Even the most advanced IT systems cannot grow and expand forever.
In the years to 2010, the Internet enters its fourth decade and is reaching two important limits which are inherent in the current architecture. Firstly, the exhaustion of previously unused IPv4 address space, and secondly a set of inter-related scaling problems in the BGP (Border Gateway Protocol) routing system which drive up the cost of the core routers used by every ISP and large end-user network – such as those of corporations and universities.
These problems contribute to the cost of Internet access for all users – at home and in business. The Internet’s Routing Scaling Problem threatens the stability of the global routing system. Together with the closely related IPv4 address exhaustion problem, the two form the Internet’s unfolding Routing and Addressing Crisis.
This crisis leads to increasingly high barriers to the entry of new ISPs and prevents millions of end-user networks from obtaining the Provider Independent address space they need to multihome their networks and avoid “provider lock-in” – by being able to choose a new ISP without having to change the IP addresses of all the computers and other devices on their networks.
At a time when many users and investors expect Internet costs to continue their historical downwards trajectory while capacity and user numbers continue to grow, we examine the problems facing the Internet, the likely solutions and the likely costs and limitations of these solutions.
A new BuddeComm biennial report in Handbook format – 2008 Technology – Internet – Volume 5 – Routing and Addressing Crisis - introduces managers and technical specialists to and important and controversial set of technical and policy challenges. The BuddeComm Handbooks provide an overview and functional understanding of important technologies which are otherwise hard to grasp without one-on-one training or lengthy study of voluminous engineering-oriented material.
The routing scaling problem and the closely related IPv4 address depletion problem directly affect every ISP’s costs, the capacity of the Internet to accommodate new networks, and the ability of end-user computers to have their own unique address. These problems are rarely discussed outside the IETF and IRTF (Internet Engineering and Research Task Forces). Until these problems are resolved, competition policy, reliability and cost problems will escalate, especially with the accelerating adoption of Internet communications in developing countries.
It is widely recognised that fresh IPv4 address space will become difficult or impossible to obtain around 2010, and that it is impractical for most users to adopt IPv6 – the second version of the TCP/IP protocols and addressing system with its essentially boundless 128 bit address space.
We review the needs of end-users and discuss in detail the scaling problems of individual routers and the BGP routing system as a whole. We then consider the role the current architecture plays in the relatively low rates of utilisation of IPV4 address space. While IPv4′s 3.7 billion addresses are a fundamental constraint, we consider how a new routing and addressing architecture could enable efficient use of these addresses than is possible with today’s techniques.
We discuss the major proposals for improving the BGP protocol and the current routing system. However, these are marginal improvements and are not a solution to the major challenge of providing an ever-growing number of end-user organisations with portable address space, multihoming – using two or more ISPs to create a robust, reliable Internet connection – and Traffic Engineering, the ability to manage incoming packet flows over multiple links. A new architecture is required so these goals can be achieved, for potentially millions of businesses and other organisations, without further burdening the BGP routing system.
Since the Internet’s inception, these have been achieved by directly using the BGP routing system so that each of the Internet’s 200k+ Default-Free Zone routers has to make a separate “best path” decision for each network’s division of the address space. There are currently 250k divisions and this number is doubling approximately every four years.
We discuss the constraints on a new architectural solution, including complete backwards compatibility for computers in networks which have not adopted the new architecture, and the requirement that there be no changes to the operating systems or application programs of user’s desktop PCs, servers and mobile devices.
We explain and compare the four current proposals for achieving these goals: LISP-NERD, LISP-CONS, APT and Ivip. These are all overlay systems, operating at the IP packet level. These involve upgrades to some routers but are intended to relieve the burden on the BGP system. These proposals are ‘Locator / Identifier Separation’ protocols, involving Ingress and Egress Tunnel Routers, with some broad similarities and important differences.
Finally, we consider some pervasive difficulties which are likely to result from any one of these new architectural approaches, including the longer headers of tunneled packets, the resulting inefficiencies for short packets (such as VoIP packets) and problems with maximum packet length and fragmentation for longer ones.
This report is intended for technical specialists and managers who require insight into the future of Internet communications in the three to fifteen year timeframe. This is the only report extant on this important topic and is intended to support readers by providing fundamental understanding of the most important principles, in order to enable them to better plan their own product and service development, investment and regulatory activities.
For more info see:
- 2008 Technology – Internet – Volume 3 – IP Techniques
- 2008 Technology – Internet – Volume 4 – IPv6
- 2008 Technology – Internet – Volume 5 – Routing & Addressing Crisis
We invite your comments: Please click here to comment