As the Internet moves forward into the next century, it is going to serve as
an information infrastructure for everyone, not just for scientists or
Project JB is a joint, high-performance, nationwide network research project in Japan. This project consists of several network computing research projects, such as the WIDE Project, the ITRC Project, the Cyber Kansai Project, and so on. They collaborate and cooperate with each other. Furthermore, Project JB has collaborated with international networking research groups, such as the Internet2 Project, the NGI Project, and the APAN Project.
The initial technological targets are IP version 6, quality of service control (e.g., differentiated service), real-time high-speed stream multicast, reliable multicast and new applications (e.g., school on the Internet). The test-bed is going to integrate all of these functions and protocol stacks into a unified platform based on the PC-Unix-based advanced network stack provided by the KAME Project.
Practical network operation with the integration of the above functions and protocol stacks will disclose some new issues and requirements for the next-generation Internet infrastructure. The deployed test-bed also works with international research networks, such as 6REN and Q-Bone (end-to-end quality of service backbone).
Keywords: DV, Digital Video, IPv6, IP version 6, multicast, QoS, Differentiated service
Internet technology provides global, ubiquitous connectivity for all computers through various datalink platforms.
Since connectivity is the Internet's own reward, the Internet has been growing numerically and geographically at a more than exponential rate.
The core technology for the Internet is TCP/IP. A network in the Internet contains computers that are connected using IP. As well as connecting computers, IP goes further and connects several networks. So the Internet is a network of networks. Reliable data transmission is achieved by the end-to-end TCP control mechanism between source and destination hosts. Therefore, it is said that the Internet with TCP/IP is inherently a distributed system.
As the Internet moves forward into the next century, it will become an information infrastructure for everyone, not just for scientists or professionals. This means that the next-generation Internet must achieve the following features.
Now, what are the technological requirements and challenges for the next-generation Internet? The main one is "scalability". Scalability has various aspects from various quantitative and qualitative viewpoints.
The initial technological targets of Project JB are IP version 6 (IPv6), quality of service control, multicast, and new applications.
IPv6 provides sufficiently larger address space (128 bits) and provides various new functions compared with IPv4. For example, IPv6 enables autoconfigurations for end hosts to accommodate everyone in the Internet.
Each person and each application may require different communication quality levels. Differentiated service, Integrated Service, and RSVP are the key architecture models and protocol stacks that provide various communication qualities for each particular packet flow over the Internet. The quality of service control must be scalable regarding the heterogeneity of the required service class and of the condition of the network.
The unicast-based multicast emulation that is the general solution in the existing Internet environment may not be numerically scalable for multicast capability.
Multicast has two different technological challenges. One is for real-time high-speed stream data transmission, e.g., digital video (DV) data multicasting. The other is for error-free data delivery for a large number of receivers, i.e., reliable multicast.
Using the new technologies described above, new applications should be researched and developed. For example, in Project JB, the School on the Internet (SOI) applications using the above technologies will be deployed.
Project JB will integrate the above-mentioned functions and protocol stacks into a unified platform based on the KAME stack Also, we will operate these above- mentioned functions over a newly deployed research test-bed, called the JB Project test-bed.
Project JB is a joint research project to explore next-generation Internet technologies. The project is organized by several network computing projects, such as the WIDE Project, the ITRC Project, and the Cyber Kansai Project.
Therefore, we are building a new, nationwide high-performance research test-bed that is jointly operated by the participating research projects and organizations. The test-bed also works with international research networks, such as 6REN (IPv6 Research and Educational Network) and Q-Bone Network operation with the integration of these functions and protocol stacks described above will disclose some of the new issues and of requirements for the next-generation Internet infrastructure.
The "Project overview" gives a brief description of the project , the test-bed topology, and its architecture. "IPv6", "Multicast", "QoS," and "New application" give a background, roadmap, and work items of each subproject.
The goals of Project JB are to develop new technologies to construct a nationwide next- generation Internet infrastructure and to point out the problems that become apparent through its operation. Furthermore, Project JB has collaborated with international networking research groups, such as the Internet2 Project, the NGI Project, and the APAN Project in IPv6, reliable multicasting, QoS, and new applications.
