Report date: October 10 '02
Dates of travel: September 18 - October 7, 2002
Travellers: Roger L. A. Cottrell
Position: Assistant Director, SLAC Computer Services
Employing Organization: Stanford Linear Accelerator Center
Address: SLAC MS 97, 2575 Sand Hill Rd., Menlo Park, California 94025 Airfare expense:
FTMS trip number: 200228834
Travel destinations: NIKHEF/iGrid2002, Amsterdam, Netherlands; CERN, Geneva, Switzerland; Padova University, Padova, Italy; ICTP, Tieste, Italy.
Vacation dates: September 20, 2002; Sunday September 29, 2002, Sunday October 6, 2002. I also worked the following full-days without re-imbursement Saturday-Sunday 21-22 September; Saturday 28 September; Saturday 5 October, 2002.
Purpose of trip: Install, configure, lead the NIKHEF/SLAC bandwidth challenge demonstration. Meet with European networking and Grid experts and scientists. Give talk on Internet measurements and performance for the ICFA/SCIC, participate in the ICFA/SCIC meeting. Meet with CERN networkers discuss European DataGrid activities, new TCP stacks for high throughput and the SLAC/CERN IEPM collaboration. Meet with collaborators at a major BaBar/SLAC/INFN tier 1 site with a view to improving Internet performance and measurements. Give talk on measuring and monitiring of Internet Connectivity to Research and educational sites at the ICTP meeting on "Developing Nations access to Digital Publishing". Participate in a panel discussion, meet & share ideas with people interestedin improving networking for developing nations and to help reduce the digital divide. Meet with networkers at INFN/Trieste.
Funding sources: NSF/DoE
Total Costs for DoE: $1360
Duty days:15 days
I attended the iGrid2002 meeting in Amsterdam and successfully set up a demo to demonstrate high throughput to over 30 sites in about 8 countries. I met with Antony Antony of NIKHEF and assisted him with extending the SLAC IEPM-BW toolkit for NIKHEF usage. I was able to coordinate efforts to provide and report on Internet measurements for the ICFA/SCIC working group. I learnt about a new INFN funded effort to deploy SLAC developed IEPM-BW measurement toolkits at 14 sites and was able to help making contacts with various INFN sites.
I was able to successfully demonstrate 3Gbits/sec from a 4 hosts to over 30
hosts in 8 different countries. This was an excellent test for the demonstration
to be done later this year at SC2002. I opened discussions with Level 3,
StarLight and Caltech about getting hosts placed at Sunnyvale for a high
speed SC2002 demonstration.
I learnt about an aggressive INFN effort to deploy 14 measurement hosts within INFN using the SLAC developed IEPM-BW toolkit. I was able to help the understanding of Internet performance for many developing country representatives.
At NIKHEF I met Cees DeLaat head of the computing/networking group there, Antony Antony who supports the network measurements at NIKHEF, Nocolas Simar of DANTE, and Dharmesh Chohan of Daresbury Lab, Maxine Brown Associate Director of the Electronic Visualization laboratory UIC, Paul Fernes of Level 3.
At CERN I met Richard Hughes-Jones of Manchester University, Harvey Newman of Caltech, Sylvain Ravot of Caltech, Olivier Martin, Wolfgang von Rueden and Jean-Philippe Martin-Flatin of CERN,.
In Padova I met Mauro Morandin, Michele Michelotto and Gravino Massimo.
In Trieste I met Fabrizio Coccetti and Roberto Percaci of INFN/Trieste, Prof G. O. Ajayi Director general/CEO of the National Information Technology Development Agency of the Federal Ministry of Science & Technology.
The demo was to show monitoring of high performance end-to-end links from iGrid2002 (Amsterdam) to 9 countries in N. America, Europe and Japan. Several people at SLAC including Connie Logg, Warren Matthews, Jiri Navratil, Jerrod Williams and myself spent several weeks before iGrid2002 developing applications and installing and configuring software on 2 hosts at NIKHEF. We were ably assisted by Antony Antony of NIKHEF. I met Antony face to face for the first time when I arrived at iGrid2002 on Saturday 21st September and we spent long hours on Saturday, Sunday and Monday setting up 3 more hosts and porting the applications. We decided to use NFS to simplify configurations (this was very successful and dramatically simplified configuring and updating, however, we had to be careful about ensuring various logs/recording files did not overlap from various hosts), added TCP parameter flushing to all hosts, set up 5 groups of remote hosts with roughly equal aggregate throughput, modified our iGrid2002 web page to provide easy access to all demo pages.
