Internet 2
May 7th, 1998
Juha Eskelin
Department of Computer Science
Helsinki University of Technology
Juha.Eskelin@hut.fi
Abstract
Internet2 is U.S. university based project that aims at developing a
high capacity network which facilitates applications that support mainly
U.S national research and educational purposes. Industry partners take
part in the project in order to spread hardware and software developed
for Internet2 to commercial networks. This paper introduces Internet2 project
as well as concepts and goals defined for this high speed, real-time network
proposed as the successor of current Internet.
Table of Contents
-
1. Introduction
-
2. Internet2 project overview
-
2.1 Organisation of I2 project
-
3. Internet 2 infrastructure
-
3.1 Topology
-
3.2 What is gigapop?
-
3.2.1 Gigapop connections
-
3.2.2 Gigapop functional requirements
-
3.2.3 Routing between gigapops
-
3.2.4 Gigapop example
-
3.3 I2 Connectivity cloud
-
3.4 Local networks
-
3.4.1 Example: upgrade at University of Central Florida
-
4. Internet 2 network services
-
4.1 Quality of service
-
4.2 Multicast
-
4.3 Measurement and optimization
-
4.4 Network management
-
4.5 Security
-
4.6 Network storage
-
5. Internet 2 applications
-
5.1 Application architecture concepts
-
5.1.1 Technical requirements for Internet 2 applications
-
5.1.2 Important architectural concepts
-
5.2 Application types
-
5.2.1 Learningware
-
5.2.2 Digital libraries
-
5.2.3 Tele-immersion
-
5.2.4 Virtual laboratory
-
5.3 Application development schedule
-
5.4 Application example: Provably Secure Videoconferencing
-
6. Conclusions
-
7. Glossary
-
References
1. Introduction
In October 1996, over thirty U.S. universities agreed on establishing a
project, which was aimed to build a next generation network. This network
was projected to be high speed, high capacity, designed to offer sophisticated
network services in order to enable new types of applications to be built.
One of the main goals of these improvements both in infrastructure and
on application level was to enable research and education community to
set future trends in the area of internetworking.
Behind this joint effort of U.S. universities was concern for the situation
of current Internet. Public, commercial and congested Internet was not
able to offer support for the research community as it had done for several
years since 1987 when NFSnet was established.
It was also seen in universities that the current Internet was not able
to offer network services sophisticated enough to carry data in broadband
networks where media integration, interactivity and collaboration are among
the largest application centric traffic areas. Therefore one of the objectives
of the I2 project was to set up a large testbed, partly isolated from the
public Internet, where new types of network services and applications could
be tested. An as soon as proven feasible, I2 project believed that universities
could catalyze in transfer new services and applications to educational
use as well as to corporations and homes.
Figure 1.1 Internet2 logo
To date over 120 U.S. universities take part in the I2 project. Among
the corporate partners of the I2 project are for example 3Com, AT&T,
Cisco Systems, IBM and MCI. Among corporate members and sponsors are e.g.
Alcatel, Apple, Digital Equipment Corporation and Nokia. The I2 project
is also closely related to U.S. federal network projects such as Next Generation
Internet -project (NGI), which aims at developing advanced network technologies.
University members of I2 project are funding the project by over $50
million per year and corporate partners are committed to providing almost
$20 million during the project
[1]. Federal agencies
and organisations participating NGI initiative are also funding the project.
2. I2 Project Overview
Since the I2 project is mainly U.S based, it aims at facilitating and coordinating
the effort ensuring that U.S will maintain it's position as, if not leading,
a very important contributor of the Internet development.
At the early phases of the project the emphasis is on developing the
network capability required by the research community. Besides increased
network bandwidth targets set for network infrastructure include for example
differentiated quality of service (QoS) determined by application requirements.
Network equipment developed within the I2 project scope is encouraged to
be deployed later in the public Internet. This goal implies that solutions
developed should also be affordable in order to guarantee widespread use
of infrastructure technologies.