Initially, Project JB focuses on the following three themes--IPv6, multicasting, and quality of service--as three independent subprojects that collaborate with each other. As an early phase objective of Project JB, these subprojects aim to establish a test-bed network that has integrated these three themes. Research and development of new applications will begin after the test-bed is available.
To build a high-speed, high-performance test-bed, Project JB has been designed adopting high-speed lines such as ATM and SDH leased line as a datalink.
The rest of this section will describe the following:
Project JB's subprojects are able to increase or decrease depending on the situation. In other words, they can integrate and divide. We call each of these subprojects a "plane" .
Figure 1 shows that these planes will in the end be unified into one single plane. This is exactly what the next-generation Internet infrastructure is all about.
Figure 1. Multiple planes unite into one plane
In a technological viewpoint, it shows a typical model of the process by which the Internet grows into the infrastructure for universal service. Technologies related to all planes are especially important, such as fusion between old technologies and new technologies or a network management that covers each plane.
At least 10 network operation centers, such as KDD Otemachi, University of Tokyo, Keio Shonan Fujisawa Campus, Osaka University, Kyoto University, Nara Advanced Institute of Science and Technology, Japan Advanced Institute of Science and Technology, Communication Research Lab, MPT, Kurashiki and Kyushu University, and at least 20 leaf-sites have connected to the test-bed network.
Figure 2 shows the initial topology of the JB test-bed network. Most of the connectivity has been provided by ATM links (OC-3 to OC-48), though SDH links have also been used in parts.
Figure 2. "Japan Backbone": nationwide backbone
The KAME stack-based PC-Unix boxes have been installed as both hosts and routers at all JB sites.
The goal of IPv6 plane is building an IPv6 network environment for the next-generation Internet infrastructure.
The IPv6 is a core protocol of the next- generation Internet. It has a 128-bit address space that is enough to cover all worldwide networks. IPv6 core stacks have already been in development for four years. During these years, a lot of network equipment vendors and research and development organizations have developed IPv6 stacks on various platforms. As a result, most of those stacks are available as commercial products or in the public domain.
Likewise, three years have passed since the significant test-bed for IPv6 stack evolution, development, and deployment, known as the 6bone, started operation. As a result of the 6bone, we have developed important technologies such as dynamic routing, source address selection, and multihoming as well as having acquired operational tips and experience.
However, if those technologies are not deployed generally throughout the world, IPv6 cannot be the Internet protocol for the next generation in its true sense. Therefore, the IPv6 plane is going to focus on establishment of a network environment using these technologies as commodity operation.
Through the construction and operation of a test-bed network for IPv6 on JB, the IPv6 plane is going to tackle the following technical issues:
Furthermore, these issues are also applicable to the deployment of IPv6 on the Internet worldwide. Especially with the IPv4-IPv6 transition issues, through actively providing an IPv4 and IPv6 coexisting environment for all JB sites, we will solve new problems and provide feedback to the Internet Engineering Task Force (IETF) ngtrans working group.
Finally, we believe that the current 6bone should be replaced with JB in order to change it from a simple test-bed to an IPv6 nationwide, general commodity infrastructure. This is because for the most part, it consists of complex and ad hoc IPv6 over IPv4 tunnels.
At present, we are going to install KAME stacks, as both router and end-host, "Toshiba CSR" that has been improved based on KAME stack, and "Hitachi GR" as routers to all JB sites.
Since May 1999, we have focused on whether these three individual research planes--IPv6, multicasting, and quality of service--were able to work with each other on the same network. In other words, IPv6 subproject had to provide IPv6 connectivity so that the other planes could work on IPv6.
This happened by July 1999. Currently, we can develop a new application on IPv6 platform at all of the JB sites.
For example, DVTS is one of the most interesting applications that has developed on IPv6. It can transfer DV over IPv4 or IPv6 with or without IPSec or multicast. We often use it as demonstration of Project JB or of a remote lecture between the United States and Japan.
We have cooperated with operation experiments for commercial IPv6 products by allowing them deployment as soon as they are ready for use. Our initial experiences were with the "Toshiba CSR" and "Hitachi GR," which are able to operate with high-speed lines such as OC-3 and OC-12.
