Distribute materials on the Common Solution Group framework forcollecting traffic performance measurements. (Monk / Almes) Guy Almes, Advanced Network Services, almes@advanced.org
Steven Bakker, DANTE-UK, steven@dante.org.uk
Vincent Berkhart, DANTE-UK, vincent@dante.net
Erik-Jan Bos, SURFnet-Netherlands, bos@surfnet.nl
Les Cottrell, DOE/SLAC-US, cottrell@slac.stanford.edu
John Dyer, Terena-EC, dyer@terena.nl
Kilnam Chon, KAIST-Korea, chon@cosmos.kaist.ac.kr
Shigeki Goto, Waseda Univ.-Japan, goto@goto.info.waseda.ac.jp
Donal Hanna, Netskills-UK, donal.hanna@ncl.ac.uk
Doug Hughes, Canarie-Canada, dihughes@canarie.ca
Harhisa Ishida, Univ. of Tokyo-Japan, ishida@u-tokyo.ac.jp
Peter Kirstein, Univ. College London-UK, kirstein@cs.ucl.ac.uk
Kazunori Konishi, KDD-Japan, konishi@lab.kdd.co.jp
Steven Ladouceur, Bell Canada, sladouceur@on.bell.ca
Thomas Lenggenhager, Switch-Switzerland, lenggenhager@switch.ch
Mark Luker, National Science Foundation-US, mluker@nsf.gov
John Martin, Terena-EC, martin@terena.nl
Tracie Monk, DynCorp/FNC-US, tmonk@snap.org
Peter Taylor, Mailbase, Newcastle Univ.-UK, peter.taylor@ncl.ac.uk
Stefano Trumpy, CNUCE/CNR-Terena-Italy/EC, s.trumpy@cnuce.cnr.it
Tan Tin Wee, National Univ. of Singapore, tinwee@irdu.nus.sg
David Wasley, Univ. of California, david.wasley@ucop.edu
Walter Wiebe, Federal Networking Council-US, wwiebe@nsf.gov
II. ACTION ITEMS
Distribute materials on the Common Solution Group framework forcollecting traffic performance measurements. (Monk / Almes)
Extend participation in the CSG project to international sites – CCIRNmember networks are to report back to the mail list concerning theirintention to participate in the CSG project and any unique requirements theyhave for their sites / measurements. (All)
Terena will extend an invitation to CCIRN members to participate in theTerena Measurement Taskforce kick-off meeting in Amsterdam, January 24th.(Dyer)
John Dyer will provide some of the operational (service related)specifications used in Terena contracts for networking service.
CCIRN chairs will be invited to send technical representatives to theNSF-sponsored Measurement Workshop in March/April 1997. (Luker) After theworkshop, there will be a session to coordinate on CCIRN measurement workinggroup / international statistics collections and analysis. (Monk)
If feasible, language regarding the desirability of conductingperformance measurements may be added to NSF’s International ConnectionsProgram solicitation. (Luker)
A CCIRN Measurement working group mail list (cmeasurement@fnc.gov) will beset up. (Monk)
III. DISCUSSIONS
a. Europe - Terena
John Dyer provided an overview of the Trans-European Research and Education Networking Association (TERENA - see: http://www.terena.org) and its plans for Performance Measurement. Terena represents its members, the national research networks of Europe. Terena’s Technical Programme brings European experts together into Working Groups whose role is to assist the member networks in the introduction and development of new technology and services.
John described the goals and work plan of the proposed new Terena task force on Performance Measurements. The kick off meeting of the TF-ETM is scheduled for January 24, 1997 in Amsterdam, following the RIPE meeting. CCIRN members interested in participating should contact Ariel T. Sobelman at ariel@terena.nl.
The task force will develop a mechanism and structure for collectingstatistics of traffic loads on an end-to-end international level,measurements of delays, flow capacity and packet loss information. Thistwo-year effort will focus on topics with significance to analysis ofnetwork performance and with relevance to Internet settlements. Activitieswill include cross-European measurements of total traffic through specificlinks and the mapping of reachable destinations covered by a route, delaymeasurements, flow capacity, continental averages of hops, and packet lossrates.