The main problems were caused by late access (Monday evening) to 2 of the measurement hosts and the large screens (this caused difficulty in installing WebStart, getting X to work with the various video cards, the Perl Tk module, emacs, & adjusting the displays to work with the limited resolution), non-uniform configurations of the various hosts, a broken software link not discovered until later, losing connectivity to some remote sites, e.g. due to hacker compromise, one site had to quarantine the remote host we had hoped to monitor, late addition/tuning of some remote hosts (2 at StarLight, one at UIUC) took time away from other tasks, an unexplained problem with the serial UTH plots locking up starting run-nw-tests on the first host, problems with ifconfig not keeping up at 1/sec readouts (we changed the software to allow an option to increase the time between measurements), the low resolution of the large flat panel displays, we were unable to take advantage of the 2nd GE interfaces in the hosts, there was a problem (Jerrod discovered and fixed) with the data extraction caused by trying to copy the data to antonia. The lessons learned will assist in the demos at SCS2002.
The first demo was at 9am on Tuesday 24th September. During this demo we showed the tomography tool, the replays of throughput measurements from SLAC, the PingWorld demo (real time round trip times from iGrid2002), real time Available Bandwidth Estimations (ABE) from iGrid2002 to about 20 sites, real time time-series plots of ifconfig interface throughputs for one host running sequentail throughput tests, and another running flood throughput tests. We were able to achieve an aggregate of about 2 Mbits/s to about 15 hosts.
The second demo was on the last day (Thursday 26th September). During this demo we achieved about 3Gbits/s with 4 hosts flooding iperf TCP data from iGrid2002 to about 24 remote hosts.
There was considerable interest in the animation maps, the tomography and ABE, and I will follow up with SLAC people at SLAC to see if the tools are portable for people at DANTE (Nicolas.Simar@dante.org.uk) and Daresbury (email@example.com). Some improvements suggested for the demos were: to display the configurations etc. of the hosts being accessed in the flood mode; provide an aggregate (for all hosts) throughput real-time time series plot independent of the central network's MRTG router displays.
Most of the demos used UDP and were video based.
The ANL folks had a lot of unresolved problems with achieving high TCP throughput from ANL to iGrid2002. We were able to achieve 200-350Mbits/sec with multiple streams in both directions for 10 second durations. For a single stream we achieved about 70Mbits/s
The Atlas/Canadian folks achieved about 375Mbits/s disk to disk per GE VLAN simultaneously over 2 GE links from TRIUMF to iGrid 2002 using 10 streams of bbftp for each GE link. they achieved a little higher than this using tsunami. I have the details in a presentation they made.
SURFnet growth since 1988 at factor of 100 in 5 years. 500,000 users 45 sites. Built of fibers and wavelengths.
Lots of latent demand, gave example of student housing tripling load from 100Mbps to 300Mbps in a couple of months. At the moment the network is faster than the computer (i.e. computer cannot keep up with 10G links). Increasing factor hundred in 5 years requires replacing infrastructure very quickly, e.g. ATM replaced by POS. Now moving to lambdas. Next step is not a simple evolution, counting on bigger routers and fatter pipes not realistic. SURFnet installed lambda to StarLight as research facility to assist in understanding international lambda networking.
iGrid2002 has 2.5 + 10G (L3) to Chi, + 10G to NY (Tyco), 2.5 G to CERN (upgrade to 10G in Dec 02 + dark fiber to Dwingeloo, CERN 2.5 G to StarLight. Astronomers can get large aperture baseline and can get results via net in real time vs weeks. In US have TeraGrid (40G) plus Pacific LightRail. Dark fiber Amsterdam - Groningen - Hamburg. Hope for donation from W. Coast to Japan goes live in Jan 03.
Imagine users/applications can route & control own wavelengths, can provision, insert, switch, route wavelengths across a carrier network. Self organization networks are being delivered in wireless world.
Customer empowered nets with carrier neutral IX (& OBGP routing) and condominium wavelengths. In Canada use Cisco add/drop muxes. Disk to disk transfer with a single hop (routers moved to edges) from Vancouver to CERN getting TB in 2 hours.
Have broadband optical amps with no regenerators.
Wireless was a problem. It may be possible to separate channels etc better. The flooding of the 10GE LAN at the end of the conference they got 8GE in and later 8GE out. Would be a good idea to have scheduled periods when everybody bangs on the network. The MRTG statistics will be made available from the iGrid2002 web page. There were problems with TCP with high speed interfaces over-running input buffers in NICs, routers. Needs re-implementation of TCP stacks, pacing etc. Need better packet analyzers etc. for GE & 10GE. Need more monitoring and analysis to understand. Most of the applications used UDP. Is there interest in keeping in touch. There will be refereed papers. The event was a huge success, 25Gbits/s capacity across Atlantic, big step up from 2 years ago.
