Figure 1. Cichlid 3-D visualization of the logical high performance connection (HPC) network across the U.S., including Abilene and Internet2. For additional information, please see section 3. more detailed view
1. Introduction
2. Network Analysis Activities (including Measurement Infrastructure Development and Management)
3. Networked Data, Documentation, Tools, and Result Presentation
4. Collaborations and Student Involvement
5. Program Plan (for 1st Q 2002 only)
Figure 1. Cichlid 3-D visualization of the logical high performance connection (HPC) network across the U.S., including Abilene and Internet2. For additional information, please see section 3. more detailed view
The development and ongoing management of the network analysis infrastructure (NAI) continued with excellent progress again this year. We continue to build support for, and add new data, tools, and features to, the NAI. Improvements (even small ones) build upon each other, creating an ever stronger foundation with which we can extend the parameters and meet the current and future challenges in high performance networking research. This report is a somewhat detailed overview of our activities in 2001. In the event that greater detail is desired, please consult our individual quarterly reports for 2001, as follows:
Activities during the fourth quarter are included here. This report is posted at http://moat.nlanr.net/Reports/Annual_report-Year4/.
The strong interest of other researchers and organizations (world-wide) in both the Passive Measurement and Analysis (PMA) Project and the Active Measurement Project (AMP) continues to build. This is reflected directly in the number of measurement monitors deployed, as well as in the large and diverse collaborations in which we are engaged. As of the end of 2001, the NAI has 135 active measurement monitors in the AMP mesh, including six on the wireless network (HPWREN), and 25 passive monitors (PMA) in place (8 of which are in the process of becoming fully installed). These figures are up from 118 (AMP) and 21 (PMA) a year ago.
Device obsolescence and technological advances became large factors for both the AMP and PMA projects. Changes in components, configurations, and site locations have been necessary in order to keep pace with manufacturing discontinuations, network growth, rapidly changing networking technology, and the resultant changes in the high performance network infrastructure.
Long term storage issues were solved first with the AMP data, then the PMA data, by arranging accounts on SDSC's High Performance Storage System (HPSS). The automation and streamlining of the process is still being fully developed; significant improvements and flexibility are resulting from this change.
Highlights for PMA in 2001 include significant changes in the PMA trace data collection procedure and the change to placing the trace anonymization on the monitors (all of the raw trace files are to be removed from the NAI server). Also included are a large number of new, very active collaborations. A primary focus for much of the second half of the year was to enrich the set of available trace data via these collaborations.
The community building effort to address and maximize PMA project planning that began at the end of 2000 continued to develop momentum. The NLANR Traces Community is a forum for announcements (such as the availability of new traces and conferences relevant to network measurements), and is intended to spark discussion and gather feedback on measurement plans and strategies and specifically what the needs of the community are, (such as which kinds of traces need to be taken, for what durations, during what time frames, etc.). For more information, see the archive of mailings at http://moat.nlanr.net/PMA/Traces/archive and http://moat.nlanr.net/PMA/traces.html.
In 2001, new PMA monitors were installed at Argonne National Lab (ANL), ADVANCENET (I2), the University at Buffalo, New York State University, and Metromedia Fibre Networks SJC#2 (commercial ISP site) [ABOVENET], with new monitors to be installed soon at sites such as the University of Maryland at College Park (nearly complete). Significant efforts were still being spent in the fourth quarter to restore some of the PMA monitors that have failed, lost fiber connectivity, been upgraded to OC12, or are being newly implemented. The most up-to-date status on PMA sites is available at http://pma.nlanr.net/PMA/Sites/.
Many of these changes promise to yield exciting sources of network research data. For example, we decided to retire two OC12c monitors at the vBNS - 12NCS and 12SDC - which saw little, if any, traffic in the last year or so, due to discontinuing use of the vBNS at SDSC and NCSA. Also retired was another system at the SDSC FDDI ring and the Cerfnet access link. New OC12 connection monitors are being installed at SDSC, one on the CalREN network connection and one on the Abilene network connection. These connections are expected to yield very useful packet traces. As a stepping-stone we are looking into implementing a small scale router instrumentation at SDSC with a total of four links at OC3c and OC12c. This will require some type of time synchronization; preparations are under way to bring a GPS signal into the machine room for this purpose.