Deliverables planned for this effort include:
2. North America
a. DOE/SLAC
Les Cottrell (U.S. Department of Energy/Stanford Linear Accelerator –DOE/SLAC) provided an overview of statistics gathering / analysis activitieson the Energy Science network (Esnet). ESnet is collecting ping data from200 hosts in 10 countries. Every 30 minutes each host is pinged with 11*100byte pings followed by 10*1000 byte pings. These data are being used tomeasure response, packet loss, and host unreachable rates and to developestimates of ‘predictability’ for specific traffic paths.
Les showed a series of time series graphics which depicted congestion incertain regions of the Internet infrastructure. The high packet loss in thenortheast U.S. was attributed to congestion at MAE-East. The significantimprovement in performance at MAE-West around the May 1996 time period wasattributed to the installation of the gigaswitch at the MAE-West facility.SLAC / DOE are beginning to publicly distribute beta-level tools forgathering these data. Analysis and reporting, however, is currently beingaccomplished using the SAS statistical package; results are depicted via X -Y plots.
SLAC used some of these measurement tools to evaluate the performance of alternative local providers of ISDN service. Through their analysis, they were able to determine that their current provider had a response time of more than 65 msec, and packet losses of approximately 8%. They therefore switched to a provider which demonstrated a response time of less than 30 msec and packet loss rates of less than 1.5%. Les’ slides are available at http://www.slac.stanford.edu/grp/scs/net/talk/ccirn-dec96. Addition information sources on DOE’s activities include the following:
Mark Luker of the U.S. National Science Foundation (NSF) describe the interests of NSF and other federal agencies in traffic measurements. NSF is currently funding tool development and specific traffic measurements by the National Laboratory for Applied Networking Research (NLANR, see http://www.nlanr.net) and by Merit (see http://www.ra.net/Tools). In February 1996, NSF funded the Internet Statistics and Metrics Analysis (ISMA) workshop at the San Diego Supercomputing center (SDSC). NSF anticipates funding a follow-on technical workshop in the March/April 1997 timeframe. Mark also spoke briefly on the timing of NSF’s International Connections Program solicitation. Participants agreed that it would be desirable for this solicitation to include provisions for traffic measurement.
c. NLANR
Kimberly Claffy of the National Laboratory for Applied Networking Research (NLANR) provided an overview of the status of measurement tool development in the U.S. A copy of her slides (with links to relevant web sites) is available at http://www.nlanr.net/Presentations/nanog.oct96.html. She provided an overview of a new flow tool being developed by the MCI vBNS team entitled OC3mon (see http://www.nlanr.net/NA/Oc3mon). This flow tool provides realtime data on traffic flows across links of up to 155 Mbps. The tool is expected to be upgraded to OC12 speeds in early 1997. Researchers from BBN Planet and ANS are evaluating similar flow statistics functionality on new Cisco routers (CflowD).
Kimberly also described an effort being initiated under the auspices of NLANR to provide ISPs with measurement and engineering tools and analyses. Details on this effort, entitled Cooperative Association for Internet Data and Analysis or CAIDA, are available at http://www.nlanr.et/Caida. Traffic flow visualizations developed under CAIDA are available at http://www.nlanr.net/Viz/Deunion/>. There is also close cooperation between CAIDA researchers and personnel at Merit who are engaged in developing tools as part of the Routing Arbitor project, see http://www.ra.net/Tools. An excellent source of Internet statistics and links to emerging tools is available at http://www.nlanr.net/INFO.
d. CSG
Guy Almes of Advanced Networking Solutions (ANS) and David Wasley of theUniversity of California described an initiative proposed by the CommonSolutions Group (CSG). The 23+ university comprising the CSG have agreed tosupport a measurement initiative which will deploy "beacons" at variouspoints in the Internet infrastructure – particularly traffic exchangepoints. Statistics will be collected on path specific metrics such asone-way delay, packet loss, throughput and path capacity. Both active andpassive tests will be employed, and packets will be distinguished accordingto the clouds which they traverse. The beacons will consist of "small,fast Unix boxes". Time synchronization among these boxes, probably usingGPS NTP, is viewed as one of the more complicated technical challenges.
