It's been 30 years since Vint Cerf, Bob Kahn, and others outlined the protocols that laid the groundwork for the development of the Internet. From a research vehicle providing universities and labs with electronic mail, remote terminal access and file transfer, the Internet has evolved into a vast commercial enterprise that has fundamentally transformed our lives.
Instant messaging, high-speed audio and video streaming, gaming, peer-to-peer file sharing, voice over IP, and conferencing tools are among the rich applications of today's Internet, changing the way we work and play. The emergence of online shopping and other e-commerce applications has leveled the business playing field, enabling the smallest companies to reach a worldwide market and some of the largest to operate without bricks and mortar. (A decade ago, amazon.com didn't exist; today it ranks 342 among the Fortune 500, with $5.62 billion in revenue in 2003.) Wireless access through a growing variety of devices is extending the reach of the Internet to remote locations and new users.
Early Evolution of the Internet
To understand the current state of the Internet (including some of its problems), it is useful to review its early evolution. The Internet was developed as an overlay to the telephone systems of roughly two decades ago. The goal was to build a unified, flexible and robust data communications network. The existing phone (and other) networks were inflexible and difficult to change, but building a completely new communications infrastructure was practically and economically unfeasible. To address the inconsistency in data connections and link the disparate networks, researchers developed the idea of "internetworking"- creating an overlay that would leverage the existing telephone infrastructure and build on top of it a new layer of functionality.
The architecture of this emerging Internet was designed to allow any user to send a packet to any other user, without restriction. This contrasted with the architecture of the telephone system, which requires callers to signal (via telephone ring) the desire to communicate and gives receivers the option of deciding whether or not to accept the call.
The Internet's architecture was highly desirable to researchers, the original users. It enabled them to communicate freely, and to quickly build prototypes, run experiments, and verify the results. This led to a rapid phase of innovation, during which protocols were honed and refined or completely rewritten. However, that same flexible architecture also makes it easy to spread worms and viruses, to initiate denial-of-service attacks, to exploit vulnerabilities and gain unauthorized access to personal and financial information, and to create other security-related problems that we see on the Internet today.
There are other problems as well, some of them by-products of the Internet's success. For example, the tremendous growth in the Internet user base has led to wide variations in access to Web sites, leading to the phenomenon of "flash crowds"- millions of users attempting to access the same Web server at exactly the same time, causing the system to fail. The current paradigm of Internet services being statically bound to hardware makes it difficult to adapt well to flash crowds.
Commercial success has also made it difficult to "fix" the Internet's problems. The Internet is a source of enormous revenues, and carriers are concerned that introducing new protocols into the network could disrupt the Internet's performance or features. One way around this impasse is to implement new protocols as overlays on the Internet in order to evolve a new, more secure Internet architecture over time. This approach was a key motivator for the development of PlanetLab.
PlanetLab: A Model for Evolving the Next-Generation Internet
PlanetLab is a network of computers strategically located at sites around the world, forming a testbed or platform for creating and deploying planetary-scale services - massive applications that span a significant part of the globe. The Akamai Technologies content distribution network is one example, and file-sharing applications are another. (As an overlay, PlanetLab itself is also a planetary-scale application.)
PlanetLab was conceived by David Culler of UC Berkeley and Larry Peterson of Princeton University. As noted earlier, one motivation for developing the platform was to evolve an improved Internet architecture by implementing new protocols as an Internet overlay. A shorter-term motivation was to respond to the research community's interest in experimenting with planetary-scale services, which were widely viewed as the likely "next big thing" to emerge since the World Wide Web.
Global distribution of PlanetLab sites. The goal is to have 1,000 nodes (computers) distributed around the world at strategic points in the network.
PlanetLab leverages the existing Internet infrastructure and adds a new layer of functionality. This parallels the early development of the current Internet, as an overlay to the telephone system that added data communication functionality. PlanetLab enables users - primarily researchers-to test and validate new planetary-scale services in an environment that is intended to replicate the environment of the Internet but does not disrupt the Internet's performance.
PlanetLab represents Intel's model of how the next-generation Internet will emerge. This new Internet will add rich computational capabilities to the existing communications functionality of the Internet. Computational resources will be embedded throughout the network, providing the capability to deliver services at specific locations in the network. Among other things, this can alleviate the problem of flash crowds by distributing resources when and where they are needed.
Launching the Projects
No university could afford to deploy the large network of computers required to create PlanetLab, and no government funding was forthcoming in the U.S., the EU or elsewhere. Intel jumpstarted the project by contributing the first 100 machines (nodes), the initial software implementation and key sites in backbone POPs and CoLo centers; establishing several international seeds; awarding research grants to a number of universities, including MIT, Rice University, the University of Washington, Princeton and UC Berkeley; and providing operational support for the project. The objective was to seed a research community and attract others to contribute in order to build PlanetLab as a shared platform that would eventually become self-sustaining.