A major project reimplementing many of the PMA tools from Perl to C began successfully with two analysis scripts fully reimplemented and several in the process (another was completed in early January, 2002). This revamp of the PMA data collection procedures became crucial due to the outage of our main trace collection server. These efforts resulted in improved efficiency and performance, and hence aid in the move towards the ability to support longer traces and higher bandwidth sites. Figure 2 below shows the performance comparison between the Perl and C implementations of tsh2ta.
Figure 2. This performance comparison graph of the Perl and C implementations of tsh2ta shows a dramatic improvement with the C implementation. For more information, please see http://moat.nlanr.net/~dgcheney/tsh2ta/Stats.html.
New long traces have been added (in addition to the Auckland II long traces): Auckland-IV (45 days continuous trace collected at the Internet access link between February and April 2001); Auckland-V (an ATM cell header trace collected at the OC3c ATM link; since removed in favor of Auckland-VII); Auckland-VI (three-point measurement, to be published soon); Auckland-VII (ATM cell header trace collected at the OC3c ATM link). All were collected at the University of Auckland. A four-hour trace was collected at Colorado State University; analysis of this data is in progress. A five-day continuous trace at NZIX (New Zealand Internet Exchange) was also published, including graphs of the data. For more information, please see http://pma.nlanr.net/Traces/long/.
Some progress was made on the router instrumentation. The Waikato Applied Network Dynamics (WAND) Research Group (New Zealand) was able to demonstrate OC48c network capturing at a commercial Internet backbone link in San Jose. Preparation and testing of a suitable computer system capable of hosting a pair of these cards in a 2U rack-mounted chassis were performed with a few trials and errors. (Size is a factor due to the anticipated space constraints at the Indiana GigaPop Abilene router [IPLS].) During the fourth quarter two monitor chassis were assembled and shipped for use in the OC48 monitor interface card development work (DAG4) by the WAND group (in partnership with its newly formed spinoff, Endace Measurement Systems Ltd.). One of these units, a 4 RU (7 inch) chassis is planned for the Mid-Atlantic Crossroads in the downtown Washington, D.C. area.
Several attempts at establishing some form of transatlantic or transpacific delay, jitter, and loss measurements with time synchronization have failed so far. The good news is that CDMA time receivers appear to be working in all the U.S. locations of interest, (including at StarTAP/StarLight [Chicago]), and in the AARNet landing point in Sydney, Australia, as well. However, local regulations regarding traffic monitoring have so far hampered an instrumentation that would make the collected data available to the general public, a central and primary goal for PMA measurement activities. For example, it has been established that public law will prevent any collaboration on passive measurements with locations in Germany.
The goal of the just beginning OC192mon project is to rapidly prototype and deploy a passive monitoring system capable of capturing IP header traces with precision timestamping at a single OC192c/10Gigabit Ethernet network link. If successful, the system can be supplemented by a second card to enable bidirectional link monitoring, thus enabling advanced functions such as application performance and flow monitoring. The Dag6 passive network measurement card is being developed by the WAND Dag research group in partnership with Endace (their commercial spinoff). We will then deploy the mature prototype to collect initial passive network header traces within the NSF NPACI DTF, the TeraGrid and the Pacific Light Rail network initiatives.
Further information on PMA activities is detailed in the Documentation/Tools, Papers, and Collaborations sections, following. Please see the Web pages for more information on the PMA project: http://pma.nlanr.net/PMA/.
Highlights for AMP in 2001 include the development and deployment of IPMP, the creation of a route display graphic, the use of HPSS for storage of AMP data, and the strong interest from other organizations in AMP's activities. The high reliability of the AMP mesh was also significant. Throughout the year, the AMP data collection functionality continued with a very high level of reliability and minimal outages. This demonstrates the maturity of the infrastructure, support systems, and procedures.
There are a number of organizations and groups interested in our measurement activities and possibly hosting an AMP machine. Deployments in progress or discussion include: FASTnet (fka MIRNet); 1GigE AMP monitors placed onto each of the four TeraGrid (DTF) sites (we are currently developing a system disk for use with a GigE ethernet card); and discussions have been held with representatives from China's CSTNet. Many items need to be addressed before progress can be made on these deployments (including technology transfer issues).