I met with Paul Fernes of Level 3 and Linda Winkler of StarLight. With Paul we discussed placing some high performance PCs with GE interfaces at the Level POP in Sunnyvale. Level 3 has connectivity from Sunnyvale to StarLight ikn Chicago. Linda will investigate whether she can connect a Level 3 link to the StarLight Juniper router so we would have access from Sunnyvale to StarLight and thence to SurfNet/NIKHEF and to CERN. Another possibility we discussed is whether Level 3 can get access to SC2002 in Baltimore. With such connections we would be able to test high speed connections between Sunnyvale and NIKHEF/CERN/SC2002.
I also talked with Harvey Newman of Caltech about the possibility of such a configuration and he was excited and mentioned placing some of his PCs and some RAID disks (he would like a TByte) at Sunnyvale.
Olivier Martin of CERN and Sylvain Ravot of CERN/Caltech have built and placed high performance clusters of PCs (6 each with 64bit PCI buses and 2*2.2GHz cpus) at CERN and StarLight. From StarLight they had 2.5Gbits/s to iGrid2002. They also have a 1GE from StarLight to CERN. If they use Cisco ONS add/drop muxes then this can be configured as a VLAN and they can use jumbo frames. The above is all for test purposes and is funded by DataTag (EU, DoE & NSF). In addition CERN has a 622Mbits production link from StarLight. The policies/ scheduling for use of the test facility were to be discussed at the DataTag meeting the following week. Olivier and Sylvain agreed to provide an account for SLAC to uise, this will replace the current sunstats and pcgiga facilities that SLAC is currently using.
I was introduced to Andy Adamson of U Michigan by Shawn McKee. Andy (firstname.lastname@example.org) has a U Michigan funded project with Tom Hacker and Roy Hockett to provide web based Authentication/Authorization using Globus (GARA) to provide access to services. We discussed using such a service to provide access to on demand "looking glass" type iperf/bbcp/bbftp measurements, or Warren's web services providing window/streams settings or throughput expectations. We are already collaborating with Shawn McKee and Shawn has the IEPM-BW software installed on one of his measurement engines. This therefore appears to be a fruitful collaboration to pursue further and Andy appeared very enthusiastic.
I met with Sylvain Ravot. He has a modified Linux 2.17 TCP stack that allows one to set the increment in the additive increment, multiplicative decrease (AIMD) congestion avoidance phase of TCP. With this he has successfully used iperf in TCP mode with a single stream and standard 1500Byte MTUs to send 250Mbits/s from CERN to StarLight. With jumbo frames he has achieved 800-900Mbits/s. He has used increments up to 30 MTUs per RTT, though this is considered overly aggressive. I have a copy of the modifications and we will install at SLAC. He is currently investigating how to adjust the AIMD backoff, in particular to make it smaller (by default it halves for each congestion event). He has not looked at implementing the Sally Floyd suggested TCP modifications that adjusts the congestion window based on the current congestion window size
I talked to Richard Hughes-Jones of Manchester. He says he is aware of three implementations of the Sally Floyd High Speed TCP modifications. Gareth Fairey of Manchester and Yee-Ting Lee of UCL have developed one, Bill Allcock of ANL has 2 summer students developing one, and it is believed taht the Net100 folks (probably Tom Dunigan of ORNL) have implemented it on their TCP over UDP user level implementation. I have a PowerPoint presentation on the Fairey/Lee modifications. They are keen to try on the trans-Atlantic link and in particular to SLAC. I also installed the UTH ifconfig real time time series measurement/plotting tool on the Manchester host (gnt4.grid.man.ac.uk), but was unable to make it work until Java is installed.
I met with Jean-Philippe Martin-Flatin, the technical manager of the DataTag project (he is in the IT division at CERN). He is arranging a workshop at CERN in February to bring together network measurement folks with network simulation folks. It looks like a very interesting and timely workshop and the attendee list appears to have most of the major players. I have been invited to give a talk and paper on our experiences with the IEPM-BW measurement infrastructure.
I met with Alberto Franco de Sa Santoro from UERJ/Rio de Janeiro, Brazil. He agree to look into getting new contacts for the PingER project in Latin America. Dr Wolfgang von Rueden (the new IT division leader at CERN), is looking for contacts in Africa and Austria. I will follow up with them when I get back to SLAC.