Main focus of the I2 project is on applications. Capabilities of broadband
network are exploited by developing applications that support research
objectives, media integration, real time collaboration, distance education
to name a few
[2]. Along with applications common middleware
framework and development tools are developed in order to facilitate and
catalyze the application development in broadband environment both in member
organisations and later on publicly.
The I2 project will be conducted in several phases during the following
couple of years. To date the project has made surveys and studies on how
infrastructure should be built in order to serve applications in the best
way. Parts of the infrastructure have been built, there is approximately
twenty operational gigapops, which is a term meaning a network point connecting
local network to backbone network. Gigapop stands for gigabit point of
presence.
2.1 Organisation of the I2 project
The I2 project is organized as two major areas, engineering and applications.
The main concern of the engineering area is how to effectively incorporate
broadband wide area network and how connections from local networks should
be established to the backbone. These issues are discussed in chapter 3
of this paper. The applications area develops concepts for building advanced
applications on top of a broadband network. Applications are discussed
in chapter 5. Between core network infrastructure and applications are
network services, which enable applications to fully exploit the capacity
of the underlying broadband network. Network services are discussed in
chapter 4.
For more detailed level the I2 project is divided into working groups.
Working groups work on specific areas, which relate to the goals of the
I2 project. Currently there are the following working groups:
-
IPv6
-
Measurement
-
Multicast
-
Network Management
-
Network Storage
-
Quality of Service
-
Routing
-
Security
-
Topology
Work done in these is described in more detail below.
3. I2 Infrastructure
Though I2 focus is mainly on developing next generation applications, examples
of which are outlined below, network infrastructure enabling new applications
is required. Therefore one of the fundamental I2 infrastructure design
goals is to specify a "common bearer service" which is analogous to layer
3 in OSI reference model and IP-layer in the exisiting Internet. This bearer
service should support both packet data routing as in current Internet
as well as switched data streams. Switches and routers in I2 should be
capable of supporting at least OC-12 (622Mbit/s) link speeds. Backward
compatibility with existing Internet should be preserved whenever possible.
3.1 Topology
Figure 3.1 shows the overall I2 architecture. From the network point of
view, there are three important entities. These are I2 connectivity cloud,
gigapops (gigabit capacity point of presence) and campus networks.
Figure 3.1 Overall I2 architecture [3]
Each campus or research entity taking part in I2 will install a high
speed circuit from their local networks (referred to as Intra-Campus connectivity
in I2 parlance) to a chosen gigapop (Campus-to-Gigapop). One gigapop will
provide service for several campuses or organisations. 5-10 organisations
can connect to one gigapop, depending on the size of the organisation.
Traffic between organisations connected to one gigapop is called Intra-Gigapop
traffic and it can be routed or switched within the gigapop's equipment.
Gigapop-Gigapop connections construct the Connectivity Cloud shown in Figure
3.1 which is the wide area part of I2. Each of these three elements is
discussed in more detail below.
3.2 What is Gigapop?
A Gigapop is a complex system consisting of both existing technology and
new I2 developed technology. Essentially, it is a regional point of network
interconnection providing access to the inter-gigapop network. Typically
it serves several regional organizations and is operated by those connecting
organizations. Physically, gigapop is a secure, environmentally conditioned
machine room where circuits from both I2 members networks and from WAN
providers terminate. Gigapop does not carry transit traffic between gigapop
and general Internet. Also inter-gigapop links are only allowed to carry
traffic between I2 sites. However, though the key function of gigapop is
to carry I2 traffic with bandwidth and QoS requirements, it is possible
that a gigapop carries standard IP-traffic between ISPs thus eliminiting
the need for I2 members to establish separate connections between members's
campus networks and ISPs. Operational staff on gigapop site is minimal.
No end user support is provided by gigapop operational staff.
3.2.1 Gigapop connections
Roughly, gigapops can be divided into two categories according to their
functionality. These types are
[4]
-
Type 1: Simple, serves only I2 members and routes their traffic to other
gigapops. Little requirements for complex internal routing or firewalling
etc.
-
Type 2: Complex, serves both I2 members and other networks to which I2
members need access, connects to several other gigapops. Need routing policies
as well as authorization mechanisms.