Additional sites were added to the AMP infrastructure in 2001; while fewer in number than in past years, they are in important locations. New active monitors were installed at NASA Goddard, the Jet Propulsion Lab (JPL), KISTI (Korea), Lawrence Berkeley National Laboratory (LBNL - a DOE facility on the ESNet), CA*net 3 ARDNOC (Canada), AARNet (Sydney, Australia), Rensellaer Polytechnic Institute (RPI), Rochester Institute of Technology, University of California, Santa Barbara, University at Buffalo, New York State University, and six monitors were placed on the wireless network (HPWREN). The AMP monitor at JPL is part of the JPL Performance Reference System. Both the Rensselaer Polytechnic Institute and the Rochester Institute of Technology involve network measurement collaborations with Bill Owens of NYSERnet.
IPMP: the IP Measurement protocol proposed by the AMP team. IPMP is a new network measurement protocol designed to facilitate the measurement of several hard to measure network performance metrics. It is intended to replace ICMP for measurement, and developed specifically for active measurement. IPMP provides a number of additional capabilities, including router support for one-way delay measurements and support for measurement of protocol-based priority queuing. It also supports forward and reverse path measurements of a single packet, allows accurate RTT measurements, supports BER measurements, and reduces the measurement overhead on the network and the potential for denial of service attacks.
As mentioned earlier, the development and maturation of IPMP has been one of the significant highlights of 2001. With the development and implementation of IPMP in the AMP infrastructure, we are able to do more sophisticated measurements which retrieve additional performance information from the network with a relatively simple packet exchange. Packet delay, clock adjustments on echo hosts, packet loss, and route asymmetry in terms of both hop count and potential delay are some of these parameters.
Since it was originally deployed over the AMP mesh in the first quarter, significant work and major progress was made on IPMP throughout the year. The ability to glean additional information from the network with IPMP required a redevelopment of the data collector infrastructure to enable the collection of this information in a scalable manner. The measurement daemon is designed to handle the collection of this information in a single executable. It schedules the measurements, collects the data, and performs compression on the data collected so that the data is stored in an efficient manner.
We modified and refined it throughout the year (including the IPMP draft itself), and discussed the protocol on the IPPM (IP Performance Metrics) mailing list. This is an important step if IPMP is to be standardized through IETF, as this group will be part of the standardization efforts. IPMP was discussed by Stanislav Shalunov, Ben Teitelbaum, and Matt Zekauskas, in "A One-way Active Measurement Protocol Requirements" at the 52nd IETF, Salt Lake City, IPPM WG, 2001-12-12 (slide 6 - http://www.advanced.org/IPPM/Meetings/ietf52/ippm-2001-12-12-52-owamp.pdf).
For more information and details on IPMP, please see http://moat.nlanr.net/AMP/AMP/IPMP/, and each of the quarterly reports.
In relation to work looking into timing issues in measurement systems, an active/passive validation of AMP was performed in the fourth quarter. The results indicate that AMP is well within the 1ms that we claim. Some interesting FreeBSD timing behavior was also discovered that will impact the way we do one-way delay IPMP measurements.
System manager: Full implementation of the system manager was achieved in the first quarter on a limited number of monitors. In the second quarter we extended the deployment of system manager over a larger number of machines, and began to understand the full impact of it with respect to bringing new machines up to speed. Beginning at this time, the procedure to complete the software set-up of newly deployed monitors consists of just a short list of steps to perform manually, then the system manager performs the other (software) tasks required. The system manager has been used extensively to update various components of the AMP mesh and has proved to be a valuable tool.
Disk failures: The redundancy provided by the amp and volt data collectors allowed full recovery with no loss of data and only minimal disruptions of service, but required quite a bit of effort on the part of the AMP staff for each failure. The utilization of the High Performance Storage System (HPSS) at SDSC (one of the main highlights of the year for AMP) is an excellent solution to the problem of continued growth of data storage needs. The use of HPSS circumvents the need to continually expand the disk storage capability of the AMP data collection servers, amp and volt. However, some issues remain. In part because of the large number of small files created by AMP, we are only slowly managing to automate this process. Solutions for these issues are being pursued with the HPSS staff at SDSC and resolutions are forthcoming.
Event detection: We performed some preliminary work on automatic event detection, which can recognize abnormal network performance and notify interested parties, often before poor performance impacts the users of the network. Unfortunately, early student work on this had to be discarded and the event detection algorithm redesigned and the code rewritten. The windows concept was maintained. We hope to develop and deploy a fully functional, in real time, event detection and notification system for our user community. Providing preemptive, as well as retrospective, diagnostic feedback to users and system administrators has the potential to significantly shorten the network fault cycle.