New web site for committee: http://icfa-scic.web.cern.ch/ICFA-SCIC/
Harvey introduced some new members
Technical Progress Reports
Alberto will provide contacts in Latin America (e.g. Mexico, Colombia), Wolfgang von Rueden will provide contacts for Austria (e.g. Vienna and Innsbruck). Need to include high performance monitoring also.
http://www.icann.org/biog/quaynor is a contact for Africa <email@example.com>s
also see http://www.at-large.com/ & www.icann.org
Core running at 10Gb since July '02, Access link upgrades are in progress. UK traffic to US now goes via GEANT. JAnet/UKERNA have weathered the failures of ISPs (e.g. KPN/QWEST) very well, and have contingency plans.
GE commodity, emergence of 10GE, utilization of dark fiber (e.g. IN2P3, Caltech, SurfNet) & wavelengths
Loss of circuits to US was a problem. Have 622 & 2.5Gbps circuits to STATLight. Have 900 Mbits/s single stream for 24 hours from CERN to US with a single stream with a modified TCP and jumbo frames. Get about 250Mbits/s with regular MTU (1500Bytes).
Had a lambda on a VLAN from TRIUMF to CERN. With 10GE card (Syskonnect) in PC and 2 * 1 GE bonded links with Sylvain Ravot's modified TCP stack and got disk to disk about 350 Mbits/s per GE interface (i.e. ~ 750Mbits/s) using bbftp with 10 streams. Also used tsunami at slightly higher speeds.
Two main providers for R&E Rbnet (Fastnet ministry of Science) operator stopped in May and traffic sent through Runet/Nordunet at 155Mbps, One week about (16 Sep) started 155Mbps by Fastnet to StarLight to Moscow. Before may was operated at 45Mbps. Discussion in end May between GEANT and Russian, principal agreement fixed in June, in July formulation of fee to be paid by Russia to enter GEANT (475K Eu/year). From October got an agreement it will be free for this year. Channel should start soon. Next year has to be discussed. Slava will send update.
Renater 3 2.5 Gbits/s backbone being started will be fully deployed by mid October, and old links will be shutdown by end October. IN2P3 CC has 1GE access. Can be upgraded. GEANT POP in Paris. Main change for HEP labs no longer use ATM so no VP to Lyon from Saclay for example. Each Lab has a private link to POP (do not go through campus networks). Lyon to CERN upgrade from 155Gbps to 1GE private link only for HEP upgraded by end of year. Is under 1 year contract. Hope to get dark fiber by end 2003. See previous notes for prices of dark fiber (cash payment of 1M Eu and guaranteed for 15 years.
Have 2*OC48 (POS) Tokyo to NY into Cisco GSRs at each end. Looking at how to get to Starlight in (Chicago). It may be difficult to carry Japan traffic from NY to Chicago. Need to discuss by email, maybe go via CERN.
Super SCinet preparation are complete. The core is 10Gb IP service, and multiples of GE. Seven universities connected with GE including KEK. Will add 3 more universities this Fall. There is no progress on connections to China.
See ICFA Standing Committee on Interregional Connectivity
Expect to start using wavelengths in 2-3 years.. There are 3 implementations of Sally Floyd's RFC (Bill Allcock, UK (UCL & Manchester) and possibly Net/Web100 (but may be built on udp)).
Showed Brazilian links via http://www.rederio.br/topologia/mapa_sensitivo with clickable drill down to more information.
Has sent out a survey to learn about networks in countries, but only 20% response. Looked at various A&R network topologies and information from NSPs and also TERENA compendium report from 2001. The topologies will provide information on the various network at a particular point in time
Global Crossing fiber around S America into Florida is in place. But grant comes to an end soon. Showed map of Latin America with important HEP sites.
The Internet Educational Equal Access Foundation (IEEAF see http://www.ieeaf.org/) is trying to bring together donors (such as Tyco) to build improved networks for education. It is very important to provide good information for them. We will provide a map of the world with HENP institutes. It will be referred to as the Global Research Exchange for Physics (GREP).
Need updated information on network capacities and topologies. Need to specify needs, get agreement and then assign someone to create and come up with procedures to keep current.
Web pages for experiences in solving last mile problems, local infrastructure, last mile (fiber IRU) pricing, vendor contacts
Systematize, classify regions, countries institutes:
Begin relationship with IEEAF?
