Seminar on Clean Slate Design for the Internet

Seminar on “Clean Slate Design for the Internet” Nick Mc. Keown nickm@stanford. edu 1

High level 2 v “Given what we know today, if we were to start over with a Clean Slate, how would we design a global communications network? ” v “Ideally, how will the network look in 15 -20 years, and how will we get there from here? ”

Prelims 3 v What’s wrong with the Internet…? v Why is the research and business community not already solving it? v What are other groups doing? v What we plan to do at Stanford v An example of “Clean Slate” design

Original Architecture 4 v A dumb connectionless packet-forwarding packetswitched infrastructure, with high-level functionality at the edge v Single, simple lowest-common denominator data delivery service (IP), with reliable stream service built on top v Fixed-size numerical addresses with {network, host} hierarchy; one per physical network interface v Later § Separation of IP and TCP (including congestion control using packet loss as congestion signal) § Subnetting, autonomous systems (EGPs and IGPs), DNS, CIDR

What is needed v Wouldn’t we like a network that we can trust to be always there, always on, easy to use, universally accessible, secure, and economically viable. v David Cheriton’s example: If the FAA carried all of its traffic over the public Internet, you'd be nuts to fly. v Some obvious desirable characteristics v 5 § Robustness and Availability § Security § Naming and Addressing: accountability vs anonymity § Predictability § Mobility § Economic Viability What else?

Prelims 6 v What’s wrong with the Internet…? v Why is the research and business community not already solving it? v What are other groups doing? v What we plan to do at Stanford v An example of “Clean Slate” design

Prelims 7 v What’s wrong with the Internet…? v Why is the research and business community not already solving it? v What are other groups doing? v What we plan to do at Stanford v An example of “Clean Slate” design

What are others doing? v 8 Background § Incrementalism and “victim of success” of Internet § New era of more radical and fundamental thinking about the future of networks and communications v New-arch (MIT) v 100 x 100 (CMU) v Geni (NSF/Gov)

New-arch (2000) v v 9 Requirements for new network § Mobility: Highly dynamic and efficient § Policy-driven auto-configuration § Highly time-variable resources § Allocation of capacity http: //www. isi. edu/newarch/

100 x 100 (CMU/Stanford/Rice) v NSF Large ITR (2003 -2008) v Questions: v 10 § Can structure be used to make networks more robust, predictable and manageable? § What economic principles drive the operation of access and backbone networks? § What security primitives must be built into the network? § Can/should network and protocol architectures be designed to take advantage of long-term technology trends? http: //100 x 100 network. org/

NSF Geni Initiative (2005) 11 v CISE major effort, seeking congressional funding of approx $300 M starting 2008 v Two parts: Research program; Global experimental facility to explore new architectures v Areas of interest: § Creating new core functionality, including naming, addressing, identity, management. § Developing enhanced capabilities: building security intot he architecture; design for high availability; privacy/accountability; design for regional differences and local values § Deploying and validating new architectures § Building higher-level service abstractions § Building new services and applications § Developing new network architecture theories

Prelims 12 v What’s wrong with the Internet…? v Why is the research and business community not already solving it? v What are other groups doing? v What we plan to do at Stanford v An example of “Clean Slate” design

Prelims 13 v What’s wrong with the Internet…? v Why is the research and business community not already solving it? v What are other groups doing? v What we plan to do at Stanford v An example of “Clean Slate” design

What we plan to do at Stanford v Weekly Seminar in Fall and Winter § v 14 Fall: Talk by professor followed by discussion Goals § To get thinking about the problem § To learn from each other § To identify some collaborative research projects

Prelims v What’s wrong with the Internet…? v Why is the research and business community not already solving it? v What are other groups doing? v What we plan to do at Stanford v An example of “Clean Slate” design How to design backbone networks from a clean slate? 15

Backbone Networks: Emerging Structure v 10 -50 routing centers interconnected by long-haul optical links v Increasingly rich topology for robustness and loadbalancing v Typical utilization < 25%, because v 16 § Uncertainty of traffic matrix network is designed for § Headroom for future growth § Headroom to carry traffic when links and routers fail § Minimize congestion and delay variation Efficiency sacrificed for robustness and low queueing delay

How flexible are networks today? What fraction of allowable traffic matrices can they support? Abilene 25% Over Prov: 0. 025% 50% Over Prov: 0. 66% AT&T 25% Over Prov: 0. 0006% 50% Over Prov: 0. 15% 17 Verio 25% Over Prov: 0. 0004% 50% Over Prov: 1. 15% Sprint 25% Over Prov: 0. 0003% 50% Over Prov: 0. 06% Verio, AT&T and Sprint topologies are from Rocket. Fuel

Desired Characteristics v Robust Recovers quickly; continues to operate under failure v Flexible Will support broad class of applications, new customers, and traffic patterns v Predictable Can predict how it will perform, with and without failures v Efficient Does not sacrifice cost for robustness 18

Backbone Design v 19 Assume underlying reliable mesh of physical circuits 1. Dynamic circuit switching over underlying mesh, or 2. Load-balanced logical network. Describing today

Approach v Assume we know/estimate traffic entering and leaving each Regional Network § v Use Valiant Load Balancing (VLB) over whole network § 20 Requires only local knowledge of users and market estimates Enables support of all traffic matrices

Valiant Load-Balancing 2 r 1 r 2 /r. N r 1 1 r. N r 2 2 3 N … 4 r 4 Capacity provisioned over existing robust mesh of physical circuits 21 r 3

A Predictable Backbone Network v v v 22 Performance: 100% throughput for any valid traffic matrix. § Only need to know aggregate node traffic. § Under low load, no need to spread traffic. Robustness § Upon failure, spread over working paths § Small cost to recover from k failures: Provision approx 2 rirj/r(N-k) § Simple routing algorithm Efficient § VLB is lowest cost method to support all traffic matrices § Similar cost, while supporting significantly more traffic matrices.

How expensive would VLB be? Cost normalized to VLB routing. Cost of switching = cost of transmission for 370 miles Abilene 25% Over Prov: 0. 026% Cost: 0. 87 50% Over Prov: 0. 66% Cost: 1. 04 AT&T 25% Over Prov: 0. 0004% Cost: 0. 94 50% Over Prov: 0. 14% Cost: 1. 12 23 Verio 25% Over Prov: 0. 0003% Cost: 0. 99 50% Over Prov: 1. 08% Cost: 1. 19 Sprint 25% Over Prov: 0. 0002% Cost: 0. 86 50% Over Prov: 0. 04% Cost: 1. 04

Open questions v 24 Worst case propagation delay doubled § Low variance in delay § There are “express paths” v (How) are multiple VLB networks connected, and how does performance change? v Economics and policy: how do operators compete?