Once the initial machines were in place, many institutions became interested in hosting PlanetLab sites. Most of the computers and all of the power, cooling and bandwidth needed to keep PlanetLab operating are contributed by individual sites.
Service Deployment
Some of the first services deployed and tested on PlanetLab were network measurement applications that monitor trends in Internet usage and growth and provide snapshots of Internet traffic. Such services are useful to planetary-scale applications that must be in many places simultaneously. For example, an application designed to stream video across the Internet must know which network connections are available and which are not.
To address the problem of flash crowds, PlanetLab researchers are exploring how services can migrate to the appropriate locations in the network where they are requested, using virtual machine technology. Thus, the code or service itself would no longer be tied to a physical machine. It moves where it needs to be, and as a result is able to provide high quality service even in the face of large flash crowds.
Beyond these services, researchers have also shown interest in applications for long-term distributed file storage (for example, for digital libraries) as well as what might be called personal mobility services- applications that enable users to migrate their computational power (operating systems, application and data files) to a remote physical location.
Rapid Expansion
Since PlanetLab was launched in August 2002, the project has expanded rapidly. At the time of this writing, there are 429 PlanetLab nodes at 181 sites throughout the world, and more sites are being added regularly. The goal is to locate 1,000 machines throughout the world, at enough strategic locations to be within a hop of the majority of traffic moving across the Internet.
In June of 2003, operational responsibility for PlanetLab was transitioned from Intel to the Princeton-based PlanetLab Consortium, whose membership includes major corporations such as Hewlett-Packard and Google. Many other institutions have joined the consortium, and the NSF [National Science Foundation] has recently contributed substantial funding.
In two short years, PlanetLab has made a significant impact on the research community, and its influence continues to grow. However, many challenges remain, among them the issue of resource allocation. Policies and mechanisms for allocating computational resources across the network are relatively primitive, leading to unpredictable contention for resources. The tools to address the problem are also fairly basic, but because PlanetLab is an open research initiative, they have improved and will likely continue to do so.
Related Research at Intel
In addition to participating in PlanetLab, Intel has initiated several research projects to support the emergence of the next-generation Internet. Some of the research involves experimental technologies that are being deployed and tested on PlanetLab. Following is an overview of four key projects.
PHI: An Information Plane for the Internet
The PHI project is aiming to improve the overall security and efficiency of the Internet, by having users monitor the condition of the network from millions of vantage points and share this data among themselves. PHI as a shared system will analyze this data in the aggregate, and deliver the results to users and their machines. The analysis can be used to pinpoint congested or faulty parts of the network and to identify Internet viruses, worms and other security anomalies.
Initially, PHI will be used to understand how worms spread across computers, identify the risky behaviors that cause the worms to spread, and to pinpoint at-risk populations of Internet users. The next step will be to educate users so they will change their behavior in order to prevent the spread of viruses and related problems, to improve the overall health of the Internet.
PHI builds on a massively distributed query system called PIER (Peer-to-Peer Information Exchange and Retrieval). Through PIER, researchers hope to develop the capability to run queries over millions of computers simultaneously through the Internet. As part of this effort, they also propose to build an attractive application that users could download to their PCs and which would provide the machines with information about current Internet activity that could be harmful, such as the latest worm signatures or virus threats, and the addresses of machines that are probing or attacking other machines in large numbers.
Researchers envision PHI as a grassroots effort to monitor the Internet and keep it healthy. Each computer that runs the visualization tool would also be contributing to the overall picture of the Internet's state.
PlanetLab is being used as a testbed for the research. PHI prototypes are already running on some 300 PlanetLab nodes around the world.
Intel Research Project
Primary Research Objective
Delay Tolerant Networking
Extend Internet access to the developing world
OpenHash
Make it easy for developers everywhere to create and test broadly useful planetary-scale applications, and for users anywhere to run them
Place Lab
Expand usage of location-enhanced computing by making it easy and inexpensive
Public Health for the Internet (PHI)
Enhance Internet security and reliability
Delay Tolerant Networking: Bringing the Internet to Remote Regions
The Delay Tolerant Networking (DTN) project is exploring how to provide reliable networking infrastructure for hard-to-reach places. As part of that effort, Intel is collaborating with UC Berkeley in a research initiative, funded by the National Science Foundation, to explore how to bring information and communications technology to the four billion people in the developing world.
To get a sense of the challenges involved, imagine a remote village where a bus visits periodically to deliver and pick up emails. This is the networking substrate that researchers have to route on. It's a unique network, involving long delays.
