PlanetLab Everywhere - "PlanetLab - future directions"

Outline

The opportunity PlanetLab provides for real-world experimentation at an unprecedented scale has sparked an explosion of research activity, with hundreds of active projects transmitting an average of 1 terabyte of data a day on the current system. At present, PlanetLab consists of 580 machines hosted at 275 sites in 30 countries. Most sites only host two or three machines. A single operations center hosted at PlanetLab Central (Princeton) is responsible for the management of all these machines. While PlanetLab has been a huge success by many measures, it continues to be limited in scale in terms of its size and the number of users it reaches.

This workshop will explore several new directions to expand PlanetLab's user base to literally every researcher or student who wants to conduct a wide-area experiment and expand its hardware base to every machine on the planet. The following will be topics at this workshop:

• Clusters meet PlanetLab
  PlanetLab integrated with cluster environments such as Rocks, Oscar, etc. Such integration will drastically increase the amount of hardware resources a site can devote to PlanetLab. There are a number of challenges in this domain: cluster resource management, security, data persistence, disk less node support, etc.
• Private PlanetLab
  The PlanetLab management infrastructure has served well to manage hundreds of widely distributed machines providing value added services. This can also be of use to large institutions or corporations (e.g., HP, Intel, etc.) to manage widely distributed machines via a Private PlanetLab infrastructure. For example, services such as end-host multicast, network mapping via scriptroute, and content distribution networks could be deployed throughout an institution. Challenges in this domain are all related to scale and include: improved monitoring of systems, further automation of remote management tasks, and better visualization of overall system status.
• PlanetLab and Machine Virtualization
  PlanetLab currently uses Linux Vserver support to virtualizes slices at the operating system level. This was primarily a pragmatic choice to bootstrap PlanetLab. Other machine virtualization systems (e.g., Intel VT, VMware, Xen, MSFT Virtual Server, etc.) are now becoming viable options, particularly given the momentum behind them. They open the door to better remote management of machines, improved availability, and even "piggy-backing" PlanetLab on regular machines at home or the office when these are idle. Challenges in this domain include: resource management, security, networking, and accountability.
• Federating PlanetLab
  PlanetLab is growing globally which introduces multiple barriers for having a single Operations center in Princeton manage all nodes. The two main barriers are: 1) vast time zone differences and 2) native language issues. It can take a long time to resolve a hardware/software support issue or clarifying security/abuse incidences. In a recent Turing award lecture by Bob Kahn and Vint Cerf, they noted that the success of the Internet was federating lots of independent ISPs. For this reason, PlanetLab Central is actively pursuing to decentralize operations by federating multiple independent PlanetLab installations into one. Challenges revolve around cross-administrative issues such as security, resource allocation, naming, etc.

The goal of this meeting is to bring together users, developers, and researches interested in these topics and develop plans to independently or collaboratively move forward with projects to develop working prototypes in each area. Further it is intended to let participants discuss technical problems and exchange their experiences to improve the quality of cooperation in the PlanetLab community.

The meeting is supported by:

    PlanetLab project - Princeton University

   Evergrow EU project

            MICS - Mobile Information and Communication Systems (Swiss National Center of Cometence in Research)

Structure

We suggest a mix of talks, panel discussions, case study presentations, experience reports, demonstrators, and "bird-of-a-feather" sessions. Participants are invited to submit proposals until Oct 13, 2005.

Agenda

Day 1

When	What	Where
10:00	Opening: welcome + schedule overview for the "work" part of the workshop (Manfred Hauswirth)	INJ 218
10:15	Overview of where we are now and where PL is going (Larry Peterson)	INJ 218
11:00	Clusters meet PlanetLab @ HUJI (Danny Bickson)	INJ 218
11:20	Private PlanetLab @ EPFL (Martin Rajman)	INJ 218
11:40	PlanetLab, Xen, and clusters (Marc E. Fiuczynski)	INJ 218
12:00	Tycoon cluster/federation (Kevin Lai)	INJ 218
12:20	Lunch (for folks to introduce themselves and briefly discuss what their overall interests are in PL, and in what way they'd like to participate, collaborate, etc.)	BC cafeteria
13:40	Federating PlanetLab (Aaron Klingaman)	INJ 218
14:00	Discussion on federation	INJ 218
14:20	Large scale challenges for PL (Rick McGeer)	INJ 218
14:40	EU OneLab collaboration (Thierry Parmentelat)	INJ 218
15:00	DIMES measurements (Yuval Shavitt)	INJ 218
15:20	ETOMIC measurements (Gábor Vattay)	INJ 218
15:40	Coffee break	INJ 218
16:00	Formation of working groups	INJ 218
18:00	End of day 1
19:00	Dinner + informal gathering @ some casual location

Day 2

When	What	Where
9:00	Parallel working group sessions Challenges (or reasons for) Private/Enterprise PlanetLab	BC 410
9:00	Parallel working group sessions Challenges in Federating PlanetLab	INJ 218
9:00	Parallel working group sessions Challenges in marrying Clusters/Grid with PlanetLab	BC 129
10:30	Coffee break	INJ 218
10:45	Parallel working group sessions (continued)
12:20	Lunch	BC cafeteria
14:00	Working group presentations	INJ 218
15:00	Wrap up the workshop + define next steps and research agenda	INJ 218
16:00	End of day 2	INJ 218

