In pursuing their mission to provide state-of-the-art networking infrastructure to the research and education communities in Taiwan, MoE and NSC will scale up TANet to meet the growing needs. On one hand MoE will expand TANet to connect high schools and primary schools, while on the other hand NSC is planning a new initiative which will warrant the research community adequate network service quality as required by advanced research projects. Together MoE and NSC will progressively upgrade the TANet backbone to the speed of OC-3 (155Mbps) at the end of two-year time frame, and to OC-12 and possibly OC-48 at the end of this 5-year project.

To provide "commodity" Internet services to education and research organizations alike, "TANet/I1" will be riding on the ATM backbone to accommodate these legacy needs. As the "old style" of Internet services, Internet-1 will continue on the physical TANet as a separate logical entity. At the same time, research projects that need high bandwidth, low latency and quality-of-service guaranteed support are ever-increasing. These applications require services and advanced protocols not available from commercial Internet today. Hence another logical network, "TANet2", will be created under the physical TANet and will focus on providing next generation Internet (NGI) and Internet-2 services to the research community. TANet2 is viewed as the counterpart to vBNS in USA and will be connected to STAR TAP. As related technology becomes available, policy-based and/or application-based routing will be enforced on TANet2 for better performance assurance and resources allocation.

With the announcement of the NGI initiative last year and Internet-2 initiative in the USA, CA*Net II project of CANARIE in Canada, the APAN project spearheaded by Korea and Japan in Asia-Pacific, as well as DANTE's TEN-34 in the European Union, TANet is also currently seeking interconnections with these R&E networks. In particular, owing to Taiwan's strong collaborative links at many levels to counterparts in the USA, Taiwan is seeking inter-continental connectivity with the USA Internet-2/vBNS backbone as a top priority to facilitate these collaborations and to increase the scale and quality of collaborative projects. TANet will support a GigaPOP which acts as the gateway to vBNS running at the speed of fractional DS3 initially and progressively increase to OC-3 speed (155Mbps) in five years.

In introducing new network technologies to TANet2 and the USA-Taiwan link, TANet will work closely with STAR TAP, and leading vendors like Cisco Systems Inc. and International Business Machines (IBM) have both indicated strong intention in transferring the latest technologies and experiences of NGI/Internet-2 to TANet2 test-bed for maintaining seamless and smooth production run.

Conceptually TANet will have a multi-tier linkage to other networks at several levels, as shown in the figure below.

Abbreviations:

APAN - Asia Pacific Advanced Network; BARRNet - Bay Area Research and Education Network; TWIX - TaiWan Internet eXchange; HiNet - local ISP; SeedNet - Local ISP; TANet - Taiwan Academic Network.

1. Inter-Continental Links to USA and other countries

At the topmost level, TANet will support a GigaPOP acting as a gateway for outbound TANet traffic. In this proposal, traffic from TANET2 to the R&E networks of USA and other countries, such as Canada's CANARIE, will go through this GigaPOP and then to STAR TAP via the USA-Taiwan high-speed link. The purpose of linking to STAR TAP is to allow more efficient routing to major research networks in USA, mainly vBNS and perhaps others. Also, with the permission from NSF and agreements with other networks, it is expected that the transit traffic to R&E networks of other countries can be routed using STAR TAP as an inter-continental exchange point, provided that the respective R&E networks and their proposed USA POP are linked to STAR TAP as well. This scheme will implicitly help STAR TAP to become the "de facto" international exchange point for advanced networks, and foster faster and more integral international collaboration.

2. Inter-Regional Links

Currently AS (Academia Sinica) has already been using its dedicate link of 512Kbps speed to USA for astrophysics research. This link is connected to California in USA. In addition, Taiwan is considering to increase its participation in APAN, as a token of proper membership in the Asia-Pacific region.

3. National Links

As of now, TANet has private peering agreements with the two largest ISP's (HiNet and Seednet) in Taiwan. In the immediate future, other non-TANet domestic commodity traffic will go through the TANet/I1 gateway to TWIX. Eventually, the TANet GigaPOP will be linked to TWIX for routing non-TANet domestic traffic, whereas international traffic will go through the USA-Taiwan link. Specifically traffic from authorized R&E organizations, that of TANet2, will route through STAR TAP.

4. TANet Links

The TANet backbone is composed of "nodes" connected by DS3 over ATM switches (http://www.tanet.net). Each node is a first-level POP for TANet and acts as the "regional center" for connecting nearby TANet institutions in a hierarchical fashion. TANet2 will connect only these regional centers, and it will be necessary to separate the research traffic for TANet2 and vBNS at these regional gateways.