New Classification: then also vendors who donated something (not just loans):
Sponsor workshops/conferences in less advantaged countries/regions that need much better connectivity, setup good networking for workshop, leave afterwards, must be willing to support region, local sponsor sponsors housing, advantaged nations sponsor time, need funding for travel. Optimistic schedule for one year's preparation.
ICFA meeting 2/2003. Overall report will be ~ 10 page report with 1 page executive summary. Refers to subcommittee reports on the web.
Monitoring about 10 pages with pictures.
Digital Divide Group: 10 pages of text, many pictures & graphs, tables ~ 2 illustrations in the main report.
Ed built 700 colocation facilities. He is the one who arranges the assets to enable the donations in IEEAF. In 2 years went from zero to > $100M in donations. 53 more pending donations; combined assets > $1B. Opportunity enhanced by hard situation of telecomms. Empowered by teachers of opportunities to create linkages; "behavioralist who unfolds diabolical plots". In world of telecomms; providers look at first high volume financial users (17 top); then multinational corporates then after that the rest - the first 2 are drivers. We (HEP) are the fathers of the development/driving force behind high performance networking (using, applying directing) at this window in time.
The effect of creating the donations is a marginal effect, they "give a straw and get a fire hose back". He "whispers in their ears for them to be socially responsible". They push for opening up market places, they fear obsolescence or a competitor taking control of the market place. He is motivated by the willingness to serve a larger community. Success is being invisible, since the minute the telecomms realize they are giving capacity to education for free, they pull back.
From our maps of HEP sites etc. he can estimate what strategies might work for various regions, e.g. where are the fibers, who owns them, how open are they etc.
I met with Mauro Morandin, Roberto (INFN/BaBar collaborators), Michele Micheloto and Gravino Massimo (compuyter support people) at Padova University. Padova is a tier 1 BaBar site and is currently engaged in the 2nd level reconstruction. This entails copying the data from SLAC and they have been running at an average of about 90Mbits/s for many weeks. They have a 150Mbits/s HENP dedicated link to INFN/CNAF at Bolgona plus a 155Mbits/s link to Milan for all other traffic. The link to Padova is supposed to upgrade to OC48 by the end of this year or early next year. BaBar also expects to double the data rate by then. They have a 131 1 unit 2*1.26GHz cpu IBM Linux cluster at Padova. They will make available a GE Linux host with an ssh account for SLAC to monitor.
I met with Roberto Percacci and Fabrizio Percacci at INFN Trieste. They have recently received a grant from INFN to procure, configure and deploy about 14 embedded Linux boxes running the SLAC developed IEPM-BW software at INFN sites. Six of the boxes will have GE interfaces and be 1 unit high, the remaining will have three 100Mbit/s interfaces and be 1 unit high but 0.5 unit wide. For reliability there will be no had disk rather they will use flash card memory. The data will be uploaded from the measurement hosts to an analysis/archive host. They currently have 4 hosts running, 2 are embedded and 2 general purpose PCs. Two are at Trieste and 2 at Milan. I pointed out to Roberto and Fabrizio that I had contacts at the Padova, and Rome INFN sites where it could be very advantageous to place the measurement engines. As a result of this the Trieste and Padova people have made contact and this looks to be a fruitful collaboration.
Held at the International Center for Theoretical Physics (ICTP), Trieste, October 4-5, 2002. ICTP is part of United Nations founded in 1963 at urging of Professor Abdus Salam. Founded in Trieste partly since at the time it was an international city at boundary of West & East.
The current meeting is a round table conference to facilitate contacts. Penelists each talk for 15-20 minutes followed by discussions of an hour or more. There were about 60 attendees. The round-table was web-casted and will be video taped. Hope to have tape ready to take when leave the meeting.
Web access was available via wired using a proxy server. They also had wireless access. Could not access VPN (probably protocol blocked).
Paper fixed:variable costs are 75:25 (variable is for copies). No control by publisher once copied. For electronic it is 99:1 and extra copy cost isn almost nothing (e.g. Napster). With appropriate precaution & contract publisher can control this better.
Gap between "haves" & "have nots" is a super set of "Digital Divide". In 1984-5 Abdus Salam proposed a bulk deal where ICTP gets one copy at full cost and remaining ones at 1/3 cost. Elsevier feared loss of control, and lack of long term scalability. Two years ago "Ideal Charter for Low Income Countries" announced at UNESCO to provide access to over 300 journals at greatly reduced prices, eligibility based on World Bank definition of low income countries. Nor philanthropy supposed to be sustainable commercial model.
Health (WHO) costs based on average country income (e.g. < $975/person/year free). Launched January 2002, access to 1400 journals, access by password through WHO to publisher's server. Next phase is for <$3K per capita countries expect to start January 2003 for modest fee access. Eligibility 42 countries 600+ possible institutions.
