Increasing Internet Connectivity Through the Development of Local Networks
Total Page:16
File Type:pdf, Size:1020Kb
Increasing Internet Connectivity through the Development of Local Networks Written for the Office of the Secretary-General of the International Telecommunications Union in reference to Administrative Region E Keertan Kini, WeiHua Li, Daniela Miao, Lauren Stephens {krkini16, wli17, dmiao, lhs} @mit.edu 1 Acknowledgements The authors would like to express their gratitude towards Danny Weitzner, Peter Lord, Hal Abelson, and Alan Davidson for their valuable guidance, mentorship and feedback on this paper. This work could have not been completed without their arduous support. This paper also benefits significantly from the expertise of Jessie Stickgold-Sarah, Michael Trice and Anna Wexler. Their advice on general argument formation was crucial to the success of this work. 2 Table of Contents Executive Summary 1. Historical Background 1.1 Scale and Inefficient Routing Problems 1.2 Inefficient Routing Phenomenon 1.3 Background on Internet eXchange Points (IXPs) 1.4 IXPs Current State 2. Problem 2.1 Internet Connectivity 2.2 Past Cable Faults Case Studies 2.3 Cost of Latency and Loss of Connectivity 2.4 IXPs as More Secure Alternative for Local Traffic Exchange 3. Policy Recommendations 3.1 Relevance to ITU 3.2 Benefits of IXPs 3.3 Past Policies Regarding IXPs 3.4 Specific Policy Recommendations: India and Bangladesh 4. Support for Policy Recommendations 4.1 Case study: Kenya 4.2 Case study: Singapore 4.3 Case Study: Brazil 5. Analysis of Policy Recommendations 5.1 Avoiding Balkanization of the Internet 5.2 Political Challenges for Implementation 5.3 Scalability of Local Replication 5.4 Effectiveness of the Local Replication 5.5 Privacy Issues 6. Conclusion Appendix I: Glossary Appendix II: Data on Submarine Cable Vulnerabilities Appendix III: Past Attempts to Mitigate Submarine Cable Vulnerabilities Appendix IV: Latency and Internet Quality Appendix V: Data on IXPs and Local Data Centers Bibliography 3 Executive Summary This document analyzes the resiliency and connectivity of the Internet in Administrative Region E (Asia and Australasia) of the Internet Telecommunications Union (ITU). We consider how to reduce the widespread outages caused by cable faults in natural disasters such as the 2006 earthquake and the 2011 typhoon. To improve the resiliency for the network, we advocate adding Internet eXchange Points (IXPs) to mitigate existing issues with submarine cables, encourage colocation of data centers, reduce latency, and decrease bandwidth costs, among other benefits. In particular, we identify two countries - India and Bangladesh - where investment in IXPs and policies to encourage competition among regional ISPs can have a huge impact on the connectivity and resiliency of the network in addition to the economic and efficiency benefits IXPs confer. The first half of the paper presents a security and vulnerability based motivation for IXPs - namely the outages surrounding submarine cable failures and analyzes the local and large-scale benefits of increased numbers of IXPs and data centers. Initially, we examine the growth and development of the Internet and issues with inefficient routing. We show these vulnerabilities necessitate the building of more robust local networks within countries by utilizing IXPs. Then, we demonstrate that the construction of IXPs can alleviate resiliency concerns surrounding the submarine cable network. The second half of the paper reviews the economic and efficiency benefits of IXPs and presents policies that will encourage investment and development of IXPs. To illustrate these policies, we identify two major types of countries that can benefit from policies to encourage the growth of IXPs. The first are fast growing nations that have both the capital and market for telecom investment, but whose infrastructure is not yet completed or unified - for example India and China. The second category are nations that have less capital and little infrastructure - for example Bangladesh. We then use the case studies of India and Bangladesh to provide the following recommendations: India: Link the existing government infrastructure by IXPs-equivalents to the smaller, private ISPs in order to create a robust backbone for domestic Internet traffic and networks. Bangladesh: Limit the control of the existing for-profit monopoly to the only submarine cable by adding a public IXP to the new submarine cable expected to be finished in 2014 in order to provide reliable competitive access to the international Internet framework and encourage further domestic competition. Following the policy recommendations, we analyze case studies illustrating the success of IXP implementation in Kenya, Singapore and Brazil. Subsequently, we present detailed analyses of our recommendations to study feasibility and highlight implementation challenges. We use this analysis to show that the ITU should promote the development of IXPs in Administrative Region E beginning with the countries of India and Bangladesh. 