Researchers in the Delay Tolerant Networking Research Group (DTNRG) are exploring the architectural and protocol design requirements for providing interoperable communications in environments, such as the remote regions of the developing world above, where continuous end-to-end connectivity cannot be assumed. Specifically, they are investigating an alternative to the standard Internet TCP/IP communications protocol. The alternative uses hop-by-hop storage and retransmission as a transport-layer overlay, and provides a messaging service interface that is similar to email in concept but is application-independent and supported by specialized reliability and routing capabilities.
OpenHash: Enabling Development of Planetary-Scale Services
Distributed hash tables (DHTs) are a novel storage technology that allows for files or data to be distributed and stored in nodes or servers throughout the network. Each node contains a portion of the data and only a partial list of other storage locations, for security purposes. Researchers develop algorithms for finding and assembling the distributed data in response to user requests.
OpenHash is a publicly accessible DHT service, initiated by Intel researchers, that makes it easy for developers everywhere to build and test new, broadly useful distributed applications, and for users everywhere to run them. Other research prototype DHT implementations are available but are complicated to deploy and maintain. Among other things, the DHT code must be run on a large group of hosts, and few developers have access to these resources. OpenHash currently runs on PlanetLab, but the goal is to eventually run the service on infrastructure hosts administered by a variety of authorities.
Place Lab: Advancing Location-Enhanced Computing
Place Lab is both a software base and a community-building activity designed to advance the vision of location-enhanced computing. The goal is to spur the development of location-enhanced computing applications by providing low-cost, easy-to-use positioning technology, thus overcoming two key barriers that have inhibited growth in this new arena of computing.
To enable low-cost positioning technology, Place Lab leverages the proliferation of existing IEEE 802.11 or Wi-Fi* hotspots in homes, businesses, university campuses, and in public spaces. Location information provided by Wi-Fi access points is not as precise as that given by technologies such as GPS, but it is sufficient for many location-enhanced applications, and it can be acquired at virtually no cost.
To make the technology easy to use, researchers developed a Place Lab toolkit for developers. Developers simply ask the toolkit "where am I" and use the response in building their applications.
Much of the previous research in location-enhanced computing has been limited to relatively small spaces, such as university campuses. Place Lab is exploring the possibility of a system that's ubiquitous to an entire city, and potentially, that spans the globe. To enable truly global deployment of Place Lab, researchers are enlisting the help of professional developers and amateur enthusiasts in identifying Wi-Fi access points throughout the world.
Place Lab researchers have designed the technology to be privacy-observant. Place Lab allows clients to determine their location entirely privately without constant interaction with a central service. Researchers are conducting user studies to gain a better understanding of what information people are willing to share about their location, with whom, and when. The team is also working on an application that explicitly exposes privacy issues to determine how to overcome them.
Conclusion
The growth and success of the Internet have spawned new problems and challenges, some having to do with the basic architecture of the Internet. The PlanetLab project is laying the groundwork for the evolution of the next-generation Internet with an improved, more secure architecture. This new Internet will supplement existing features and functionality with computational capability embedded throughout the network, enabling the delivery of services at any location in the network.
PlanetLab is Intel's model of how the Internet will evolve, as well as a platform for innovation, enabling researchers to deploy and test a variety of planetary-scale services. Intel was an early contributor to the project and continues to be an active participant.
In addition to participating in PlanetLab, Intel is engaged in a variety of research efforts to enable the evolution of the next-generation Internet. Intel also provides technology and products to enable a range of low-power, lightweight mobile and wireless devices that will bring Internet access to new groups of users throughout the world.
The Internet will inevitably be transformed into a new and more useful entity for computing and communications, as current problems are solved, new services are developed, and a more secure architecture evolves. Only the timing of the transformation remains unclear. Over the next two or three years, changes in the Internet may be indistinguishable to the majority of users, although its problems (especially security-related problems) may become more obvious. Within a longer time frame- perhaps five years and beyond, as the next-generation Internet emerges-users will have access to a variety of new services, delivered in news ways, using virtual machine technology, in an environment that is far more secure than that of the current Internet.
Team
Members of the PlanetLab project team within Intel:
PlanetLab Overview: (WMV file, 18.9 MB) Video features Mic Bowman, Mothy Roscoe and Brent Chun describing global testbed and framework of Planetlab. Testimonials from Dr. Vivek Pai, Princeton and Dr. Steve Hand, Cambridge.
Place Lab: (WMV file, 22.9 MB) Video features Anthony LaMarca, Yatin Chawathe and Prof. David McDonald, University of Washington. Overview of location enhanced computing.
Numerous research projects related to PlanetLab are underway. Four representative projects include
Netbait, the worm detection service described in the opening paragraph;
OceanStore, a global storage service;
ScriptRoute, a distributed Internet debugging and
measurement tool; and CoDeeN, a content distribution service used by PlanetLab
to distribute software updates to nodes across the network. More than a dozen other
projects are slated for PlanetLab deployment in the near future.
Visit the PlanetLab site for more information or to join in the fun of developing the next Internet.