Many activities and procedures were performed throughout the year to improve or maintain the performance of AMP system. Several of these measures significantly strengthened and improved the efficiency of the AMP mesh and data collectors. Many of these changes were transparent to data collection and end-users (researchers).
Some analysis of the "Code Red" and the "Nimda" worms and their effect on the AMP sites was performed.
During the second quarter, we reported that the system board used in the AMP monitors was becoming obsolete and therefore, unavailable. Tests and investigation into replacement boards resulted in the selection of the ASUS CUSI-M board with the SiS 630E chipset. A side benefit of this change was that the new system board selection allowed for the AMP monitor chassis size to be reduced from a 4 rack unit (RU) size (7 inches high) to a 1 RU size (1 3/4 inch). The advantage of needing much less rack space is that it makes remote site installation easier and participation more attractive to new sites. However, computer component obsolescence continued to affect AMP into the fourth quarter. In the short time since our decision, the board manufacturer, ASUS, announced that this board/design has been discontinued. At this time we are testing a new board, manufactured by Gigabyte.
Plans have been developed regarding upgrading the operating system on the AMP machines to FreeBSD 4.4 (to be implemented early next year). [At the time of this report, difficulties have been found with GigE card drivers and 4.4, therefore the upgrade will likely be to the 4.5 version.]
Further information on AMP activities is detailed in the Documentation/Tools, Papers, and Collaborations sections, following. Please see the Web pages for more information on the AMP project: http://moat.nlanr.net/AMP/.
The need for network performance research on the wireless network (HPWREN) became greater in 2001 as more people began using the Internet and its networks. By studying the HPWREN performance parameters we are able to understand a network and its response to improvements in its infrastructure and/or expansion of its capabilities. This enables the optimization of the network's flexibility and utility for an increasingly diverse base of users.
The High Performance Wireless Research and Education Network (HPWREN) measurement and analysis infrastructure is being developed for this purpose. We perform a number of network performance measurements on the HPWREN; all measurement data is available online in near real-time. We currently have five primary sets of data:
The development of each of these data pages took place over the year with refinements, additions, and modifications to each as necessary. Many changes have been to make the data easier to access and use. New scripts are being written for further analysis of existing data, as well as to refine the processes of data collection and retrieval. Activities of particular note include:
For more information on the HPWREN network performance analysis efforts, please see http://stat.hpwren.ucsd.edu/.
Figure 3. This image is a Cichlid 3-D visualization system generated graph showing the analysis of a trace file. Trace files are 1.5 minutes to 2 minutes in length (time, x-axis).The analysis shows the network statistics for the trace file; each metric is shown per second/10000. From left to right on the y-axis: smtp bytes [red], ftp bytes [green], html bytes [blue], udp bytes [brown], tcp bytes [1st gray]; the number of bytes and packets form the last two gray rows, respectively. Please click on image for larger view with more detail.
The Cichlid 3-D Visualization System is a software toolkit which facilitates real-time visualization of remote data sets. Users can transform raw data into compelling images and animations, then explore and interact with the high-quality 3-D data sets as the information changes over time. It was developed by Jeff Brown, then an undergraduate student researcher with our group, (now a Computer Science Ph.D. student at UCSD).
Improvements to the Cichlid system and servers continued through 2001. Highlights of changes and additions include the following:
| Cichlid home page: | http://moat.nlanr.net/Software/Cichlid/ |
| Cichlid gallery (sample images): | http://moat.nlanr.net/Software/Cichlid/gallery/gallery.html |
|
Repository for Cichlid Servers (of all types): | http://moat.nlanr.net/~jfields/cichlid-servers.html |
Documentation and tool development for all aspects of the Network Analysis Infrastructure (NAI) continues to progress and develop with a number of new tools and Web pages for each of our activities. As always, we continue to make all data, that we reasonably can, available on a regular basis, through our Web site (and via ftp). Part of the evolution and expansion of the site as a whole involves making data of all types more easily accessible (and therefore, usable) by the HPC research and engineering community, as well as by a wider range of users, including students and other scientists. These goals are being pursued by creating database referencing for our data, as well as soliciting user input through a variety of methods. Each database is dynamically designed to include current data; additional improvements to each of the datasets were made throughout the year, in response to our and the users' needs, as well as to evolving technology considerations.