African Virtual University (AVU) initiated by World Bank, for Sub-Saharan countries.
Speaker does not believe free distribution is not a lasting option., price differentiation (e.g. by GDP) more promising. However, publishers funding models in developed countries is still evolving with digital world. The technological problems remain. Are research journals the highest priority of developing countries (e.g. are textbooks more important, & distanced learning). Steps so far have been useful, need usage feedback to help chart future.
Problems for not-profit publishers are similar to those of for profit publishers, both technical and financial. The APS has always had tiered pricing, members had lower (factor of 1:10 member:library) cost access. There are real costs, in particular editing, receive over 100 articles per day. APS is not for profit but also not for loss.
Technical problems: absence of (or very expensive) internet connectivity, no cost distribution (with ICTP) of CDS; very slow internet connectivity solve via email distribution of papers. Block auto bulk downloads of complete archives (blocks access for others as well as make all information publish).
Financial problems: inadequate funds for purchases of access (of course); need for professional societies to cover costs; not only responsibility of society to provide access, better off institutions should support access for less well off colleagues; consortium arrangements can help accomplish broadened access; societies can provide low cost or free electronic access when revenue is not imperiled.
Long term goal: finding a funding model that assures recovery of costs and permits barrier-free access throughout the world.
Institute of Physics based in England. Publishes 6 magazines, 36 research journals, 600 books/year. Put all journals on web in 1996, web access free to subscribers. In 1997 made journals free to Africa, published via email and via Ghanaian physicists, got 4 institutes, have republished, now have only 14 institutes, mainly in S. Africa. Problem is Internet Africa. In 1999 launched new journal on physics made freely available on the web. In its fourth year. In 2002 made all journals to all Russian institutions at advantageous terms (nationwide consortium).
Many similar evolution and problems to those encountered by APS.
Real growth in future will come from knowledge based industries, and science is a major driver of this. Thus it is very important form disadvantaged countries improve their access to science information.
JHEP is a scientific journal written, run & distributed by electronic means, encompasses all areas of HEP. Owned by SISSA. Has automatic electronic peer review, and an advisory board. Now journal with highest distribution and looks like it will soon have the top citations. The cost is established at $0.5M/year/scientist in developing countries. Scientists have a strong say in how it is run. Cost to users is free.
Recently starting to form the JCAT (Journal of Cosmology, Astronomy and ?)
A.S. is a review of all fields of science & engineering. Want to create a low-bandwidth searchable archive of current and past American Scientist. Want membership to be about $5/year/scientist. Have to register and sign-up. Issues in developing countries: isolation, poor access ... Want 2 way access. Will have fully illustrated as well as low bandwidth capabilities.
Digital divide is widening more quickly than other divides, however, it is probably easier to close since it is a technical rather than social issue. Subscribers need to sign copyright agreements. Tiered costs based on per capita GNP. Can be free for <$1000 pc GNP.
Grew out of post communist world needs. Big effort to enable countries to help themselves, e.g. working with governments on policies to push ICT issues up on the agenda.
Aim to support capacity building in the research sector in developing and transitional countries. By delivering information, disseminating local research, enhancing ICT skills and strengthening local publishers
Making Journals available. Also working on training information and strengthening local publishing. Challenges: projects are making impact, usage is going up, need for hardware, training. Channel for subscriptions.
FAX Kofi Anan secretary general of UN sends greetings and a long message was read encouraging improving ICT for developing nations.
Support at ICTP are 2 people with support from others and Hilda Cerdeira coordinating. Problems not enough bandwidth, to load in real time, lack of necessary hardware, cost of paying for connection. Little real-world effort to study extent of connectivity at universities and institutions in developing countries, except for a few cases and with a specific purpose. If such a study were done we might have a better idea of the research potential of those institutions.
Purpose is to provide individual scientist with online scientific articles, which they can search for & download using only email, and allows them to follow hyperlinks as if they were surfing the web via a live Internet connection. Program depends on relationship with publishers. Only registered scientists working in 3rd world countries can use it and only after accepting certain requirements.
There is overlap between program goals. Typically not too much overlap between programs in a given country.
Goal: Use technology to strengthen organizations working to improve the lives of people in Africa. Access to learning opportunities, access to Internet services, networking knowledge & sharing
Kabissa has 10K subscribers to news letter on social justice. 200% growth in 2001. 450 member organizations from 33 African countries.