5. Campus Networks

Campus networks of R&E organizations on TANet2 will be encouraged by NSC to upgrade their internal networks to high-speed backbones. As of now, TANet already has connected more than ten universities with OC-3 speed networks in production. Specially in the Hsinchu area, inter-campus OC-12 links have been installed among NCHC, NCHU and NCTU. The campus networks are connected to regional centers or smaller "sub-regional" centers and to the TANet backbone.

The objectives of the linkup between TANet and vBNS as proposed are to:

1. Provide state-of-the-art network connection between institutions on vBNS in USA and on TANet in Taiwan, for the purpose of improved advanced research and education collaborations;
2. Implement new Internet technology on TANet and the USA-Taiwan link, for the purpose of realizing advanced network technology for the future;
3. Devise long-term collaboration mechanism and develop improved environment upon the upgraded USA-Taiwan link;
4. Identify potential topics for joint research projects that can demonstrate the potential capabilities of advanced network.

TANet is a non-profit network funded by MoE and NCHC of NSC, and is operated jointly by major universities and research institutions. Under this charter, only non-commercial activities will be involved, and only organizations of the following categories are entitled for connection:

1. public and private educational institutions, namely, K-12, colleges, and universities;
2. government funded research organizations, such AS, the "National Laboratories" of NSC, Industrial Technology Research Institute (ITRI), NHRI (National Health Research Institute), and Information Industry Institute (III);
3. instruction-based hospitals;
4. social education agencies, such as the "culture centers" of city and county governments;
5. public museums and performing art centers, such as the National Museum of Natural Science, National Concert hall, etc.;
6. public libraries of all levels;

The large number of institutions on TANet gives indications on the future potential scale between TANet- and vBNS-based collaboration when the high-speed interconnection plan is in-place. Only institutions with concrete joint research or education projects will be allowed to participate in the current proposal and route network traffic via vBNS. Institutions identified at the initial stage are given in Section C.8-2 of this proposal. Specifically, the AUP (Acceptable Use Policy) as required on vBNS from NSF will serve as the guideline for the projects. NSC and AS will be responsible for identifying, promoting, and screening projects and institutions in Taiwan for this purpose. Institutions not qualified for connection to vBNS will use the "conventional" Internet channel to route their traffic.

Initially, TANet2 will provide an uncongested "bearer" service to give reasonable assurance that there will be adequate bandwidth for carrying out advanced Internet applications for R&D. The final goal is to implement policy-based and/or application-based routing when the related technologies are available. In line with Internet-2 objectives, it will progress from a best-effort service to a differentiated communication service. Queuing mechanisms, such as priority queuing, weighted fair queuing and weighted random early discarding, will be used as RSVP, IPv6, and QoS become mature. Using these leading technologies TANet2 will support an advanced communications and computational infrastructure that will meet the demands of Taiwan's growing R&E community. In the meantime, TANet will design, create, implement and operate a GigaPOP for connection with vBNS via the Chicago STAR TAP for inter-continental R&E connectivity.

*: Router to perform "institutional-based" routing

Major TANet members, i.e., the regional centers, will link up to the GigaPOP through TANet backbone from their production networks. Other TANet members can link up at the minimum of DS1 speed to TANet regional centers initially. The TANet GigaPOP will provide an fractional DS3 link (45Mbps) and progressively increases to full DS3 and OC-3 interconnect with vBNS.

Initially the underlying ATM structure of GigaPOP will support and provide both IPv4 and ATM accesses to all connected institutions. PVCs and SVCs will both be supported when the compatibility issues between vendors are resolved. As IPv6 and RSVP mature, the GigaPOP will make use of them to provide increased control over quality-of-service (QoS) for diversified applications from research projects. From the start, the underlying ATM structure of the GigaPOP will offer coarse-grained QoS, such as bandwidth-provisioned virtual circuits, to end-to-end ATM applications.

To match the requirements of Internet 2 GigaPOPs, equipments and mechanisms provided for at the TANet GigaPOP will include:

1. high capacity advanced packet data switch/routers capable of supporting up to OC-3 link speeds and switched data streams and packet data routing.
2. switch/routers supporting IPv4 and IPv6, advanced routing protocols and QoS protocols such as priority queuing, weighted random early discarding and RSVP.
3. SONET or ATM multiplexers to enable bandwidth allocation for different services.
4. ATM inverse multiplexers to support link speeds in between DS1 and DS3.
5. traffic measurement and accumulation of metrics and statistics for monitoring network performance and managing network resources.
6. help-desk and trouble-ticket systems for facilitating the trouble shooting process.
7. public web pages for disseminating operational status, performance-monitoring results and narratives, and progress reports at regular intervals.

