JISC Internet2 Pricing Networks Project (with the M3i project, Hewlett Packard, British Telecom and the University of California, Berkeley and UKERNA) First 6 month report A Bouch, J Crowcroft, A Sasse, et al This report starts with the 6 month diary of the project so far, then reports the plans for the next 6 months of work Project Diary ------------- September: ---------- I. Deliverables -(need to check proposal where these fit):- a) QofIS/M3I meting in berlin - The Qofis workshop in at GMD Workshop in Berlin held a minitrack on Internet charging on Sept 27. Angela Sasse gave a presentation summarising results of empirical work on user preferences for quality and charging conducted at UCL over the past 3 years, and plans for a trials and evaluation of congestion based-charging under this Internet-2 project. Angela Sasse was a member of the final panel, and Jon Crowcroft made the closing remarks on the day. The talk generated interest from HP (Joern Altmann, who works on the INDEX project), and collaboration was agreed with him and Bob Briscoe, BT. b) we've been looking at the possible communities we could experiment on We have established contact with the High Energy Physics group (via Peter Clark at UCL) who are interested in collaborating. c) we(I) are trying to figure out what is the minimal setup to provide the M3I framework, without havin to have all the details deployed i) measurement ii) access control (e.g. web proxy desgn and interaction with routes) iii) accounting framework II. Next month Work: a) have advertised job Anna Bouch will start working part-time on the project this month, and will be developing the framework for evaluation and data collection instruments (interview schedules and questionnaires). We will hire additional staff to work on the user-response side once data collection is ready to commence. b) will talk to ukerna about what Internet2 US end can help with as c) need to specify some congested paths and meter points (e.g. New York end of uk-us link d) need to talk to UKERNA operational folks about loss monitoring, netflow accounting (level of detail and so on) and ECN deployability, and maybe deployment of HP smart internet usage s/w. e) most important, start thinking about experiment design details! aside:- iii. UKERNA has some US contacts who can help with their end, and is gong to an US internet2 meeeting in end of october - if we can draft some ideas for the "kind" of things we might need, they can present... iv. UKERNA asked when we might go (or send someone) to an interne2 US meeting - sometime after xmas, we'd try to sort out probably 2 (try to combine with some other things like IETF) - Angela Sasse will be going to LA early Nov. and Minneapolis early Dec., in case it's anywhere near that. October: -------- 1. Anna Bouch is now working full-time Project, and is carrying out user level requirements capture etc. 2. Gave detailed exposition of technical requirement and deployment plan to UKERNA for discussion with US Internet 2, and for requirement on ULCC for UK-US link allocation. Somewhat dismayed to hear that UKERNA may not be able to deliver their part of the service - this means that most Internet2 collaborations will be non technically realizable (e.g. the immersive VR one at UCL will suffer from jitter when working with MIT and UNC - we will not be able to offer differentiation for our priced users). 3. Made contact with Computing Support Edinburgh (Graeme Wood), who have a differential charging scheme for Web access (cached vs. direct access) They agreed to let us distribute a short description of projecta and recruitment call for users to participate in face=to=face interviews and focus groups We are currently preparing these and will send them to Edinburgh within the next couple of weeks 4. We're interviewing in the next week for the programmer/technician 5. Plan to discuss work with Cambridge Uni. experts on pricing 6. Will seek to meet with UKERNA for clarifications on 2. in November, as this will impact our project(s) - may need to redesign technical side altogether. JISC will have to be informed of this problem that UKERNA have in delivering capabilities similar to those enjoyed in the US. 7. Work on 3 can continue despite this problem. November: --------- i. we are on track ii. we have a meeting with M3I next week at BT labs - we'll discuss the experimental plan then iii. we are happy to take part in the networkshop (mar 26...?) - we can either present an experimental overview, or even run a track on pricing...depends on interest and time avaialble. iv) we'd like to be paert of the may internet II event, of course v) we need to contact the heads of computing services - Angela will compose a message and send to Rina, so it can be injected to Ukerna's lists via officially sanctioned path! We will also be happy to contrbute to the next UKERNA newsletter in this regard, and in general about the project! vi) the info about the proposals/discussions on the UK-US/Abiline upgrade ideas was most useful - we'd favour (especially for the VR work) the direct link from UKERNAs NY router to Abiline, of course! vii) we would like to stay informed of the JISC discussions on possible extension to the Internet 2 scheme - As well as having several ideas for followup/continuing studies on pricing, Angela has some new ideas on users and security which we think would be highly relevant there. December: --------- 1. We have carried out a literature review of previous work into users' perceptions of charging for Internet QoS. From this review we have identified central questions important for I2 to answer. Examples are: 1.1 Research with telephony shows that users have a strong desire for simple pricing schemes How can we map this requirement to data networks, where QoS is far more dynamic in the case of congestion pricing? We must understand which QoS parameters are important to people in times of congestion (critical period). These might be different than in times of non-congestion. 