Achieving high utilization with software-driven WAN

Abstract

We present SWAN, a system that boosts the utilization of inter-datacenter networks by centrally controlling when and how much traffic each service sends and frequently re-configuring the network's data plane to match current traffic demand. But done simplistically, these re-configurations can also cause severe, transient congestion because different switches may apply updates at different times. We develop a novel technique that leverages a small amount of scratch capacity on links to apply updates in a provably congestion-free manner, without making any assumptions about the order and timing of updates at individual switches. Further, to scale to large networks in the face of limited forwarding table capacity, SWAN greedily selects a small set of entries that can best satisfy current demand. It updates this set without disrupting traffic by leveraging a small amount of scratch capacity in forwarding tables. Experiments using a testbed prototype and data-driven simulations of two production networks show that SWAN carries 60% more traffic than the current practice.

Joint QoS and Congestion Control Based on Traffic Prediction in SDN

“…Recently, several works are done in routing over the SDNs [1]. Several recent works have exploited the global visibility offered by SDN controllers to distribute traffic across pre-computed paths in optimal and suboptimal manners such as Software-driven wide area network (SWAN) [2] and B4 [3] that the former one uses k-shortest paths for routing across the SDN switches while the latter one presents a greedy heuristic to ensure fairness; however, both of them lack sufficient flexibility and path diversity to handle unexpected situations, i.e., traffic bursts, link failures. Recently, authors in [4,5] tackled models (e.g., oblivious routing for wide-area network traffic engineering [6,7]) and estimate traffic tools in order to cope with these problems.…”

Virtual machine cluster mobility in inter-cloud platforms

“…The solution supports cloning of VMs but not virtual networks. In (12) authors present SWAN that enables an efficient and flexible inter-DC (datacenter) WAN by coordinating the sending rates of services and centrally configuring the network data plane by focusing on updates of a network policy without causing transient congestion (12). In (13) authors present B4 that is a private WAN connecting Google data centers across the planet.…”

Traffic engineering approach to virtual-link provisioning in software-defined ISP networks

“…interconnect different sites of a company or IoT devices with Cloud services). In order to provide the necessary QoS, ISP providers usually adopt a strategy of link overprovisioning [1][2]. However, due to high costs associated with Wide Area Network (WAN) links, such a strategy cannot be economically justified in the long run.…”

“…Namely, in most of the existing production Multi Protocol Label Switching (MPLS) networks, ingress routers greedily select routes for traffic flows. As a result, the network usually stucks in globally suboptimal routing patterns [1]. To overcome limitations of classic MPLS TE, the Internet Engineering Task Force (IETF) proposed Path Computation Element (PCE) architecture, where a dedicated element is in charge of centralized path computation [4].…”

“…SDN TE mechanisms for Data Center (DC) networks, inter-DC networks and small-scale WANs have already shown a good performance [1][2][3], [6][7][8]. However, these solutions cannot be directly applied to ISP networks, which are characterized with unique requirements and properties.…”