of networks. Many DCNs support also network virtualization that separates the traditional Internet service providers into two new roles: infrastructure network providers (InP) and service providers (SP). The role of the SPs is to offer various services using virtual networks that are obtained by leasing resources from InPs. Combining these novel features of SDN and virtualization, DCNs offer the opportunity to better control data sharing and to provision virtual machines (VMs) with unprecedented flexibility. However, this setting raises many new challenges. One major challenge is to optimize the use of physical resources by a careful allocation of VMs to hosts, continuously balancing between the conflicting objectives of high performance and low operational costs. Another main issue is energy consumption that is playing an increasing role due to its costs and environmental impact.
DCNs have been explored so far mostly by the networking community. For some of the notable optimization problems in this setting, such as VM placement, mostly heuristic algorithms have been proposed. Since the underlying optimization problems are often NP-hard, it is natural to seek approximation algorithms with provable performance guarantees. Moreover, many of the optimization problems arising in this setting can be cast as variants of classical problems on graphs, assignment problems, or scheduling and load balancing problems. This calls for applying more elaborate techniques and graph theoretical tools, commonly used by the algorithms community, thus leveraging the long and successful history of interaction between the networking and the algorithms communities.
This workshop aims at bringing together researchers working at the intersection of the theory and practice of networking, with a special focus on design of efficient algorithms for DCNs. The workshop will assemble a diverse community of industrial and academic researchers that span these areas and will also provide opportunities for students and postdocs to engage leading researchers in the study of algorithms for DCN management. The workshop will provide a venue for a stimulating interaction between the networking and the algorithms communities for a mutual study of algorithms for DCNs. As part of this, one outcome of the workshop will be a document containing the views of the workshop attendees on the most important algorithmic problems involving DCNs. This will hopefully stimulate research in the way past "open problems" lists spur further research in other areas of algorithm design and networking.