- Speaker : Paul Francis
- Title: Dirty-slate approaches to scaling global Internet routing
- Date: 24/08 14h-17h (2h)
- Abstract: The Internet research community has in recent years taken a renewed interest in the long-standing global Internet routing scaling problem. Much of this research has taken the form of much-needed "clean-slate" research, where ideas are pursued unconstrained by the installed base. Our research group takes an opposite but complementary approach. We look for solutions that are economically motivated and therefore have a decent chance of actually being deployed. One such solution is Virtual Aggregation, an approach that exploits tunneling and aggregation to reduce the size of hardware routing tables (the FIB) easily by an order of magnitude. This work is being pursued inthe IETF. More recently, we are working on adapting Virtual Aggregation to shrinking software routing tables (the RIB) as well. In this session, we will discuss the root causes of scaling problems in the Internet, review the history of scaling solutions, and present Virtual Aggregation in detail.
- Speaker : Lars Eggert
- Title: Future Internet Standardization Activities in the IETF/IRTF
- Date: 25/08 8h-11h (2h)
- Abstract: The Internet Engineering Task Force (IETF) and its research arm, the Internet Research Task Force (IRTF) have been developing and standardizing the majority of the core Internet protocols over the last 25 years. This talk will give an overview of the two bodies, highlight Future-Internet-related activities that are already underway or are likely to kick off in the near future, as well as describe how new research results can be brought forward for standards consideration.
- Speaker : Bryan Ford & Janardhan Iyengar
- Title:Rethinking Transport Layering
- Date: 25/08 13h30-15h30 (2h)
- Abstract: Current Internet transports conflate functions driven by underlying network requirements, such as endpoint naming and congestion control, with functions driven by application requirements, such as retransmission and reordering. Dividing these function areas into layers separated by a clean interface addresses several limitations of the Internet's architecture, enabling middleboxes such as firewalls, network address translators, and performance enhancing proxies to operate without interfering with end-to-end semantics or security mechanisms such as IPsec and HIP. We explore two alternative instantiations of this architectural principle: one in a ground-up redesign of the Internet's transport suite, the other a mere reorganization of existing transport protocols. The former approach offers efficiency advantages such as smaller headers and fewer round-trips to initiate a connection, while the latter approach may be easier to deploy quickly.
- Speaker : Bruno Quoitin
- Title:Realistic interdomain routing simulations with C-BGP
- Date: 25/08 16h-18h (2h)
- Abstract: C-BGP is a BGP routing solver. It was developed as an alternative to traditional packet-level simulators for playing with BGP on large scale topologies. C-BGP computes the outcome of the BGP routing decisions based on a fairly detailed description of the layer-3 topology, the interconnection of BGP routers and the BGP routing policies.
In this session, we will first briefly describe the design of C-BGP,
its recent evolution and the kind of experiments it can be used for. In a second time, we will present two different applications of C-BGP: modeling an Internet-like topology and modeling an ISP's backbone network.
- Speaker : Bob Briscoe
- Title: Practical Microeconomics and Internet Resource Sharing Protocols
- Date: 26/08 9h-12h (3h)
- Abstract: This tutorial gives a primer on the microeconomics of sharing a pooled resource like Internet bandwidth. This knowledge is then used in 2 ways:
- To assess how the economics has led to unintended consequence, given today's resource sharing protocols: TCP, weighted fair queuing, volume capping and deep packet inspection (and similar problems with the recent proposals XCP & RCP).
- To motivate an approach to resource sharing based on understanding of the economic drivers and pitfalls. Kelly's proposal to use Floyd's Explicit Congestion Notification (ECN) for pricing will be explained, as well as the recent re-ECN proposal. Further unintended consequences of these approaches will be outlined, including potential market failures.
- Speaker : Mathieu Lacage
- Title: Network experimentation and simulation with ns-3
- Date: 27/08 14h-17h (3h)
- Abstract: Intended to eventually replace the popular ns-2 simulator,ns-3 aims to retain the successful features of ns-2, while remedying some complaints about it. In this presentation, we will describe the design of ns-3 and how it relates to some of its long-term goals, such as improving the integration of simulation tools with real-world applications and experimentation testbeds.
- Speaker : Timothy G. Griffin
- Title: From Semirings to Metarouting
- Date: 28/08 9h-12h (3h)
- Abstract: The metarouting project is developing a system for the declarative specification of routing protocols. A metalanguage is used to specify "routing algebras", structures related to semi-rings. Algebraic properties, such as monotonicity, are automatically inferred from metalanguage specifications. Code is generated and linked with algorithms selected from our library, which includes generalized versions of standard Internet routing protocols that we have extracted from the Quagga code base. The inferred properties are used to ensure that only a correct matching of an algebra and an algorithm can take place.
- Speaker : Dimitri Papadimitriou
- Title: Compact Routing: Challenges, Perspectives, and Beyond
- Date: 28/08 13h30-15h30 (2h)
- Abstract: Compact routing schemes were until recently considered as the main alternative to overcome the fundamental scaling limitations of the Internet routing system. Such schemes have been introduced to address the fundamental and unavoidable trade-off between the stretch of a routing scheme and the routing table size it produces. Recent studies have shown that static routing on topology-dependent identifiers (that includes "some" topological information) scales (poly)logarithmically on the number of nodes in scale-free graphs such as the Internet. Nevertheless, evolution of the Internet routing system also demonstrates that i) topology-independence of the numbering scheme becomes a fundamental requirement for e.g. multi-homing, nomadicity, and mobility, and ii) "static" routing is unable to handle dynamic graphs, i.e., routing update messages (triggered by topology-driven, policy-driven, and other protocol-driven events) must be exchanged to timely inform remote routers such that each router maintains a consistent view of the non-local topology.This tutorial starts by reviewing the fundamental dimensions of routing schemes. It then introduces the compact routing theory and its application to scale-free graphs. Next, it describes and analyses the consequences resulting from i) routing on topology-independent numbering space and ii) dynamic routing information exchanges and resulting costs. This tutorial will conclude by outlining the open research coupled-challenges (stretch x routing table size x dynamics) and possible direction(s) to address them.
Internet Reloaded