PlanetLab has proven to be an invaluable platform for learning about network-wide phenomena, creating new network protocols, evaluating new and existing network services, gaining experience with network systems running at a global scale, and in the end, deploying novel network services that enhance the capabilities of the Internet.

Impact on Research

PlanetLab has become an essential tool for network and distributed systems research. Researchers that make claims about protocols and services running on the Internet use PlanetLab to demonstrate how their designs hold up under realistic network conditions. We have compiled a partial bibliography of papers that have taken advantage of PlanetLab. The bibliography includes nearly 200 citations, the vast majority of which have been published at the very best networking and distributed system venues: SIGCOMM, SOSP, OSDI, NSDI, INFOCOMM, Usenix, EuroSys, IPTPS, IMC, and HotNets. It is also noteworthy that the papers have been evenly distributed over time, with papers published at the same rate in 2007-08 as in 2004-06.

It is difficult to quantify the broader impact of this research, but there is strong anecdotal evidence that research leveraging PlanetLab is having a far-reaching impact. A small sampling includes:

• The iPlane and Hubble network measurements projects have been a valuable resource for the networking research community, with more than twenty research projects using the structured network topology information produced by the systems.  Further, RIPE-NCC, the organization that provides technical and administrative support for network operations in Europe, views Hubble as a valuable trouble-shooting resource and is working towards re-implementing Hubble and adding it to their suite of network management services.

• BitTyrant is a highly optimized and strategic BitTorrent client whose development was aided by extensive experimentation on PlanetLab.  BitTyrant was publicly released about a year ago and since then has been downloaded by more than a million users.

• PlanetLab was used for experimentation with localizing optimizations for peer-to-peer systems.  Out of this work came a new proposal, P4P, an interface that allows ISPs and peer-to-peer systems to coordinate and optimize for both network-level efficiency and application-level performance.  The P4P technology is currently being evaluated by five major ISPs (Verizon, Telefonica, AT&T, Comcast, and Bell-Canada) on real-world swarms, with initial experiments promising substantial improvements for a wide variety of network performance objectives.

• The PlanetLab deployment of the Coral, CoDeeN, and CoBlitz content distribution networks carry 6+TB of data each day on behalf of approximately 1M end clients. This has proven to be a valuable service for many non-profits with content to distribute.  For example, CoBlitz is used by the UChannel to deliver videos of public policy lectures, by CiteSeer to deliver technical papers, and by Fedora to distribute Linux software releases. Similarly, Coral is used by SpamAssassin and AdblockPlus (FireFox) to distribute rule set updates, by websites overwhelmed with visitors from viral news communities like Slashdot and Digg to mirror their linked content, and by individuals to distribute breaking news such as pictures and videos of the Katrina disaster or the Asian Tsunami.

• The CoBlitz content delivery system, developed and hardened on PlanetLab, is being evaluated by commercial Telcos as a technology for delivering both traditional web content and streaming video. In one such trial, CoBlitz was deployed on the Polish Telecom network, and used to deliver video of their presidential debates (Fall 2007), which at the time was the largest on-line event in Poland's history.

• The CoDoNS and CobWeb services on PlanetLab ran Andrew Tannenbaum's highly popular website (electoral-vote.com) during the 2006 elections. The website, which received 700,000 visitors per day, had previously been the target of denial-of-service attacks, but was protected by these resilient and self-organizing distributed systems.

• PlanetLab has provided a level playing field for evaluating alternative Distributed Hash Table (DHT) protocols, resulting in a robust implementation---OpenDHT---that synthesizes ideas from many of these original designs. OpenDHT, in turn, has been used to build Adeona, an open-source, privacy-preserving device tracking system for stolen or lost laptops.  Adeona has received significant commercial and public interest, with more than 20K downloads over the last month.

Impact on Teaching

PlanetLab has hosted over 4700 users in its six-year history, approximately 3700 of which have been students. Whether these students are working on their PhD research or doing course assignments, they are gaining valuable experience with network systems running in at a global scale---experience coping with transient failures, differences in connectivity cliques, variations in latency and bandwidth, abuses inflicted by real users (some of which are malicious), and so on.

More narrowly, a set of graduate and undergraduate courses have been designed to take advantage of PlanetLab. Links to over 40 of these classes are available at http://www.planet-lab.org/courseware .

Multiplicative Effect

Because PlanetLab is now a software distribution as well as a public deployment, there is an opportunity for explosive growth in the network resources that can be made available to researchers, and used to run the innovative network services they are creating. Effectively federating these autonomous deployments remains a significant challenge, but the deployments already under way point to an unprecedented opportunity.

One such deployment is VINI, which extends the PlanetLab code base to exploit layer-2 circuits, and has been deployed on the National Lambda Rail (NLR) and Internet2's NewNet backbone. Another is MeasurementLab, which widely deploys servers that can be probed from the edge of the Internet. Similar deployments are as follows.

International Deployments

Catalyzed by the GENI initiative in the US, which was itself inspired by PlanetLab, there are a host of regional efforts to deploy PlanetLab-based network substrates. At the time of this writing, the PlanetLab team at Princeton is working with the following international groups:

• The OneLab2 project in Europe, funded under the FIRE initiative, operates PlanetLab-Europe, an independent entity that federates with the public PlanetLab.

• The CoreLab project in Japan, with participation and support from NICT and WIDE, is in the early stages of operating PlanetLab-Japan, also an independent entity that will federate with the public PlanetLab.

• The G-Lab project involving a consortium of five German universities and with funding from the German government, is building a PlanetLab-based layer-2 network, modeled after VINI.

• Similar but less mature efforts are also underway in South Korea, China and Brazil.

We are planning the deployment of a lightweight version of a PlanetLab node, called PlanetBridge, into neighborhoods, villages, schools, and homes around the world. Our partners are in Ghana (Kokrobitey Institute & EcoBand Inc. & KACE), Brazil (Universidade Federale Fluminese / RUCA2), India (National Institute of Technology Karnataka), British Columbia (University of Victoria), South Africa (Meraka Institute), Belize (Imaging the World), and Costa Rica (CruTec).

Commercial Deployments

Corporations, in many cases Telcos, are deploying PlanetLab on their internal infrastructures.  Examples include AT&T, Polish Telecom, BT, and Intel. There are three general motivations. One is to provide a "network sandbox'' for the company's researchers, who often have less latitude exploring new ideas on their corporate networks than academic researchers have on the Public PlanetLab. A second is to evaluate technology developed on the Public PlanetLab for commercial viability. This includes both network services running on PlanetLab and PlanetLab itself as a programmable platform for delivering future services. A third is to exploit a PlanetLab capability for an existing corporate need, such as monitoring Internet and intranet activity from multiple vantage points.

Note: If you have other impact bullets and you would like to add to this page, please send an e-mail to info@planet-lab.org