Figure 3.2 illustrates the variety of connections that may be established
to one gigapop.
Figure 3.2 Gigapop connections [4]
ATM switching element shown in the middle of Figure 3.2 may connect
with direct SONET circuits to campus ATM switches or to other gigapops.
Connections may also be full ATM service from commercial providers. ATM
switching elements multiplex with either switched virtual circuits (SVC)
or with permanent virtual circuits (PVC) depending on the characteristics
of the connection. ATM switching element allows optimization and separate
bandwidth allocation for I2 testbed traffic and production traffic [4]
or for other requirements.
IP routing element shown in Figure 3.2 provides the primary service
of a gigapop. It makes all decisions concerning QoS support or IP routing.
IP routing element can feed traffic directly to/from SONET/PPP, high speed
synchronous circuit or PVC/SVC links into ATM fabric.
3.2.2 Gigapop functional requirements
In order to satisfy its key functionality, I2 traffic exchange, gigapops
must satisfy several functional requirements. At least the following aspects
have to be supported by an I2 gigapop [4] :
-
Supported protocols
-
Since common bearer service of I2 is IP, all network layer devices must
support IP. However, though IPv4 is the current standard of IP, it is assumed
that I2 project can help in the emergence of IPv6. Therefore gigapops must
support both IPv4 and IPv6 as soon as stable implementations of the latter
are available. Also other protocols of TCP/IP family are to be supported,
including multicast support protocols like IGMP (Internet Group Management
Protocol) and RSVP (Resource Reservation Setup Protocol).
-
Routing
-
Gigapops must be responsible for either allowing or denying packet flow
via certain link. This requirement is discussed in more detail in section
3.2.3
-
Speed
-
Bit rates to and from gigapop may vary. Gigapops have to be sized in a
manner that enables minimal packet loss within gigapop.
-
Linkage
-
Layer 2 connectivity must support ATM PVC/SVC links as well as raw SONET
links.
-
Costing and use measurement
-
Gigapops have to be able to save and share usage statistics for cost allocation
between I2 members.
-
Enable hierarchical aggregation
-
If necessary it should be possible to aggregate gigapops to form hierachical
structures. This allows capacity scaling whenever required.
-
Technology transfer
-
Gigapops should be designed in manner that enables upgrading to new technologies.
This requires communication between gigapop operators.
3.2.3 Routing between GigaPoPs
As stated earlier, IP is a common bearer service for I2 networks. Both
IPv4 and IPv6 are supported. As gigapops are formed and managed by a consortia
of universities, the I2 network will be built by linking gigapops which
are under separate, but somehow coordinated administration. Furthermore,
it is likely that intra-gigapop traffic will consist of traffic that is
aimed to stay within the gigapop as well as of traffic that is aimed to
flow to other gigapops. Thus, intra-gigapop and inter-gigapop routing policies
are likely to be separate.
As two versions of IP are used as bearer service for I2 network and
support for direct ATM links is encouraged routing has to be done according
to any of these. When QoS based routing in traffic inter-domain traffic
is possible, it requires routing consideration.
Whenever it comes to IPv4, I2 project strongly recommends not to route
anything from commercial Internet to any other gigapop. That is, only traffic
originating or terminating to gigapop's members may be routed to public
Internet. This decision is made in order to avoid congestions of public
Internet. For IPv4 BGP (Border Gateway Protocol) and IDRP (Inter Domain
Routing Protocol) are to be supported as well OSPF (Open Shortest-Path
First Interior Gateway Protocol) eventhough there is no QoS support in
them. Integrated Public Network-to-Network interface (I-PNNI) is one possibility
for QoS based routing protocol for both IP versions as well as for ATM.
For IPv6 routing I-PNNI, IDRP, OSPF, RIPv6 (Routing Information Protocol)
or BGP4++ (Border Gateway Protocol) may be used [3].
Routing on the ATM layer will also be needed, since QoS-related functions
may need dynamic resource allocation which is done at the ATM layer. Both
PVCs and SVC will be used, though the use of SVCs will be prefererable
for example to avoid effects of network problems. Most of the ATM products
have QoS support, but no policy filtering. However, since the amount of
sites working with direct ATM is and will likely to remain small, the use
of PNNI will suffice for the near future.