Several interesting additions, changes, and refinements were made in 2001 regarding PMA tools, Web pages, and documentation. Significant work was done to the PMA trace data collection and storage infrastructure. We also continued work on improving the PMA Web interface to make it more productive and intuitive for our users. The following is an overview.
Many interesting additions, changes, and refinements were made in 2001 regarding AMP tools, Web pages, and documentation. Of particular note is the route display graphic, one of the highlights of the year for AMP. The following list is an overview of these efforts, beginning with the route display graphic. A page showing the AMP graphs - RTT, Route Display Graphic, and Event Graphic - working together is available at:
http://amp.nlanr.net/active/cgi-bin/daily.cgi?amp-aarn/HPC/data/amp-asu/101.12.19.gz.
Throughout the year, other areas of our Web site were also developed, in some cases with new pages created. The following is an overview of these efforts.
Among the highlights this year, 13 new papers by our group were published, many of which included students as coauthors. Four additional papers have been accepted for PAM2002 (three will be presented, one will be a poster, all will be published in the proceedings). Another paper has been accepted for SAINT-2002, the 2002 International Symposium on Applications and the Internet, 28 Jan - 1 Feb 2002, Nara City, Nara, Japan. Of particular note, AMP project leader Tony McGregor has been invited to give the keynote address at the symposium.
A list of our total published papers appears elsewhere in "Fastlane" and is not reproduced here. This list is also available on our Web site: http://moat.nlanr.net/PubsResources/pubslist.html.
We continued to publish the Network Analysis Times, our newsletter which we use as a forum for presenting our activities and encouraging collaborations and the use of our data. Issues were published in April and December. Printed copies were created for distribution at conferences and meetings. A special effort was made to distribute the latest issues at the PAM2001 workshop in the Netherlands and at ITC17 in Brazil. http://moat.nlanr.net/NATimes/
We also continued to publish the NLANR Packets newsletter which covers the activities of all three NLANR sites. A new issue was released (posted to the Web page) in January of 2001; printed Table of Contents Summaries were created for distribution. http://www.nlanr.net/NLANRPackets/v2.1/
A new PMA poster was completed in the first quarter: http://moat.nlanr.net/PubsResources/posters.html.
TheNetwork Analysis Times and NLANR Packets, as well as our posters (NLANR, AMP, PMA, Cichlid) are on display at various locations. The NATimes was distributed at more than a dozen conferences and meetings throughout the year, across the U.S. and internationally. At most of these conferences and meetings, the NLANR Packets summary pages and various posters (11x14 posters of AMP, PMA, Cichlid, and NLANR) were also distributed. In addition to the events listed in the following Presentations and Meeting Participation section, distribution included:
The NLANR Measurement and Network Analysis Group was represented at the following conferences and meetings held in 2001:
ACM SIGCOMM IMW2001, Internet Measurement Workshop, San Francisco, November 2001
SC2001, Denver CO, November 2001
10GigE Workshop, SDSC, October 2001
Internet2 Virtual Fall Member Meeting, Internet2 End-to-End Performance Initiative
JET meeting, August 21,2001
CISCO Workshop, Miami, Florida, August 15, 2001
SIGCOMM Traffic Measurements for IP Operations
ACM SIGCOMM meeting, UCSD, August 27-31, 2001
AMPATH meeting, Miami, Florida, August 16-17, 2001
High Performance International Internet Services (HPIIS) Workshop, SDSC, August 23-24, 2001
We hosted this workshop in August in conjunction with NSF's High Performance International Internet Services (HPIIS) Project. High performance network researchers and administrators from all around the world attended. The workshop focused on the development of measurement metrics to quantify the use of HPIIS links in addition to possibly identifying research collaborations. Domestic and international advanced networks were involved. On the second day, Todd Hansen presented a demonstration on creating your own AMP network. For more information on this workshop, please see the Web page: http://moat.nlanr.net/Workshops/HPIIS-2001/.