Organization Committe

Local organization:

• Manfred Hauswirth (EPFL, Switzerland)
• Renault John (EPFL, Switzerland)

Organizing committee:

• Danny Bickson (HUJI)
• Manfred Hauswirth (EPFL, Switzerland)
• Marc E. Fiuczynski (Princeton, USA)
• Aaron Klingaman (PlanetLab Consultant, USA)
• Thierry Parmentelat (INRIA, France)

Participants

Kevin Glynn (UCL, Belgium)

Pierre Reinbold (UCL, Belgium)

Maoke CHEN (CERNET Center, China)

Thierry Parmentelat (INRIA, France)

Thomas Neumann (Max-Planck Institute, Germany)

Sebastian Michel (Max-Planck Institute, Germany)

Antonatos Spiros (FORTH, Greece)

Gábor Vattay (Eötvös University, Hungary)

Danny Bickson (HUJI, Israel)

Assaf Ben David (HUJI, Israel)

Tom Alsberg (HUJI, Israel)

Elliot Jaffe (HUJI, Israel)

Danny Shaked (TAU, Isreal)

Ariel Biener (TAU, Isreal)

Yuval Shavitt (TAU, Isreal)

Akihiro NAKAO (The university of Tokyo, Japan)

Yoichi Shinoda (JAIST, Japan)

Erik Aurell (KTH, Sweden)

Miika Tuisku (CERN, Switzerland)

Tapio Niemi (CERN, Switzerland)

Mikko Pitkänen (CERN, Switzerland)

Predrag Buncic (CERN, Switzerland)

Karl Aberer (EPFL, Switzerland)

Manfred Hauswirth (EPFL, Switzerland)

Martin Rajman (EPFL, Switzerland)

Ivana Podnar (EPFL, Switzerland)

Renault John (EPFL, Switzerland)

Anwitaman Datta (EPFL, Switzerland)

Michela Spada (EPFL, Switzerland)

Francis Lapique (EPFL, Switzerland)

Hung Nguyen (EPFL, Switzerland)

Pascal Felber (Univ. of Neuchatel, Switzerland)

Marc Schiely (Univ. of Neuchatel, Switzerland)

Lorenzo Leonini (Univ. of Neuchatel, Switzerland)

Pablo Rodriguez Rodriguez (MS Research, UK)

Marcin Pilarski (Warsaw Univeristy of Technology & Polish Telecom, Poland)

Larry Peterson (Princeton, USA)

Marc E. Fiuczynski (Princeton, USA)

Aaron Klingaman (Princeton, USA)

Andy Bavier (Princeton, USA)

Tom Anderson (UW, USA)

Rick McGeer (HP, USA)

Kevin Lai (HP, USA)

Sheri Brodeur (HP, USA)

Timothy Roscoe (Intel Research - Berkeley, USA)

Hotels

Hotel Continental
http://www.hotelcontinental.ch/
Place de la Gare 2
1003 Lausanne
(directly at the main train station)
Tel: +41-21-321 88 00
Fax: +41-21-321 88 01
Rates: single - CHF 130.20, double CHF 204.20 (1 Euro is approx. 1.5 CHF)
reservation@hotelcontinental.ch

Hotel Aulac
http://www.aulac.ch/ Place de la Navigation 4
1006 Lausanne
(directly at Lake Geneva)
Tel: +41-21-613 15 00
Fax: +41-21-613 15 15
Price : single - CHF 122.00, double CHF 144.00 (1 Euro is approx. 1.5 CHF)
aulac@cdmgroup.ch

We pre-reserved 15 rooms in each of the two hotels. You have to book before October 17, 2005 and mention "PlanetLab Everywhere" to get the special rates.

If you want to check for further hotels or want to get information about Lausanne, please visit the Lausanne Tourism web site at http://www.lausanne-tourisme.ch/.

Venue

Getting to Lausanne

For comprehensive information please check the web site of the Swiss railways at http://www.sbb.ch/en/index.htm.

Getting to EPFL

Please be careful when you buy your ticket for the metro: You need the exact amount in Swiss Francs (coins) as the wending machine does not return change (some of the machines accept Euros). The price for the ticket is CHF 2.40,-

• Timetable M1 (TSOL)
• Timetable M2

From the EPFL metro stop to the meeting

• Get off M1 (TSOL) and take the exit in the direction you arrived. Go down and then turn left underneath the rails.
• Walk up the stairs and you'll be between 2 large buildings.
• Walk straight ahead until you see EPFL's large main square to the left!
• Follow "La Diagonale" (guess why it is called that way :-) slightly to the right to the end until you hit the computer science buildings (IN)
• Enter the building and go to the second floor, then you will have to take the INJ corridor.
• The meeting is in the INJ 218 which is the room at the end of the INJ section, on your left.

Facilities

Basic infrastructure

Please do not forget to bring a converter for your electricity plug!
Check http://kropla.com/electric.htm for comprehensive infos. Switzerland uses this kind of plugs.

Internet access

It is NOT possible to provide Internet access via Ethernet (cable).

You MUST bring a WLAN card if you want to have Internet access.