3.2.4 Gigapop example
As an example case, hardware and software at FloridaNet's Florida Distributed
Gigapop (DGP) is described. Figure 3.3 gives an overview of how DGP is
constructed
FloridaNet Institutional Gigapop
[5]
FloridaNet gigapop consists of one Cisco Lightstream 1010 ATM switch
which serves those who need direct ATM connections. For IPv4, Cisco 7507
Router is used. Router supports also IGMP and RSVP [5].
To satisfy the needs for remote out-of-band access and performance data
collection additional equiment is needed. For intra-gigapop routing, BGP4
is used and upgrades to its successors are done whenever required. Routing
policy is implemented with a pair of route servers. Virtual paths (VP)
are used between each DGP edge site and the ATM switch. Inside the VPs,
virtual circuits are established to provide direct path from each edge
site to every other and to core router. Network operation center (NOC)
services are supplied by the University of Florida. Monitoring data is
captured via RMON, NETFLOW, SNMP and OC3MON. This allows the studying of
the traffic profiles in both normal and problem cases.
3.3 I2 Connectivity Cloud
Connectivity cloud in I2 terminology means interworking between gigapops
and connections from gigapops to backbone network. This implies wide area
requirements considering that the number of gigapops will be limited for
now. Main requirements for inter-gigapop connections include [3]:
-
Very high reliability
-
High capacity (bandwidth)
-
Support for selectable QoS
-
data-collection and circuit management tools for administrators
It is assumed that most usual wide area transports will be provided over
either SONET or ATM signaling. For now, the initial connections are made
with NSF vBNS network. However, over time it is assumed that there could
be other forms of connecting gigapops. For example, if there are specific
bandwidth or service needs, gigapops can build point-to-point links. Point-to-point
links can, for example, be direct ATM. Multicast routing and data transport
are required between gigapops.
In addition to vBNS inter-gigapop linkage to other national network
clouds is possible. This would allow experiments in multi-QoS network of
multiple network providers after QoS problems between single network vendor
gigapop-gigapop traffic have been solved. Multi-QoS traffic in a multi
vendor network is likely to introduce new questions in e.g. accounting.
3.4 Local Networks
Unless campus networks are upgraded to support advanced I2 applications,
effort done on gigapop and WAN level cannot be realized. Requirements at
different campuses may vary. Choosing either a cell-switching campus backbone
or frame-based ethernet solution can be suitable for some campuses. Also
prioritization and bandwidth reservation decisions can be done on either
link or network layer. In practise, most of the campuses may have to transfer
from FDDI to ATM. In any case, upgrading campus networks will cover large
portion of investment required from I2 members.
Generally, every campus will establish a high capacity circuit to the
nearest gigapop and select an advanced functionality router as the campus
gateway. If an experimental service testbed is required, campuses might
also install an ATM multiplexer or switch between the campus and the gigapop.
It is also assumed that campus-to-gigapop connection will carry non-I2
traffic, which is directed to separate link at the gigapop. During 1998
it is assumed that every campus connecting to a gigapop will establish
at least a limited I2 capable network segment which allows testing and
experimenting of I2 applications. By 2000 it is assumed that nearly every
campus backbone is I2 connected [3], even though not
every workstation will support I2 and some of the QoS questions may remain
unsolved. Gradual transfer requires careful planning and management at
the campus level.
3.4.1 Example: upgrade at University of Central Florida
One example of how campus networks can be upgraded to be I2 compliant is
presented here. The university of Central Florida chose to deploy a campus-wide
ATM network. The ATM cloud covers all major campus buildings with 155Mb
OC-3 links. In the data center of the university, there is a pair of Cisco
Lightstream 1010 ATM switches. In campus buildings Catalyst Ethernet 10/100
switches are used as edge devices. They supply either 10Mb shared Ethernet
or 100Mb Fast Ethernet to desktops or labs. If required, direct ATM connections
can be provided to the desktop. [5]
In order to allow testing of I2 applications, an additional research
ATM layer has been added to the production backbone. This is possible since
as a result of rewiring project multiple category 5 cables are available
to each desktop and fiber optic cables are installed to connect the wiring
closets. Both single-mode and multi-mode fiber are installed.