After consultation with Chicago STAR TAP staffs, a possible model of the initial architecture in the interconnection with vBNS is as follows.

This proposed architecture will support and provide for registered autonomous Internet networks with BGP-4 routing to all connected TANet members and manage the edge routers and switches connected to TANet from premises of TANet members. This includes support of all necessary routing management for interconnection to other international Internet-2 type networks which may ultimately be connected to TANet. The multiple layers of services will be accommodated in a single physical network TANet, using IPv4 routing, IPv6 routing, ATM switching, digital video, and integrated services. Support for the use of advanced networking protocols and performance measurement tools like RSVP, MOSPF, MPOA, Tag Switching, IPPM, and OC3MON can be expected progressively as technology transfer takes place. The experience in these advanced technologies from the academia of Taiwan will backup the implementation of the network. In-depth descriptions of past activities in this area are given in Appendix-J.4. In addition, there would be other advanced services which commercial ISPs may not be able to provide, including:

• 6-bone tunneling;
• IPv6 DNS;
• Hierarchical web proxy caching servers;
• Mbone with true multicast IGMP support;
• Audio & Video Broadcast Streaming Servers;
• Video-on-Demand Servers;
• Network services such as route filtering, route serving, and network firewalls;
• Other multimedia servers.

A network operations center (NOC) will be set up to maintain the service quality of the TANet GigaPOP with adequate provision for backup power supply, climate control and security, with the appropriate quality staff for round the clock support. A liaison staff will be available for immediate contact by the STAR TAP operations round the clock to ensure that any network emergencies can be quickly resolved. Network security will also be emphasized through active participation in Computer Emergency Response Team (CERT), Forum of Incident Response and Security Teams (FIRST) and other international Incident Response Teams (IRTs).

In preparing for the smooth implementation of new network technologies when they become mature and available, extensive interoperability tests and experience exchange among related parties are needed in the process. To expedite the sharing of experience and adoption of new technology for collaborations, this proposal will leverage the strength of the STAR TAP engineering team, and will also include related industry leaders.

Even as the advanced network connection is readily available, a great deal of challenges remains to be met before meaningful applications can be implemented. In planning for a long-term sustainable collaboration relationship between the research communities of USA and Taiwan, it is necessary to examine the components of the "ideal environment" of collaboration. Moreover, it is equally important to devise a "proper mechanism" for bringing projects from the two sides together. Only under this approach can the collaborations between USA and Taiwan be carried out in a more organized and efficient manner.

Strategy-1: build ideal environment for collaboration

This proposal views the entirety of international research collaboration of advanced network applications as the composition of several layers. By NGI's definition, the basic layer is the "high-performance network fabric", upon which the layer of "advanced network service technology" is built. Here the "revolutionary applications" of NGI are further divided into two layers, namely the "application environment" layer that lays the foundation for and enables specific "meritorious applications" of a higher layer. Included here in the definition of "environment" are the utilities and tools as well as software systems and packages that enable meritorious applications. The application environment functionally is equivalent to the API (APplication Interface) found in many systems, but has contents more enriched and high-level in nature. The figure below illustrates this layered environment.

The "application environment" is a subtle but vitally important layer that is overlooked by many. In order to promote application researches to make better use of new network technologies, it is important not to over-burden the research teams of various disciplines with too much in-depth knowledge of networking technologies, and that these scientists are provided with a consistent and efficient environment to work with. The responsibility for developing this interface layer really falls on "double-talented" institutions that provide advanced computing and communication facilities. It is the intention of this proposal to coordinate the collaborations between high-performance computing facilities of USA and Taiwan to develop a better environment for the future.

Strategy-2: build mechanism to facilitate collaboration

As important as identifying a large pool of potential individual research topics for collaboration is to devise a mechanism that will enable long-term collaboration on applications of advanced network. This is because most research projects are phased for 3-years at most, which is shorter than the effective duration covered by this proposal. A mechanism allowing incorporation of new projects not viable at this time would reduce the likelihood that the latter half of this 5-year proposal being void of any collaboration. In addition, pioneering research topics will emerge from time to time as new disciplines are created. This type of project requires concentrated and concerted attention from the international research community as a whole. However, because of the uncertainty involved with the "frontier" type of research projects, they can be proposed only at proper timing in the future. A mechanism would have its merit in this respect as well.

