Minutes and Notes

A summary of the 8th Joint Techs Workshop, cosponsored by Internet2 and NLANR, is at: http://www.internet2.edu/events/may-2001-joint-techs/

• Ana Preston reviewed the challenges that were delineated earlier (see earlier presentation at http://ncne.nlanr.net/training/techs/2001/0514/presentations/200105-preston1_fil es/v3_document.htm). She also mentioned the workshop to be held at the end of the Fall Internet2 meeting in Austin, TX. Mark your calendars for Oct. 4-5, 2001; see upcoming note to the list.
  
  
• Gregg Goddard, U of Florida and Ana have begun creation of an FAQ for p2p. It is found at: nersp.nerdc.ufl.edu/~gregg/projects/p2p-faq.txt This *very* rough draft is a first step at consolidating some of the topics/faqs that have circulated in the p2p list (listserv.utk.edu/archives/p2p.html). Contributions and contributors are welcome. Contact either Gregg (gregg@nersp.nerdc.ufl.edu) or Ana (apreston@utk.edu) about content or participation
  

Case Studies and Presentations

• Univ. of TN - chosing not to do anything at all and give bandwidth away; experiences. (Ana Preston) (see presentation at volnet.utk.edu/p2p) Ana talked about p2p on the U of Tennessee-Knoxville campus. UTK gave the dorms their own DS3. A dialogue was initiated by the students halfway through the semester. The students indicated that they would organize an educational campaign and that they wanted to identify the bandwidth hogs, especially since they had seen that their own DS-3 was saturated. The students also indicated that they needed help in being responsible users of the network.
  
  Discussion item: Steve Wallace told us that Indiana U provided a DS3 to the residence halls for their exclusive use. A discussion ensued regarding rate limiting and having the students pay for any extra bandwidth used
  
  
• Univ. Wisconsin-Madison - FlowScan and rate limiting adventures (Michael Hare) (see presentation at volnet.utk.edu/p2p)
• Michael mentioned that Napster constituted 50% of their outbound traffic.
• Michael gave a very interesting overview of how Univ. of Wisconsin has tackled rate limiting and p2p applications on their campus
• An overview of FlowScan and recent updates to it was provided
• FlowScan is a network analysis and reporting tool. It processes IP flows recorded cflowd-format raw flow files and reports on what it finds. For more information, please go to http://net.doit.wisc.edu/~plonka/FlowScan/



• Indiana University - Ramblings Regarding P2P (Steve Wallace)
• Steve mentioned that it was the 10th anniversary of the first distribution of browser software from CERN
• He indicated that p2p is a distributed intelligence approach to harnessing the power of devices. Napster and other p2p tools provided a wake-up call to networking personnel. Users had bandwidth-hungry applications (hadn't we in the I2 community been waiting for that?). Dorm rooms could essentially be turned into music sharing entities for the world. Indiana U has a tremendous amount of outbound traffic to off-campus addresses. IU realized that the model previously used (essentially unlimited bandwidth for all) was not a sustainable model. They looked at three alternatives:
  
  1. rate limiting to the dorms
  2. blocking certain ports
  3. separate pipe to the dorm
• Rate limiting p2p applications b/c they may be infringing, leads to slippery slope. Effort should be in managing these resources and their use.
• They look at networking as a resource much like the telephone or electricity. While IU has an interest in the subsidy of the resource, they aren't interested in being involved in the content identification of the resource
• Steve predicted that all traffic will soon look like web traffic and it won't be possible to determine whether a user is utilizing a p2p application or checking the University's site for information. - Therefore, IU is putting efforts into reasonably allocating the resource that they have available
• Steve also cited a local broadband provider that is using a simple plan-they employ increasingly tighter constraints based on usage and let the consumer pay for resources. He made the point that the efforts of the provider shouldn't on p2p applications but rather, consumption of resources. It is a slippery slope to attempt to rate limit p2p applications because there may be copyright violations associated with them. During the discussions surrounding the lawsuit with IU, there appeared to be a perception that IU was assisting users in transferring copyrighted material. The music industry is concerned about this now. The movie industry will be concerned about this soon. Napster and other p2p applications are easy to use-it is easy to find songs and it is easy to obtain songs because of the directory services centralized at Napster. A service that provides easy search and retrieval of music and is also scalable will be difficult to stop. Gnutella does not scale and Napster also has scaling issues associated with islands of users. Any service that allows users to use substrings to search for song titles will not scale.
• Most p2p applications have focused on distributed computing. However, distributed intelligence is a p2p application that bears investigation. Boeing is investigating the use of intelligence in planes for air traffic control. Their system would allow planes flying into all but the most congested airports to be controlled by Boeing satellites with intelligence in the planes. This would allow the planes to fly in the most direct airspace, saving fuel and, perhaps, time.
• Steve ended his ramblings by mentioning other interesting aspects of p2p technology including computing grids, weather, toys and games, and PDAs