4. I2 Network Services
Many of the network services supported and required in I2 are introduced
in other chapters of this paper and this chapter collects them to provide
more complete view of I2 network services.
4.1 Quality of service
Most of the network services are closely related to quality of service
issues (QoS). In the current Internet there is only one service level,
best effort. For real-time applications this is hardly enough. Therefore
I2 project specifies five QoS dimensions that are likely to be needed in
advanced applications:
[3]
-
Transmission speed
-
The minimum data rate and optionally target average and tolerable maximum
data rates. As an example, an application might request connection data
rate of which never falls below 50Mbps.
-
Bounded delay and delay variance
-
For real-time signals such as video only some limited interruptions are
allowed. An application might request a connection that the gap between
successive packets is 0.4s at maximum.
-
Throughput
-
The amount of data transmitted in given time period.
-
Schedule
-
Starting and ending times for service requested.
-
Loss rate
-
The maximum packet loss rate tolerated within a given time interval
Along with QoS, other questions rise. It is clear that high quality of
service requires more network effort and a larger portion of overall network
capacity. Thus, it is likely that different QoS levels will be charged
differently. As a consequence also costing and accounting mechanisms have
to be supported in the network. Also a user, or an application requesting
high level of service has to be authorized to do so. This requires both
authorization and authentication mechanisms, if individual users are to
be charged for network usage. Authentication itself requires a lot of consideration
and introduces more problems in form of address spoofing and such. One
aspect that should also be considered is the end-to-end property of service
level. Thus, if a site is providing services that require high level of
service, in addition to the user requesting that service also the users
of the service providing site are affected by the QoS support. Since QoS
affects performance of the local network, local users may have to suffer
from lower level of service [6].
4.2 Multicast
If Internet technologies are to be scaled to very large sizes, a good inter-domain
multicast routing has to achieved. Otherwise increases in bandwidth will
be consumed before built. In multicasting area, same issues that are recognized
in unicast inter-domain routing, should be taken into account. Since the
Internet Engineering Task Force (IETF) is active on this area, the I2 project
has chosen to follow the work done in IETF and participate in IEFT working
groups
[7].
4.3 Measurement and optimization
As stated before, different service levels are likely to be charged differently.
This requires that billing elements are included in the network. In the
current best effort networks costs have been predictable since organizations
have paid for bandwidth. When quality of service options are introduced,
cost estimation becomes one concern of organization. Billing schemes and
cost models in the broadband networks are not clear yet. Therefore, one
goal of the I2 project is to develop and test possible cost allocation
methods. This is done according to three rather obvious principles [4]:
-
The cost for a service should be predictable
-
Higher level of service should cost more than lower
-
Resources consumed for accounting should be minimized and therefore accounting
methods should be simple
4.4 Network management
The network in I2 environment consists of multiple separately operated
entities, since each of the gigapops will likely to be operated by its
member organizations. However, from the end users point of view the network
is a single system. Coordinating end-to-end service requests involves multiple
organizations and their cooperation. Currently, network management tools
typically monitor individual links and devices instead of considering performance
from the end user point of view. Management tools need to be developed,
which take into account the end-to-end characteristics in multiple service
levels.
4.5 Security
Some of the security features can be provided on the network layer, though
end-to-end security features will also be needed. Achieving security often
requires compromising usability. In the I2 network environment, at least
the following three categories of security have to be considered [4]
:
-
Network system attacks
-
A malicious person takes actions with a clear intent to degrade services
provided by the network or cause intentional failures in the network. These
attacks result in denial of service from the authorized users and they
can take place in the form of flooding the network, spoofing network control
protocols or illegally accessing network management systems.
-
Unauthorized use of network
-
Since different levels of service are provided, different resource controls
and fees are associated with a level. Network must use authentication and
authorization for obtaining services. Also mechanisms providing authentication
and authorization services have to be immune to attacks.