NLANR researchers such as Jamshid Mahdavi, Matt Mathis, and Vern Paxson are participating on this project. Addition funding is being sought from NSF to support the project in addition to funding provided by the CSG universities. Guy agreed to provide more information on this initiative as it is publicly available. Overview slides presented by Almes at this CCIRN meeting and at the October CSG meeting are available at http://www.fnc.gov/hpcn_almes.ppt.
David answered questions about the CSG effort. The beacons will initiallyrely on NTD, which is adequate for some measurement (e.g., delay), butinadequate for accurate high bit rate path capacity measurements wheremicrosecond resolution is required. The platform will be defined byend-January. Global positioning systems (GPS) will eventually be installedon each of the measurement beacons in order to achieve very accurate timecorrelation among them.
Guy also provided an overview of the activities of the IETF IP Performance Metrics (IPPM) working group and the draft metrics-related RFCs to be discussed at the December IETF meeting. For details, see: http://www.advanced.org/IPPM.
e. Canada
Steven Ladouceur of Bell Canada provided an overview of Bell Canada’sstatistics activities. They will also conduct measurements over theEurope-Canada link. Steven felt that Canada would be interested inparticipating in a global measurement infrastructure, including installingbeacons at some of its exchange points – particularly at the end points onthese intercontinental links. Time synchronization, he agreed, will be avery complicated technical issue.
3. Asia
Kilnam Chon described the activities of the Asia Pacific Networking Group http://www.apng.org. Specifically he address topics relating to caching in Asia; Japan’s role as a hub for both caching and measurement activities; internationalization of Asia’s research networks; increased participation by developing countries; and commercial networking in Asia. He also describe the status of the Asia Pacific Advanced Networking http://www.apan.net as it relates to network design and operation planning; network research; and emerging networking applications, particularly biomedical applications.
Chon reviewed the plans for and status of APAN’s network, including plannedT3 or greater links between Seoul, Tokyo and North America. Other Asiancountries would interconnect with the trans-Pacific pipe via the Seoul orTokyo hub (through the AI3, RWC, or other networks). These connectionswould generally be E1 or less. Some would be via satellite link. Chon alsodiscussed the possibility of a high bandwidth satellite connection betweenTokyo and Europe. Chon described the need for performance measurement ofthese intercontinental links (10,000 miles or more). He described the needfor measurements which would permit "tuning" of these ATM links.
Professor Goto described measurements being conducted in Japan currently, including general end-to-end path measurements on round trip delay and packet loss, as well as link usage and traffic across specific hubs. Measurement will be an important feature of APAN’s efforts in 1997, with experience from Japan’s IMNET being applied to APAN, including the collection of information on BGP routing, general TCP traffic statistics, and mbone traffic statistics. For details on IMNET statistics activities, see http://inoc.imnet.ad.jp/mgmt.
IV. GENERAL
Participants discussed the need for improved mbone traffic measurement tools. Links to currently available nascent tools can be found at http://www.nlanr.net/INFO, including MRMap, MRHop, MTrace (the latter being the primary multicast measurement tool). The problem with these tools is that they require recognition from the MBONE servers; also there is a strong need to ensure that servers are running the current version of MRouteD.
Walter Wiebe discussed the importance of Internet measurements beingconducted within the context of service architectures. We are currently ina period of transition from buying "pipes" and best effort services, to anenvironment with differentiated qualities of service. Our ability toevaluate user requirements and map them to specific classes of service(basic and higher grade) is still in its nascent stages. Tools for modelingthese networks and evaluating services received against pre-definedbaselines also need to be developed and further refined. The group agreedthat these "design" and evaluation concepts provide context for futurediscussions of traffic/metrics measurement.
Last updated on 23 January 1997