Discussion: Jerry Sobieski made the comment that digital content is moved easily. Steve Wallace added that the main purpose of PCs is to allow the user to view content. With specific devices for specific purposes, it would be possible to limit what is available to users. Viewing devices present a problem that is solved by having hard drives that limit access to specific things. Specific devices will become popular when a way to make money by offering them is determined. Ana asked what academia can bring to the table

• P2P apps in state networks and sponsored participation implications (James Deaton, OneNet, Oklahoma)
  
  Higher ed residence hall rooms and Napster present an interesting combination. They have used Flowscan to determine what constitutes various types of traffic on the campuses. Napster appeared to constitute 50% of the traffic. A decision was made to educate the University and community college communities regarding use of bandwidth. He made the point that use of bandwidth affects the entire state rather than just the local campus. Oklahoma schools are heavy users of H.323 video and they found that Napster was starving out H.323. They decided that they needed a solution first and to institute a policy second. They started by using CAR. They are looking for solutions that will work for the entire state. They are identifying software and researching trends in the traffic.
  
  
• Univ. Oregon -- p2p application management at the Univ. of Oregon (Joe St Sauver, Univ. Oregon)
  
  see Joe's previous presentation, "Oregon GigaPop - Traffic Characterization" at: http://ncne.nlanr.net/training/techs/2001/0514/presentations/200105-sauver1_file s/v3_document.htm
  
  Joe mentioned that they are seeing applications other than just Napster and Gnutella, including non-IP file sharing on local subnets. His institution has constrained commodity bandwidth and they visit with and encourage users to be good network neighbors by using the network responsibly. They have a local AUP that works in conjunction with a state AUP as well as a state law. They perform no filtering.
  
  
• Question via email : How is tracking done?
  
  Tracking is done using cflowd. They are using an application that was written locally. IU does something similar. The usage seen is much like modem use -- few use a great deal, some use moderately and many use sparingly.
  
  
• U of Texas
  
  check out muti host traffic grapher (think of MRTG) Mhtg.the.net
  
  
• University of Florida -- Dan Miller
  
  Dan began by indicating that the computing group at the University of Florida chose the path of least resistance. They did nothing and waited for their administration to make a decision regarding p2p use. The administration did not act. When two DS3s became full, they decided to rate limit on ports. They didn't announce that they were going to do this-they just took action. They are using CAR and waiting for CAR to limiting to be adjusted by time of day. They are also buying more bandwidth. They treat their housing users separately but they are not segregated.
  
  
• Univ. of Utah -- Gnutella and creating a p2p application for security purposes (Dave Packham)
  
• Excellent overview on how Gnutella works and a view into the protocol and how it could be made more efficient
• Dave described a tool to determine who is using Gnutella. He indicated that he believes that Bearshare and Limewire are the best network clones currently available
  
  
• AS Path, Chris Rapier, PSC
  
  Paul talked about how, last fall, Iowa State chose to rate limit all traffic. They didn't care what the traffic was, they just rate limited it. He expressed concerned for the September effect during the fall of 2001. He has found that traffic doubles from April to September and then tends to remain constant during the year. During the fall of 2001, they will separate residence hall traffic from the rest of campus traffic.
  
  
• Iowa State -- Paul ??
  
  Paul talked about how, last fall, Iowa State chose to rate limit all traffic. They didn't care what the traffic was, they just rate limited it. He expressed concerned for the September effect during the fall of 2001. He has found that traffic doubles from April to September and then tends to remain constant during the year. During the fall of 2001, they will separate residence hall traffic from the rest of campus traffic.

Thanks to all of our presenters and all of you who attended the session. Please contact Ana (apreston@utk.edu) or Linda (lroos@oar.net) if you have any questions.