4 1. Historical Background 1.1 Scale and Inefficient Routing Problems The Internet currently functions as a medium for spreading ideas and information. It has unparalleled depth and breadth as a tool for spreading pluralism and encouraging the marketplace of ideas. As the amount of data trafficked through the Internet has grown exponentially over the last decade and shows no sign of slowing down, the need for a robust and efficient network system has grown increasingly important. Unfortunately, the design of Internet infrastructure not evolved nearly as quickly. Due to the fact that physical facilities are governed by a number of different stakeholders, ranging from commercial corporations to government agencies, collaboration efforts to improve the existing international network infrastructure have been less than successful. This leads outdated facilities and inefficient traffic routes. As more and more bottlenecks occur on the network, data transfers experience larger delays, which ultimately hinders the dissemination of information over the Internet, and undermines the fundamental principle upon which the Internet was built. In this section, we first describe an inefficient routing phenomenon caused by outdated infrastructures, and explore methods to mitigate the issue. 1.2 Inefficient Routing Phenomenon Early on in the evolution of the Internet, there was a general lack of domestic network infrastructure in most countries. Since a majority of the network traffic is international, Internet Service Providers (ISPs) did not see a need to invest in construction of local connectivity. This scarcity of physical infrastructure imposed significant costs on domestic connections between ISPs, in particular where domestic connectivity is not very competitive (and therefore expensive). As most ISPs purchase international transit in order to provide global connectivity anyway, some find it convenient to include domestic traffic in these links, particularly in countries where access to domestic infrastructure is limited and prices are high. As a result, domestic traffic, such as an email between neighbors, may leave the country in order to be exchanged - a process sometimes known as tromboning. This process is detailed in the figure below, in which each of the ISPs in Country A uses international transit to exchange traffic with one another, as well as to exchange traffic with foreign ISPs and content providers. However, there are performance and cost issues with this method, as more and more local traffic has to be transmitted across countries before coming back to the same region. 5 Figure 6: Illustration of Tromboning, where Domestic Traffic Uses International Transit Lines While tromboning does not pose a problem when there is little domestic traffic, it is beginning to cause great inefficiencies in networks today. As the Internet market penetration grows and more homes are equipped with broadband connections, there has been a clear incline in the amount of domestic data transfers. Evidently, it would be significantly efficient if local communication such as emails between neighbor can be exchanged locally without leaving the country, unlike the illustration in Figure 6. Internet eXchange Points (IXPs) can be introduced to serve this purpose. 1.3 Background on Internet eXchange Points (IXPs) In the real world, IXPs are simply physical points that can be a switch in a basement, or a large network spanning multiple data centers, with elaborate power backup and cooling systems. These points facilitate local information exchange between different ISPs’ networks. Figure 7 below provides a simple illustration of this new setup. 6 Figure 7: Illustration of Local Data Exchange using IXPs As the figure above demonstrates, IXPs provide a point of local exchange that circumvents the process of tromboning - an email between neighbors can be simply transferred over two different ISP networks via an IXP. Not only does this data transfer take less time than internationally routing the traffic, it also reduces the traffic burden on international transit lines. In addition, there are extra performance and economic benefits associated with IXPs that will be explored in detail later in our policy recommendation section. Finally, we highlight the issue of IXP governance - how IXPs are typically implemented and administered. It is important to note that IXPs do not have significant impacts on existing network infrastructure, as they do not increase the burden domestically. Given that the traffic was already present in the first place - IXPs simply reroute