-
Inappropriate use of network
-
This category covers actions that do not harm the network itself but the
objects available via networks. These are mostly information systems and
breaking into them is also considered under this category.
Achieving and maintaining security requires knowledge of network operators
and also requires coordination in the forms of organizations such as CERT.
In order to avoid misuse, network operators should publish information
on good operating procedures and problem solving.
4.6 Network storage
Network storage such as web caching is also considered as a network service.
In the I2 environment when data amounts transfered can possibly be very
large caching and replication services are even more essential than in
current environment. One aspect of the network storage service working
group is to discuss the definition and use of Uniform Resource Name (URN)
which would allow requesting data from the nearest available location.
5. I2 Applications
The popularity of today's Internet has risen from applications built on
TCP/IP-networks. Today's Internet with MIME e-mail, WWW browsers and newsreaders
is a basis for limited multimedia applications. As the popularity of current
applications has risen, the demand and expectations for more sophisticated
applications have grown. Applications have larger requirements for communication
technologies, that is they require more network bandwidth and multicasting
abilities. They also have larger requirements with respect to computing
power and real-time constraints.
The main focus in the I2 project is to facilitate building applications
which can be used for research and educational purposes. Therefore, investments
done on applications can only be realized after I2 network infrastructure
has spread to all universities, schools, workplaces and of course homes.
5.1 Application architecture concepts
It is assumed that many of the new trends in programming and application
development will affect greatly the environment where I2 applications are
being developed. Such trends are e.g. object-oriented programming, software
components, object request brokering, dynamic run-time binding and multi-tiered
applications
[8]. It is also assumed that fully distributed
computing will emerge during the I2 project. Traditional client/server
architecture is seen as restrictive and network bandwidth consuming [8]
and therefore the use of new realms of computing is encouraged. However,
since the development of application development tehniques is fast and
the I2 project is building the infrastructure at the same time, it is too
early to define exactly what kind of an application architecture will be
chosen. Due to that I2 application area staff only outlines some concepts
that should be taken into account when developing I2 applications.
5.1.1 Technical requirements for I2 applications
I2 applications area recommends the following guidelines to be followed
in applications development projects: [8]
-
Open APIs and services should be used
-
Best practices are promoted to become standard rather than waiting for
slow standardization procedures.
-
Stateful network environment must be assumed by application developers.
-
Applications must be written to be bandwidth-aware
-
For example, different CODECs can be used depending on the network.
-
Quality of service properties of I2 network infrastructure should be
exploited
Taking into account the above mentioned recommendations implies that I2
client software should be running on a high end workstation computer with
multi-threaded, multi-tasking operating system and high bandwidth network
connection. In the client side high bandwidth connection means 25Mbps network
capacity. However, though desktop computer will be the most usual client
software platform, other devices must also be considered as client platform.
Among these devices are e.g. PDAs, portable phones and set-top boxes.
5.1.2 Important architectural concepts
Even though all details of the application development platform are not
yet known, it is understood that applications should be layered in a manner
that allows specific middleware to identified as part of applications.
With the middleware concept multi-tiered applications can separate data,
process and presentation functions [8]. Middleware
concept also allows developing APIs and toolkits that release applications
developers from thinking about the networked environment. Therefore at
least details of network performance and latency as well QoS details should
be implemented as part of the middleware toolkit or as operating system
functionality. Server side models of I2 application include multi-tiered
servers where one client side application can use multiple servers for
different functionality.
As a whole I2 application architecture makes use of the following concepts:
[8]
-
Components
-
Objects, modules and possibly other forms of separate, but interworking
software entities
-
General purpose component classes
-
Providing means for authentication, authorization and validation services
as well as directory and presentation services
-
Brokerage services
-
Services for component, object and informations service location
-
Run-time requesting and binding service which allows components and objects
to construct fully operable applications
-
Application programming interfaces
-
Standardization of general purpose application support and service operations
-
Information resource metadata services
-
High level directory concept allowing application to locate the needed
information resources
-
Network service negotiation capabilities
-
Services allowing applications to negotiate the network related resources
like required bandwidth, QoS and latency levels
-
Integrated network-wide applications performance, costing and utilization
metrics
-
Services allowing utilization metrics such as testing, tracing costing
and performance measurement
Applying these above mentioned concepts and principles in application development
allows building of libraries and service packages that ease post-I2 application
deployment and accelerate the spreading of applications which are desribed
below.