With these philosophy, this proposal places emphasis on the "mechanism" as much as on the "content" of collaboration. The mechanism is built up through two approaches. The first channel is based on the existing agreement between NSF and NSC, which allows research groups from the two sides to apply to NSF/NSC for supports on their respective activities under a common topics of collaboration. In the past the agreement has been carried out in full and solicitations of projects have been made to the academia each year. In the future, projects identified for collaboration from this channel would be examined for potential in employing the advanced network infrastructure implemented by this proposal. The second approach involves using the channels of high-performance computing facilities, such as NCSA, to their research communities, as below:

NCSA (National Computational Science Alliance):

"Application Technologies Teams" is one of the critical components in the NCSA partnership program, where dedicated leaders in a wide range of computational science and engineering disciplines are assembled to work on a research and advanced development plan of a national scale. This application team will therefore consist of the foremost research projects using NCSA's facility, and hence constitutes an ideal base for collaborations using advance networks.

NPACI (National Partnership for Advanced Computational Infrastructure):

under the recently formed NPACI led by SDSC, a number of "thrust area" are identified. Each thrust area will be comprised of a multidisciplinary team of applications scientists, computer scientists, and technology developers from multiple partner sites. Projects will leverage ongoing, separately funded research projects to ensure rapid deployment and robustness of the resulting infrastructure. This focused approach, consistently validated at SDSC over time, has led to greater progress overall in infrastructure development and has benefited the broadest array of disciplines. By way of thrust areas, project identified for collaboration will be able to take better advantage of advance networks.

NCHC:

for the past years, NCHC has been coordinating a "high-performance computational technology" research program in Taiwan for promoting applications of high-performance computing and communication. Under this program NCHC outlines a master plan that suggests the future focuses in each professional discipline. This master plan is used as the basis for "request for proposal" issued each year to research institutions, and special funding are set aside for awarding worthwhile projects. Projects eligible for support under this program are characterized by potentials in resulting in advanced computing algorithms, large-scale numerical models, or practical software systems. As of the time of this proposal, over 150 projects in various disciplines have been funded. Project investigators from this research program will serve as a good base for potential collaborations.

It can be seen that the "thrust area", "application teams" and "high-performance computational technology research program" are all already in the forms of "mechanism" for each facility to identify high-quality project topics utilizing state-of-the-art computing and networking technology. Although officially the newly awarded NPACI and NCSA have not yet begun to function, the importance of collaborating mechanism is recognized and is basically in-place. As it is the natural intent of most research groups to seek collaborations with parties with similar endeavor, NPACI/NCHC and NCSA/NCHC will only have to perform the catalyzing steps below, and collaborations between research groups will likely occur as a result.

Based on the spirit given above, this proposal outlines the application research collaboration under two topics: the first is the "application environment" that can enable more efficient collaborations, the second is the collection of some tentative "joint research projects" that will make use of the advanced network infrastructure in the initial years of proposal.

The "application environment" type of projects are extremely important, especially in the initial years, because the current Internet technology is unable to deliver the necessary tools for collaborations efficiently. Significant investment of talents and facility are needed from the HPCC community to build up the required base upon which the research community can use transparently. In this respect, NSC in Taiwan will dwell upon the National Center for High-performance Computing (NCHC) to coordinate with NCSA (National Computational Science Alliance) in USA on the following topics:

1. Integral Meta-Computing Test-bed

By "integral", the essential ingredients "computing", "data storage" and "visualization" of meta-computing are implied. Integrated together and driven by proper applications, they represent the most demanding requirements to a network, and will stress the performance of new technology to the ultimate. In the "computing" arena, there are already tools and emerging standards available to address the basic needs, and focus would be to develop more practical software packages (see below). But the computing spectrum is incomplete before issues of distributed file storage and remote visualization are also resolved.

(a) Distributed Data Storage

To support collaborated data-intensive computing demands, it is necessary to provide capabilities of distributed file storage, mass data storage, and data migration across long-haul or international networks, and should be addressed properly early on.