Provide email access, Cyber cafe user support (web based services). www4mail
Requests/day 4200 requests in last7 days, 1500 files requested, data transferred=654MB, rarely used from Africa mainly Russia, Ukraine & Cuba.
Need to improve image, simplify the user interface, add support for other languages, gather data on usage, user registration.
3 colleges, started email in 1994 through S. Africa, 3 kbits/s to single host. Now > 300 people attached now 14.4kbits/s. Still using a 1994 model computer.
In some countries (mentioned Zaire & Cameroon) access to information is becoming worse. Email works very widely.
Send email to special email address and it provides access to medical databases (e.g. free medical journals).
Development of science & research cannot be done today without Internet access. Academic community is obvious starting point for efforts that will permeate the society. Underlying problem is lack of qualified human resources able to handle new systems. ICTP has carried out training activities for scientists from developing countries in particular for radio communications, goal recently has been to enable the establishment of digital communication computer networks in developing countries.
Working on eJDS.
The gap: leaving some groups behind. Ownership of a computer will not solve these problems. Also need ability to manipulate technology through a variety of media. Gap continues to enlarge. Despite many efforts. ICT far from being at disposal of all. More than 80 countries have less thatn 10 phone lines per 100 inhab. 3 over 5 have less than one person that uses the Internet. Sci competitiveness especially from remote areas, requires access to net at low costs and tailored services like eJDS.
It awareness: how many online: cf NUA Internet Survey: Africa: 2.58 M users (0.85%) vs. USA Canada: 136.8, 45%. World total 304 M. as of March 2000. in May 2002, Africa 0.76%, 4.15 M. USA down.?
Quality of service: bandwidth: insufficient basic infrastructures in Africa. To help, need training; meaningful content; free, reliable open source technologies; secure transactions.
Also need: literacy: Burkina 18%; US, 99%. Hampers spread of computers and networks.
Economic: can’t pay rent, how can buy computer? Pentium system costs 350, average income in $6/mo in some countries. Some countries have fewer than 3 engineers per 10k people. Few training facilities. Often less than 10 standard networking labs. Even networking experts have poor access to tools and materials.
Inadequate power supply: often outside rated limits; power surges; frequent outages. – makes for equipment damages, unreliable networks, loss of reputation and money.
Why real time monitoring? Become aware of problems developing; provide data for optimal network configuration; monitor growth of traffic, plan ahead; set objectives; security; accounting.
What can be monitored? Status; link status; security;; performance; traffic, what is where; traffic flows
PingER (end to end reporting). Internet End to end Performance Monitoring.
Why active monitoring of Africa? Need to plan virtual laboratories – for instrument control at a distance, data sharing, collaboration, comm. – need stable network for this.
Better distr of money resources.
Need to draw special attention to low band networks. Need forecasts. Central archive of this.
Need monitoring to quantify digital divide realities in the region. Help create awareness of lack of facilities, networking, content.
Has assisted 35 countries in Africa. Doing this for last 15 years. Consultant. Advises. Evaluation of internet projects.
Connectivity Mapping in Africa.
General trends. % of people who are internet users 1998 .04% internet users; 2000 0.4%. world 6.7%.
ICT utilization: 1 in 160 use the Internet. More mobile phones than fixed lines. 0 to 24M in last 5 yrs. Because it’s the easiest way. Often takes years to get fixed line phone. Can get mobile in half an hour. Pay high proportion of income to get them. Of 800M people in Africa: 1 in 4 has radio, 1 in 13 have TC, 1 in 35 have GSM phone, 1 in 130 have PC, 1 in 160 have network, 1 in 400 have pay-TV. Big differences in Africa between countries. Best connectivity (apart from Seychelles) is Gabon, Botswana, S. Africa etc.
Not very accurate information. Census by polling isp’s. but a lot of people share accounts; don’t know how many. Multiple dial up subscribers by 3 – but there may be 5-8 in Africa. Doesn’t count cyber cafes, or corporate organizational facil. Counting number of users doesn’t tell you about extent or quality of use. Probably just checks email. Now thinking of looking at bandwidth used – gives you the amount of traffic. 70-90% of traffic is int’l – only abour 100 int’l isp providers in afr; if take total int’l bandwidth, get bits per capita. Wealth vs. connectivity plot goes from Gambia to Botswana to Gabon very wealthy (Seychelles off chart). Almost all near bottom, below $1000/person, and below 1 bit per cap ?.
Fiber – in last year, Atlantic 2 cable, connects Latin America to Senegal; and West African cable, connecting a number of countries (Cameroon…Nigeria, Gabon). Traffic from Ssia and S. America can now come directly to afr rather than having to go up to Europe. Additional options for S-S connect.