5.2 Application types
There are four main types of applications, developing of which is targeted
in the I2 project. Each of these types is described here briefly
5.2.1 Learningware
Learningware stands for instructional, networked software that can be used
in any level of education. To date there are very few examples of this
kind of software and the I2 project aims at developing an architecture
for delivering and distributing instructional multimedia data. For learningware
most of the general application concepts, as stated above, apply.
Instructional Management System (IMS) is a term for learningware concept
which takes into account the learning process and provides both standards
and services for incorporating multimedia rich instructional material and
learning. The I2 project follows EDUCOM's National Learning Infrastructure
Initiative which will create the standards for IMS on the Internet and
I2. IMS standards will define data elements that belong to IMS applications.
These include learning styles, learning modules and such.
[9]
.
An example of I2 learningware application includes remote jam sessions
of music students and music teachers over high speed network connections
[9].
5.2.2 Digital Libraries
Digital libraries in I2 scope mean access to online catalogs, abstracting
and indexing databases and the content itself such as journals in digital
format. In this application area the I2 project follows ARPA/NASA/NSF-based
Digital Library Program and it's efforts
[10] .
While current Internet is sufficient for certain types of digital libraries
multimedia libraries require more bandwidth and reliability than the current
Internet can offer.
A related research activity on this area is data visualization and applying
visualization analysis to non-textual data. This is also one of the focus
areas of digital library applications.
5.2.3 Tele-immersion
Tele-immersion means a cave style immersion technology as in MUDs and MOOs.
It uses high speed telecommunication for collaboration support. It also
contains methods for recognizing movement and presence in the cave which
allow realistic projection of environment and interaction with other entities
in the cave. Tele-immersion holds potential for numerous types of applications.
For example video-conferences might be replaced with tele-immersion virtual
proximity. Tele-immersion concepts belong to many other I2 application
types but its requirements for network and client software are high.
5.2.4 Virtual Laboratory
Virtual laboratory means a heterogeneous, distributed problem solving environment
[11]
where researchers from different geographical locations can work efficiently
on common projects. Virtual laboratory also contains also the tools required
for specific research area such analysis tools for research data. By the
I2 definition
[11] the virtual laboratory contains
the following components:
-
Computer servers for large scale simulations
-
Databases containing research information
-
Network connected scientific instruments
-
Collaboration tools, maybe containing tele-immersion
-
Software assets for simulation, data analysis, discovery and visualization
Bandwidth and real-time requirements of virtual laboratory applications
are high and multicasting protocols are essential for larger research groups.
5.3 Application development schedule
Since I2 applications rely heavily on network services introduced by the
I2 network infrastructure, application development is scheduled to take
place later. At this point some demos and experiments have been carried
out concerning mainly conceptual matters in I2 applications. The I2 network
has been modeled up to the application level and QoS experiments are done.
Objectives for year 1998 include initial production applications, QoS toolkits,
and large scale experiments
[12] and year 2000 will
bring large scale production applications which should introduce properties
not possible in the current Internet.
5.4 Application Example: Provably Secure Videoconferencing
In I2 member meeting held in October 1997 several application examples
were demonstrated. Most of these were, as assumed by the project goals,
research and education supportive and some of them more general purpose
demonstrating the use of advanced network services. Among the demonstrations
a videoconferencing system that delivers audio and encrypted full-motion
video was demonstrated. This system is developed by the university of Michigan
with IBM and Bellcore. It is built as an extension to VIC, which is videoconferencing
tool used in MBONE. System allows the encryption changed on the fly. Supported
encryption systems include XOR, DES, RC4 and Bellcore's provably secure
VRA
[13].