(b) Multi-User Shared Visualization and VR Environment

Scientific visualization and virtual reality (VR) has been known as an indispensable technique to envision large data sets, and is the basis of effective interpretation for research scientists. The Immersadesk and CAVE are pioneering environment from the Electronic Visualization Laboratory (EVL) of University of Illinois at Chicago (UIC) that allows user-data integration of this fashion. A further step is to employ the visualization or VR technology among collaborating groups across the network. Nevertheless, much is still under experiments and more pilot applications that can best utilize CAVE are still to be uncovered. NCHC will establish a Immersadesk/CAVE facility and leverage the experience and assistance from EVL/UIC, NCSA, and Argonne. In addition, this proposal adds an important concept of the "distributed" nature to the Immersadesk/CAVE environment. This will allow more efficient collaborations for research works that require in-depth visualization during discussions among researchers. The results will then be implemented domestically in Taiwan among NCHC and interested institutions. Currently applications in the area of numerical wind tunnel, transient flow field analysis and visualization of complex molecule structures are being planned.

2. Pioneering Application Software and WWW User Interface

There are certain disciplines that software system used by the application research are still far from ideal, e.g., that of CFD, molecular modeling, or bio-informatics, especially when "parallel" or "distributed" computing is concerned. With the web-based environment becoming increasingly popular, there is also a great need for friendly user interface to complement the computational kernel. These software, if designed with work patterns of collaborations over network in mind, will not only leverage new technology more effectively, but will also enhance the research productivity and facilitate efficient collaboration. As software development and testing involves highly specialized manpower, it is highly desirable that both facilities work jointly in complementary to each other to accelerate the progress.

There are certain disciplines that software system used by the application research are still far from ideal, e.g., that of CFD, molecular modeling, or bio-informatics, especially when "parallel" or "distributed" computing is concerned. With the web-based environment becoming increasingly popular, there is also a great need for friendly user interface to complement the computational kernel. These software, if designed with work patterns of collaborations over network in mind, will not only leverage new technology more effectively, but will also enhance the research productivity and facilitate efficient collaboration. As software development and testing involves highly specialized manpower, it is highly desirable that both facilities work jointly in complementary to each other to accelerate the progress.

3. Intranet over Next Generation Network

Projects with asterisk (*): Detailed abstracts are given in Appendix-J.1.

List of Abbreviations:

USA Institutions:
ALS: Advanced Light Source
ANL: Argonne National Laboratory
BIMA: Berkeley-Illinois-Maryland Association millimeter-wave array project
BHNL: Brook Heaven National Laboratory
CIESIN: The Consortium for International Earth Science Information Network
CIW: Carnegie Institute of Washington
CMU: Carnegie Mellon University
CWRU: Case Western Reserve University
GCRIO: US Global Change Research Information Office

FSL/NOAA: Forecast System Laboratory, National Ocean & Atmosphere Agency
HCFA: Health Care Financing Administration
HHC: Honolulu Heart Center
INEL: Idaho National Engineering Laboratory

IRIS: Incorporated Research Institutions for Seismology
JHU: John Hopkins University

JPL: Jet Propulsion Laboratory (California Institute of Technology)
MIT: Massachusetts Institute of Technology
NASA: National Aeronautical & Space Agency
NCAR: National Center for Atmospheric Research
NCBI: National Center for Biotechnology Information
NCSA: National omputational Science Alliance
NCSU: North Carolina State University
NIH: National Institute of Health
NLM: National Library of Medicine
NPS: Naval Post-graduate School
NSLS: National synchrotron Light Source
ORNL: Oak Ridge National Laboratory
OSU: Ohio State University

PMEL : Pacific Marine & Environment Laboratory
SFI: Santa Fe Institute
SRI: Stanford Research Institute International
SUMC: Stanford University Medical Center
SUNY: State University of New York
TAMU: Texas A&M University
UC: University of Chicago
UCB: University of California at Berkeley
UCSD: University of California at San Diego
UIUC: University of Illinois at Urbana-Champaign
UoI: University of Iowa
UoL: University of Louisville
UNM: University of New Mexico
UoP: University of Pittsburg

Taiwan Institutions:
AS: Academia Sinica

ASCC: Academia Sinica Computer Center
IAA/AS: Institute of Astronomy & Astrophysics, Academia Sinica
IBS/AS: Institute of Biomedical Sciences, Academia Sinica

CGCMT: Chang-Gung College of Medicines and Technology
CWB: Central Weather Bureau
CYCU: Chung-Yung Christian University
INER: Institute of Nuclear Energy Research
NCCU: National Chung-Cheng University
NCHC: National Center for High-performance Computing
NCHU: National Chung-Hsing University
NCKU: National Cheng-Kung University
NCTU: National Chiao-Tung University
NCU: National Central University
NHRI: National Health Research Institute
NSYSU: National Sun Yat-Sen University
NTHU: National Tsin-Hua University
NTOU: National Taiwan Ocean University
NTU: National Taiwan University