Telecom is biggest limitation to expanding connect: only about 20M lines – if take out N and S. Afr, 650M people, only about 4M lines.
Call charges too expensive: local call charges $2.50/hr or up to $6 – only 18 countries have local call charges for Internet access nationwide; but worrying: local call charges are increasing. “tariff rebalancing”. Most of lines in capital cities. Seychelles has given 70% discount for calls to internet.
New trends: revenue sharing with telcos – ‘free isps’, no monthly fee; gsm data integration. Can send SMS to websites – from mobile to web. GPRS – packet calls at higher speeds – don’t pay for time on line but packets you move on internet. Low cost. More interesting: user-financed infrastructure – WiFi – wireless networks of their own + satellite. Like Seattle Wireless. Don’t have monthly telecom charges. Setup high but payback quick. For $2-3000 can get satellite dish, $2-300 / mo.
Open source: popular, reducing cost of software.
Hybrid systems – Epost Senegal: send email to po box, it’s printed out; also wireless + broadcast satellite, one-way by low-speed wireless telemetry link, by standard digital home pay TV dish, good enough for cyber café.
Public access phone and Internet services:
Franchise model, e.g., AfricaOnline’s E-Touch
Small phone shops; use for email
Wireless local loop phone lines, galvanized iron shack in middle of nowhere. Slow to access. 4800 baud. Four poles, lines.
Another public access facilites, container based, local entrepreneur.
Internet business center: phone, fax, email, secretarial service – long distance call to get to internet, but still cheaper than long distance phone call.
Mixed business, e.g. hair salon & Internet café. The most sustainable model.
2003 and beyond, fiber: cables being planned Africa One; network for east and southern Africa ; should see that part of continent covered well. Rest of continent no plans. Contel project planning fiber up through sudan and above.
Satellite – covers the continent, and more coming up. Prices expect to drop.
Limiters: electricity, import duties, education, trade restrictions. Most lectricity is in lagos, a few coastal cities, S. Af., cairo. On map looks black vs. Europe and mid east.
Push for competition in basic telecom services, allow users to set up their own wireless or satellite links.
Use broadcast media – DSTV + Worldspace – can receive journals
Optimise bandwidth utilization – often no compression; no local links
Reduce startup costs – promote open sources, recycled PCs (World Computer Exchange, Computer Aid) ($50-100 for machine capable of using internet)
Les Cottrell, asst director of computing services, SLAC
High energy physicists want to be able to work with facilities but not live there – collaborate remotely. So have to know about connection.
Uses ping infrastructure. Has been running since 1995. so have fair history of internet.
Round-trip time improving 10-20%/yr. Replacing satellites with land lines (satellite about 500ms).
Typical lower limit today 0.3 speed of light min fiber.
Impact of loss on applications: bad for typing or voice. If get above 3% loss. Back in 1995 there was high loss everywhere; getting much better. Fewer sites have dreadful performance (12% loss).
It’s improving – losses 50% a year; throughput by 80%/yr. Mainly in developed countries. Reduced use of satellites. Uganda loss reduced from 10 to 3%, but round trip time constant at 10ms.
Home site: www-iepm.slac.Stanford.edu
Workshops and conferences produce slides in HTML since open, and provide a CD with all information before people leave.
Only a few universities have >=64kbps connectivity. Nigeria 120M people, more than 1M GSM subscribers in 1 year. Access to Internat limited to cities. Policy is to liberalize telecommunications, now have 4 GSM operators, privatizing national carrier (NITEL). Declared ICT as a national priority. Launch of Nigerian satellite system. Want Nigeria to be key player in Information Society by 2005. Use IT for education, creation of wealth, eradiate poverty, job creation, global competitiveness. Infrastructure will use satellites and wireless distribution in country. Network at IlepIfe university in Nigeria grown from 2 subnets (1996) to 15 subnets today
Internet reached Senegal in 1980 with email and file transfer main services. In 1992 university became manager of .sn TLD. Look on OSIRIS site for information on Internet and Senegal. Internet connectivity is currently 53 Mbits/s up from 64kbps in 1996., ISPs 14, Internet subscribers si 12K, number of Internet customers is 100K, number of .sn declared domains is 1150 (effective is 160), number of Internet access points is > 150. Official minimum salary in Senegal is about $50. Typical $1/hr in Cybercafes. University (Cheikh Anta diop) access is 2 Mbits/s for all university, which is limiting. Modems connections are expensive.