Several OSPF version 3 implementations are in progress. They have been tested since the fourth quarter of 1999. At present, there are two candidates for developing platform. One is "gated," the other is "zebra." As both of them are available in the public domain, we will also make available the results of development, such as codes and documents.
As the Internet will have to go through the same transition in near future, the transition of the main IPv6 test-bed of Japan from 6bone to JB is a most interesting event, not only for us, but also for IPv6 researchers all over the world. Actually, this transition is going on at present, and we have had several interesting experiences. We will be supplying feedback on these experiences to the IETF community.
The goal of the multicast plane is the establishment of reliable technologies on a worldwide scale. For example, to provide multicast communication capabilities for high-quality multimedia streaming such as DV and audio(cf. "Internet Telephony"), and for guaranteed reliability.
The most suitable situation to make use of the striking features of multicast such as reduction of redundant traffic and similarity to existing broadcasting media is high-bandwidth communication within a widely distributed area. Therefore, development of IP multicast technology on super high-speed networks is necessary for the deployment of multicast communication in general.
MBone, which is a worldwide test-bed for IP multicasting, has long supported the development of many IP multicasting technologies such as DVMRP, which is a multicast routing protocol based on distance vector algorithms; RTP, which is a multimedia transport protocol for real-time communication; and some reliable multicast protocols for reliable communication.
In this way, MBone has been connected with hosts of various performance and nature. However, there is an important problem.
MBone consists of many tunnels laid out on an Internet based on "best effort." Because of the overhead needed to maintain these tunnels, MBone is not suited for functions such as a high-bandwidth multimedia communication test-bed.
Therefore, Project JB has built a live test-bed for multicasting to experiment on multicast communication with the next-generation Internet environment in mind.
Furthermore, another important study concerning IP multicasting technology is the development and deployment of a scalable multicast routing protocol such as PIM and solving the problems discovered during this process.
In other words, it is for these reasons that IP multicasting has not seen worldwide deployment despite its superior technological value.
On Project JB, IPv6 is ready to operate from the beginning. We assume that it is the core protocol for development and operation.
When considering scalability, we hypothesize several different size scales for the Internet and develop and deploy several appropriate multicast routing protocols for each accordingly.
Practical, reliable multicast protocols are also going to be designed and implemented by making use of QoS guarantee technology. And as an application of these, technology for multipoint delivering will be developed.
Finally, the multicast plane is going to focus on the establishment of high- speed datalinks, IP technology, multicast routing, and applications taken as a whole on the next-generation Internet infrastructure. The technologies developed by Project JB will be adoptable on the next- generation Internet immediately.
At the beginning, we are going to focus on establishing a multicast capable test-bed of about 20 sites on JB, using PIM-SM as the multicast routing protocol.
Also, we are going to focus on a establishment of technologies for high-bandwidth multimedia transport such as DV streaming.
Beginning in April 1999, we focused on a simple implementation of PIM-DM multicast routing protocol daemon. It became available in June 1999, and we deployed it to all JB sites. From July 1999, we began to develop PIM-SM daemons and continue to experiment through the test-bed. Currently, we use PIM-SM as primary multicast routing protocol.
In November 1999, we succeeded in DV multicast streaming over IPv6 from Kurashiki to several JB NOC/leaf sites for a WIDE Project meeting. It provided reasonable quality and did not have any major problems.
We will review and reschedule our plan in March 2000 and then continue to tackle any remaining or new issues.
The goal of QoS plane is building a quality control capable network for the next-generation Internet infrastructure.
There is a growing demand to multiplex voice, video, and data into a single infrastructure, without sacrificing scalability of the Internet. The IETF's differentiated service (diffserv) working group has been trying to solve the problem and standardize protocols for the differentiated services mechanism. In the diffserv framework, interpretation of the IP type-of-service field has been redefined so that various queuing mechanisms and packet-drop mechanisms can be deployed at routers.
However, the standardization of an IP-based QoS framework does not tell us how best to classify packets and prioritize them within the given traffic mix of voice, video, and data.
While a lot of work has been done in the area of QoS, very little work focuses on flow aggregation and their probabilistic QoS guarantees. Little is known about the probabilistic QoS guarantee on aggregation of flows; there is a large possibility that network engineers must fill the gaps between application requirements and the underlying mechanisms, such as traffic class, drop preference, queuing mechanism, and so on.