Passive and Active Measurements (PAM2001) - Amsterdam, the Netherlands, April 23-24, 2001
STAR TAP International Advisory Committee Meeting, INET 2001 - Stockholm, Sweden, Tuesday, June 5, 2001
NRDM workshop - Santa Barbara, CA, May 2001
14th ITC Specialists Seminar on Access Networks and Systems - Barcelona/Gerona, Catalonia, Spain, April, 25-27, 2001
NLANR/Internet2 Techs Workshop - 14-17 May 2001, Lincoln, Nebraska
New Zealand Computer Science Research Students' Conference, University of Canterbury, New Zealand, 19-20th April 2001
NPACI/SDSC All-Hands Meeting, San Diego, February 26, 2001
APAN/TransPAC/NLANR/Internet2 Techs Workshop, East-West Center, Honolulu, Hawaii, 28-31 January 2001
Spring Internet2 Member Meeting, Washington, D.C., 7-9 March 2001
The NLANR meeting (Feb. 14) and NSF-NLANR meeting (Feb. 15)
I2 End-to-End Planning Meeting, Ann Arbor Michigan, Jan. 9, 2001
In addition to the formal meetings listed above:
We continue our strong tradition of collaboration for both the AMP and PMA projects. The Network Analysis Infrastructure (NAI) serves as a platform with which we support our collaborators, and the HPC community, by making all of our data, analyses, tools, and techniques publicly available for use by network researchers, engineers, systems administrators, and students.
As the size and scope of our various measurement and analysis projects has grown, we have expanded our outreach efforts to continue to meet the needs of the high performance community (HPC), as well as obtain feedback on future strategies and direction.
We have numerous continuing collaborations, some of which we have maintained for years. The following list includes both long-term relationships, as well as newly developing ones. This year we continued, or began, work with the following organizations and researchers:
Per a suggestion from Bill Cleveland of Bell Labs, we have had discussions with Mark Johnson regarding joint measurements in the NCREN network (North Carolina). Working with NCNI and Don Smith of the University of North Carolina, Chapel-Hill was also suggested. (4Q)
We have met with a group from China's CSTNet. They are interested in our measurement activities and possibly hosting an AMP machine. We are also working with FASTNet (fka MIRNet) regarding the deployment of an AMP machine, or AMP mesh.
Beginning in the third quarter, a primary focus for PMA is to enrich the set of available trace data via collaborations. For example, new traces are expected to arrive from the Bell Labs Internet access link as well as a first multipoint trace from the TransPAC. In addition, we are developing an illustrated trace archive. Our collaboration with Bell Labs is the first step of drawing other researchers' analysis results to the data, slowly building up a "knowledge base" of Internet header traces. The hope is that we draw more people into this activity over time.
Network routing activities were focused on the collaborative effort with the SKIDL group (the evolving data mining group at SDSC). The purpose was to explore the application of data mining techniques and commercial tools to networking data. Specifically, GateD BGP routing data (collected and archived by NLANR) was analyzed. Considerable time was devoted to characterizing the data and deciding what type of analysis could be performed, given the nature of the data set. Ultimately, the collaboration used Perl to format the data, SQL databases, and Intelligent Miner (IBM's commercial data mining software) to generate a model for specific phenomena evident within the data, and then tested the accuracy of that model. (1Q, 2Q)
Ronn Ritke participated in the planing, coordination, and organization of the Extreme Networking WORKSHOP 2 -TECHNOLOGIES, held January 8-9, 2001, at San Diego Supercomputer Center (SDSC), UCSD. (1Q) Ronn also participates in other SDSC activities throughout the year.
Attendance at workshops and meetings helped begin new (and strengthen ongoing) collaborations. We hosted the High Performance International Internet Services (HPIIS) Workshop, in August, at SDSC. High performance network researchers and administrators from around the world attended this workshop which provided new opportunities for networking and collaboration. We have since confirmed a passive monitor installation at the AMPATH access link at Florida International University, as well as implementation of an AMP mesh on CANARIE (see above). (For more information on the workshop, please see the Conference/Meeting Participation subsection earlier in this report and/or the Web page: http://moat.nlanr.net/Workshops/HPIIS-2001/.)
Attendance at ACM SIGCOMM at UCSD in August proved a very fruitful place to make new contacts and learn about the research work of other groups.
