[talks] Xin Jin will present his Pre FPO on Monday, January 11, 2016 at 4pm in CS 302

[Nicki Gotsis]

Xin Jin will present his Pre FPO on Monday, January 11, 2016 at 4pm in CS 302.

The members of his committee are:  Jennifer Rexford (adviser), David Walker (reader), Ratul Mahajan (Microsoft Research, reader), Nick Feamster (non-reader), Aarti Gupta (non-reader).

Everyone is invited to attend his talk.  The talk title and abstract follow below:

Title:
Dynamic management of software-defined networks

Abstract:
Network management is a difficult problem, since network administrators have to continuously update network configurations to alleviate congestion by routing more efficiently, to detect and block cyber-attacks, to perform planned maintenance, etc. Network updates are painful because operators unfortunately have to balance the tradeoff between the disruption caused by the problem (e.g., congestion and cyber-attacks), and the disruption introduced in fixing the problem. In this talk, I will present my research on building new network management systems that can (1) efficiently compose configuration updates from multiple network services, (2) quickly apply them to switches, and (3) effectively handle topology changes.

I will first present CoVisor, a new kind of network hypervisor that enables, in a single network, the deployment of multiple network services written in different programming languages and operating on different controller platforms. CoVisor provides a simple API that allows network administrators to define the composition of multiple controllers, to choose the virtual topology exposed to each controller, and to constrain the packet handling capability of each controller. We developed a new set of efficient algorithms for composing controller applications and for compiling configurations of virtual networks into those of the physical network. We built a CoVisor prototype and showed that it is several orders of magnitude faster than a vanilla implementation without our optimizations.

In the second part of the talk, I will present Dionysus, a new network update system that can quickly and consistently update network configurations. Dionysus uses a novel tool called a dependency graph to capture the dependencies between individual update operations. Based on the dependency graph, Dionysus dynamically schedules operations according to runtime conditions, and guarantees the consistency of the updates. We built a prototype of Dionysus and extensively evaluated it with testbed experiments and large-scale simulations. Results showed that Dionysus improves the median update speed by 53–88% as compared to prior methods.

Finally, I will present Owan, a novel traffic management system that optimizes wide-area big data transfers with centralized joint control of the optical and network layers. Owan can dynamically change the network-layer topology by reconfiguring the optical devices. We developed efficient algorithms that dynamically reconfigure the optical circuits in the optical layer so that most transfers can use one or few hops in the network layer to reach their destinations. Testbed experiments and large-scale simulations showed that Owan completes big data transfers up to 4× faster on average, and up to 1.35× more transfers meet their deadlines, as compared to prior methods that only control the network layer.