The research community has considered in the past the application of Artificial Intelligence (AI) techniques to control and operate networks. A notable example is the Knowledge Plane proposed by D.Clark et al. However, such techniques have not been extensively prototyped or deployed in the field yet. In this paper, we explore the reasons for the lack of adoption and posit that the rise of two recent paradigms: Software-Defined Networking (SDN) and Network Analytics (NA), will facilitate the adoption of AI techniques in the context of network operation and control. We describe a new paradigm that accommodates and exploits SDN, NA and AI, and provide use cases that illustrate its applicability and benefits. We also present simple experimental results that support its feasibility. We refer to this new paradigm as Knowledge-Defined Networking (KDN).Comment: 8 pages, 22 references, 6 figures and 1 tabl

During the last years several operators have expressed concerns about the continued growth of the BGP routing tables in the default-free zone. Proposed solutions for this issue are centered around the idea of separating the network node's identifier from its topological location. Among the existing proposals, the Locator/ID Separation Protocol (LISP) is the one with the biggest momentum. In LISP, a mapping system is required to provide bindings between locators and identifiers. In this paper we present a new mapping system: LISP-TREE. It is based on DNS and has a similar hierarchical topology: blocks of identifiers (EIDs) are assigned to the levels of the hierarchy by following the current IP address allocation policy.It can work with the existing DNS implementations, providing the benefit of 20 years of operational experience. In addition, we present measurement-driven simulations of mapping systems' performance, assuming deployment of LISP in the current Internet. Index TermsRouting scalability, locator/identifier split, mapping system.

Concerns regarding the scalability of the inter-domain routing have encouraged researchers to start elaborating a more robust Internet architecture. While consensus on the exact form of the solution is yet to be found, the need for a semantic decoupling of a node's location and identity is generally accepted as the only way forward. One of the most successful proposals to follow this guideline is LISP (Loc/ID Separation Protocol). Design wise, its aim is to insulate the Internet's core routing state from the dynamics of edge networks. However, this requires the introduction of a mapping system, a distributed database, that should provide the binding of the two resulting namespaces. In order to avoid frequent lookups and not to penalize the speed of packet forwarding, map-caches that store temporal bindings are provisioned in routers. In this paper, we rely on the working-set theory to build a model that accurately predicts a map-cache's performance for traffic with time translation invariance of the working-set size. We validate our model empirically using four different packet traces collected in two different campus networks.

Abstract-OpenOverlayRouter (OOR) is an open-source software router to deploy programmable overlay networks. OOR leverages the Locator/ID Separation Protocol (LISP) to map overlay identifiers to underlay locators, and to dynamically tunnel overlay traffic through the underlay network. LISP overlay state exchange is complemented with NETCONF remote configuration and VXLAN-GPE encapsulation. OOR aims to offer a flexible, portable, and extensible overlay solution via a user-space implementation available for multiple platforms (Linux, Android, and OpenWrt). In this article, we describe the OOR software architecture and how it overcomes the challenges associated with a user-space LISP implementation. Furthermore, we present an experimental evaluation of OOR performance in relevant scenarios.

Since so far, most studies on path-vector routing stability have been conducted by means of ad-hoc analysis of Border Gateway Protocol (BGP) data traces. None of them consider the specification of an analytic method including the use of stability metrics for the systematic analysis of BGP traces and associated meta-processing for determining the local state of the routing system. In this paper, we define a set of stability metrics that characterize the local stability properties of path-vector routing such as BGP. By means of these metrics, we derive a stability decision criterion that can be applied during the BGP route selection process. Results obtained using real BGP datasets show that 90% of the routes are not affected by a path length increase when selected based on this criterion. Moreover, among the remaining 10%, a significant fraction of the routes is covered by a path length increase of one-hop. These results corroborate the assumption that enforcing stability would not come at the detriment of increasing the stretch of the routing paths.

No abstract

In this paper, we define a set of metrics that characterize the local stability properties of path-vector routing protocols such as BGP (Border Gateway Protocol). By means of these stability metrics, we propose a method to analyze the effects of BGP policy-and protocol-induced instability on local routers.

Traditionally, efficient inter-domain data delivery may be implemented either as a network or application layer multicast service. However, while the former has seen little uptake due to prohibitive deployment costs the latter is widely used today, but often without a minimum guaranteed performance. In this paper we present Lcast, a network-layer single-source multicast framework designed to merge the robustness and efficiency of IP multicast with the configurability and low deployment cost of application-layer overlays. The architecture involves no end-host changes and only requires the upgrading of a small set of routers to support the Locator/ID Separation Protocol (LISP), an incrementally deployable enhancement to the current global routing infrastructure. Content distribution over the Internet's core is done by means of a router overlay while within domains, end-hosts interface with Lcast using conventional multicast protocols. The overlay's scalability and topological configurability is sustained by logically centralizing group management. We illustrate the versatility of our solution by designing and assessing the scalability and performance of three management strategies for low latency content distribution. Our analysis is based on large scale simulations supported by realistic user behavior and Internet-like network topologies. The results show Lcast's low management overhead and ability to optimize delivery to meet various operational constraints. Notably, we find that it can deliver traffic with latencies close to unicast ones, independent of overlay size.