Attendance at PAM2001 (April) led to a collaboration with Piet Van Mieghem, Delft University of Technology, see above for more information. (3Q)
Several attempts were made to find collaborators for collecting data in commercial networks, but these efforts have failed so far. We will continue to pursue opportunities; however at the same time, we are focused on making the best use of our existing installed monitoring basis. (2Q, 3Q)
Students continue to make significant contributions to the project in many areas (script writing, displaying data graphically, Web page development and maintenance, technical support regarding data, note takers for the HPIIS workshop, as well as participating in other activities). Several students wrote papers for conference submission this year. Of the six papers submitted by the group (as a whole) to the ACM SIGCOMM IMW2001 workshop, five students were closely involved (three as first authors). (2Q)
David Cheney (UCSD undergraduate) serves as technical support for the PMA project and the trace users who have questions or problems with our data, formats, methods, and/or practices. He developed a FAQ in this regard. He has also worked on the PMA Web interface and retrieval and analysis tools for the data, developing several new PMA pages. The second half of the year he spent considerable time working to rewrite our analysis scripts in C (from Perl). Please see Sections 2 (PMA activities) and 3 (Documentation, Networked Data, and Tools) for additional detail on this important project reimplmenting many of the PMA tools.
Neil Cotofana (UCSD undergraduate) who had been working primarily on Network Routing activities with us for over two years, decided that for his last quarter of undergraduate education, he was going to work with SDSC's Data Mining Group - SKIDL. He had previously (1Q) initiated a collaboration with SKIDL to explore the application of data mining techniques and commercial tools to networking data.
Jamie Curtis (University of Waikato graduate student) presented his paper entitled "Review of Bandwidth Estimation Techniques" at the New Zealand Research Students Conference (end of April). It placed in the top six (6) papers and was therefore accepted to a special edition of the New Zealand Journal of Computing, now published. (2Q)
Stephen Donnelly looked into timing issues in measurement systems as a University of Waikato Ph.D. student (Computer Science); he finished his thesis work in November. He performed an active/passive validation of AMP. The results indicated that AMP is well within the 1ms that we claim. He also discovered some interesting FreeBSD timing behavior that will impact the way we do one-way delay IPMP measurements.
Building on Cichlid developer Jeff Brown's initial work, Justin Fields created a new Cichlid image of the logical high performance connection (HPC) network across the U.S., including Abilene and Internet2. (See Figure 1 above.) He represented the NLANR Measurement and Network Analysis Group with demonstrations of Cichlid at SC2001 and the NPACI/SDSC All-Hands Meeting. He maintains the Web page of the Cichlid servers: http://moat.nlanr.net/~jfields/cichlid-servers.html.
Kuo-Wen Lo (UCSD undergraduate) works on the measurement and analysis data from the wireless network (HPWREN). He presented his work at the SDSC student poster session, where he displayed a variety of graphs and charts that he had developed during the year. (3Q) By the fourth quarter, as a result of learning and understanding in more detail about the wireless network, its configuration, and how the routers, switches, etc. work together, he was able to recognize errors in some of his earlier scripts and correct them. He automated all scripts, such that when new components are added to the wireless system, they are automatically accounted for and data can be polled immediately.
Matthew Luckie, a University of Waikato Ph.D. student (Computer Science) returned to New Zealand in June, after having spent nearly six months with us at SDSC. While here, he spent time on the AMP project, particularly on the development of IPMP. He continues to work with us. He and Tony McGregor presented an IPMP paper at the ACM SIGCOMM Internet Measurement Workshop, November, San Francisco, California. Another paper on IPMP has been accepted for PAM2002 (where they will present this spring).
Jose Otero began graduate school (PH.D. student in Geophysics) at the Scripps Institute of Oceanography (UCSD) this past Fall. Jose continues to work with Frank Vernon (SIO geophysicist, HPWREN Co-PI) on seismic activities, including dataloggers, GPS activities, and other tasks. During the summer (2Q), he worked with Frank Vernon and Cathy Constable on a number of GPS related activities. These efforts will enable the use of more efficient means to align earthquake sensors.
Pavana Yalamanchili, UCSD graduate student, completed her Masters's thesis on "Study of Unlicensed Band Wireless Data Network," defended, and graduated. She continues to work with us on an informal basis. (2Q)
This section is not a standard program plan for a year's anticipated efforts, but includes only those activities which we have planned for the first quarter of 2002; after which, this cooperative agreement will be replaced by the new one (sometime in April, 2002). Our proposal for the new funding serves as the Program Plan for April 2002 to March 2003.
We will continue to manage and develop the NAI, as well as adding new activities. For both the AMP and PMA projects we will continue to be responsive to the changing needs of the HPC community and deploy new monitors as needs arise. The following is a list of our planned activities for the quarter (plus one or two weeks); it is possible that there are more tasks than can be completed in the time period.
-30-