The opaque nature of the diffserv framework leaves a good number of design alternatives for network engineers. While the standard can be directly adaptable to enterprising networks demanding assured forwarding of voice traffic or expedited forwarding of business transactions, the scale and multiplexing nature of the Internet makes it difficult to apply the standard to the backbones of the Internet. For example, an ISP must be able to handle such questions as "if we prioritize a specific customer, what kind of quality will the other customers receive?"
If diffserv is going to be widely available across the Internet, we must have at least one working service model for the differentiated services.
Project JB is going to bridge the gaps between application practitioners and protocol designers by providing a live QoS infrastructure. Our focus is on both software implementation and network deployment, since it enables fast feedback to the software/protocol development process.
Our goals are as follows: 1) provide open-source implementations for a QoS infrastructure, 2) accumulate working knowledge of QoS-enabled network design, 3) develop algorithms as well as implementations for intra-domain QoS-routing, admission control, bandwidth allocation across multiple administrative domains, and flow aggregation at domain boundary, and 4) provide live QoS-enabled infrastructures to researchers in various science fields.
Since April 1999, we have focused on intra-domain QoS frameworks. Our work spans across routing, forwarding, QoS signaling, and admission control. The bandwidth broker is name of one of the frameworks for QoS developed by organizations such as Internet2. After a one-year focus on intra-domain QoS frameworks, we will be able to move on to implementation and deployment of inter-domain QoS frameworks starting in April 2000.
We are going to develop a service model incrementally through actual deployment. Our initial applications are Internet telephony and DV transmission, as well as live IPv6 data traffic.
Several people are studying the effects of flow aggregation at DS boundary routers.
COPS Common open policy service implementations are in progress. They have been tested since May 1999. Interactions between different types of clients and servers, as well as service location mechanisms, will be tested here.
DSCP marking at DS boundary routers is another issue. While it has been easy to identify packets based on TCP ports, it becomes impossible with IPSec encapsulated packets. Interaction between COPS, IPv6 flow-label, and DSCP will be studied here.
The goal of the new application plane is implementation of and experimentation with application systems and software for the next-generation Internet. There are two themes are in progress. One is DVTS, which consists of two programs, digital video sender and digital video receiver.
The other is School on the Internet, which is a learning and education program based on a real time and archived remote lecture system.
The DVTS is a system that provides DV transmission ability using normal PC and DV VTR/Camera. DV streaming needs 25Mbps bandwidth to transmit at full rate. It is reasonable to evaluate this application for the high-speed/high-performance environment that will be available with the next-generation infrastructure.
Since DVTS is well-design and often used, it can work on both IPv4 and IPv6 and also can work with both IPSec and multicast. So, we are able to apply it various situations.
The SOI Project, with a working group in the WIDE Project, was started in September 1997 because of the change of the role of the Internet in society. Our goal is to provide higher education and opportunity for all the people in the world who have the will to study using Internet-based technologies, eliminating traditional limitations.
In October of the same year, we started WIDE University, School on the Internet, as a test-bed of our new concept (http://www.sfc.wide.ad.jp/soi). Since then, we have been doing research on topics such as education based on the Internet, development of the university environment, and the education system of the new age.
SOI is an environment to learn about the Internet on the Internet. It is difficult for just one educational organization to gather enough teachers that it can teach about this whole new subject and also provide a sufficient educational environment for people who want to learn about the Internet systematically. The establishment of SOI has been an important guideline in setting up this new educational field by coordinating the work of different universities.
The activities of SOI are all on the Internet. Lectures and speeches provided on the Internet by demand are all about Internet and computers and are given by the professors of the WIDE Project. Almost 1,500 people, including university students and other adults, entered this university. The average number of students accessing the lecture page per month has been as high as 200,000.
Project JB is an advanced network research and educational project in Japan. It has focused on high-speed, high-bandwidth, well-designed, and reliable technologies for the next-generation Internet infrastructure. It especially focuses on IPv6, multicast, QoS, and new applications.
It is therefore called the "Japan Backbone" (which is one of the things JB stands for).
Although we described the background, roadmap, and work items of each plane, the JB Project's work is still in progress. However, it will provide robust, high-quality, reliable, and high-speed service for everyone in the near future.