1.2 The INDEX project has succesfully used personalised support tools for in the form of intelligent agents to manage charging for users We currently do not know what actions users need control over, and which they expect to be encapsulated. 2. Drafts for research to address these questions have been developed and refined. Amoung other research, we plan to: 2.1. Conduct a preliminary questionnaire to assess users' reactions to charging mechanisms and the further investigate the relationship between price and QoS. 2.2. Conduct a focus group, ideally with the same sample, to explore these issues more deeply and relate them to previous work on charging done at UCL 2.3. Conduct and experiment: Our idea is to set up a game where several users try and access a particular site to gain some information. The game could be based on an adventure game where they should get different pieces of information to get to the end. They are in competition to get to the site. This creates congestion (we can simulate different congestion to users, it doesn't have to be real). We should give them a real-money budget. The strategies we program into the game are based on our findings from our focus groups. We can log their behavior (e.g. how long did they wait, how much did they sacrifice). We should also construct a dynamic comment box for them to enter how they feel about the interaction. The game should take several visits to complete. 3. We have negotiated access to department reps and individual end-users at the University of Edinburgh, which has a 2-tier charging system for Web access. We will refine our interview questions and questionnaires in consultation with Computer Services staff there (Bill Byers and Graeme Wood), who have kindly agreed to help us gain access. We hope to have the detailed experimental plan before end of calendar year (on or before Q1 deliverable date). Anna Bouch is gong to present some of the work at the Jan workshop described at: http://www.ncne.nlanr.net/news/workshop/2001/010128/mtgnotice.html January: -------- 1. Jon Crowcroft attended the Tequila european (IST) project workshop on Service Level Specificaiton, at Terena in AMsterdam, and gave a talk about the pricing project to the attendees - Also present was Ben Teitelbaum from the QoS WOrking Group of the US end of the Internet II program - we had discussions on the desired outcomes of the JISC Project. Several interesting problems came up about the assymetry of the link and the initiation of a traffic flow - for example, it is not obvious whether a TCP connection with data mainly from US to UK is caused by the receiver or the sender. 2. The experiment design is nearly complete. 3. Staff - we currently still need to hire a 2nd web/tech programmer person to complement the human factors expertise - this is not a problem yet, but will be importasnt after next month. 4. It is not clear how we are going to carry out the actual experiments unless there is some form of managed/assured bandwidth on the UK-US link. Recent feedback from UKERNA was not encouraging here. We can carry out pure web based experiemnts (as discussed with Glasgow and others), but these do not reflect true network costs, so do not have the general interest of the community - for example, the US partners are interested in actual throughput, whether it is for VOIP or GRID use, as well as delay influences on user "willingness-to-pay". Febuary: -------- Negotiatiing via UKERNA wuith NOSC about enabling either ECN, or loss metering at the UK and US end of part of the transatlantic capacity - not getting ery far with this (NOSC aren't really aware of what is needed, or even what ECN is!). Plan B (fall back) is to do the whole thing web based and distributed... March: ------ The aim of the research is to investigate users' mental models with regard to pricing with focus on congestion pricing. One of the functions of these models is to describe the criteria for acceptance of pricing mechanisms. Existent and proposed pricing models can therefore be evaluated against these models. To date we have: 1. Conducted a large scale questionnaire with users with difference network experience. The results have shown us that services should be seen as objects associted with particular commercial brands. As objects are static, payment for them should be a one-off, and dictated by the user's task. 2. Conducted focus groups to test the assumptions of emerging services, such as 3G. The results are beginning to show that there is a complex socio-cultural structure that dictates users' expectancies of services, and how much they will pay for them, and how. We intend to conduct groups with users from different cultures to test this theory. The aim is then to discover a generic users' model and disaggregate this into culturally-specific models/markets. Project Plans for next 6 months ------------------------------- We will attend the UKERNA Internet2 Project day (was to be in May, now september). We will host a pricing workshop at UCL June 18/19 - M3I partners and UKERNA and US Internet2 folks to be invited. M3I experimental plan to be expanded to include UKERNA work. Papers in first 6 months reporting work: The papers we had accepted were for IWQoS and It-COM'01: 1. Bouch A., & Sasse M.A. (2001). Why value is everything: A user-centred approach to network Quality of Service and pricing. http://www.cs.ucl.ac.uk/staff/A.Bouch/BouchSasseCRC.pdf http://www.cs.ucl.ac.uk/staff/A.Bouch/Bouch&SasseCRC.doc 2. The abstract from ITCOM (full paper is not yet written): Bouch, A., & Sasse, M.A. (2001). Not all bits have equal value. To be presented at IT-COM'01. Denver, Colorado, August 19-24, 2001. http://www.cs.ucl.ac.uk/staff/A.Bouch/SpieIT301-Bouch.doc We've submitted more recent empirical work for review at ICQT'01. Anna BOuch has starting to build the simulations that test pricing schemes preferred by users, as discovered in the ICQT work.