For this demonstration the desktop computer had switched 10Mb Ethernet
connections to the campus edge system, which in turn was connected to a
gigapop with 155Mb ATM connection. From the gigapop, 155Mb vBNS connection
to other parties of the demonstration was available. The setup and network
equipment used can be seen in Figure 5.1.
Figure 5.1 Secure videoconference network setup
[14]
As an application platform IBM 42T RS6000 with integrated Ultimedia
Services MJPEG was used. Hard coded encryption keys were used, though smartcard
key exchange will also be possible [14]. A specific
video snooper device was included to demonstrate third-party interception
in unencrypted VIC session. Demonstration setup is illustrated in Figure
5.2.
Figure 5.2 Secure videoconference demonstration setup
[14]
6. Conclusions
The I2 project can be thought of as large scale testbed for the successor
of the current Internet. However, in the infrastructure area the project
seems to be traditional rather than modern. Protocols used are IETF standards
and the I2 project has no aims to develop protocols parallel to those of
IETF. Instead, I2 participates in IETF work in order to ensure the compatibility
with the rest of the Internet. However, project seems to be active in the
quality of service area. Applications which are developed in the I2 project
require differentiated quality of service levels and the project feels
that IETF is not active enough in that area.
Eventhough goals defined for the project are rather U.S. centric, it
should be noted that the I2 project also aims at deployment of approved
technologies and concepts in the global Internet. Also, since most of the
corporate partners of the project will eventually manufacture the equipment
for I2 organizations and they work in the global market, I2 technologies
will spread across the public Internet. However, it is odd that no international
co-operation is included in the I2 project agenda. This might cause resistance
in Europe, at least if there are competing, local solutions available.
Gigapop is a concept introduced by I2. The benefit of the gigapop is
to provide a single external connection point for different types of networks.
This concept can simplify organizations' networks and make operations easier.
It also enables smaller organisations to form a collective effort to share
both costs and operation's workload. It can be assumed that the concept
could be adopted by universities across the world. However, it is unlikely
that companies would join their forces to form a shared gigapop. In the
I2 model, the gigapop is operated by its members. This is also unlikely
to happen in the commercial world, but the operation will be provided by
the network operators. Still, the gigapop concept can be useful and it
can to some extent be adopted by commercial network operators as a point
of customer connection.
Establishing a gigapop requires investments. However, there are calculations
showing that gigapop members can achieve significant economic benefits
compared to situation where each of the gigapop members establishes an
individual connection to the backbone network [15].
According to Gartner Group [16] , by 2001, 75%
of the Internet services will be priced by usage. It is also known that
even though there are lot of industrial partners participating the project
paybacks will be made when concepts and equipment developed in the I2 project
are extended to the commercial world. Most of the QoS issues are still
in the drafting stages and it can be assumed that only after deployment
of IPv6, QoS in IP-networks can be guaranteed on a level different than
best effort. One interesting aspect of I2 infrastructure is ATM. ATM has
been seen as complicated and expensive but offering better QoS support
than other alternatives. The I2 project seems to have a great deal of faith
in ATM and if it is found as a viable solution, the I2 project can give
new boost for ATM-to-desktop visions.
Generally, the impact of I2 can be assumed to be notable. Most of the
leading network equipment vendors of the current Internet are involved
in I2 project. As soon as the project produces equipment or protocol specifications
that can be commercially utilized, vendors will add these features to their
products. Therefore, Gartner Group
[16] recommends
that organizations should rely on flat-rate pricing for some three years.
During this period of time, network managers should start to measure and
analyze current traffic to plan the budgets for the switching to usage-based
pricing.
7. Glossary of Terms
BGP
Border Gateway Protocol. An Exterior Gateway Protocol defined in RFC 1267
and RFC 1268. Its design is based on experience gained with Exterior Gateway
Protocol (EGP), as defined in STD 18, RFC 904 and EGP usage in the NSFNet
backbone, as described in RFCs 1092 and 1093.
CERT
Computer Emergency Response Team. An organisation formed by DARPA in November
1988 in response to the needs exhibited during the Internet worm incident.
The CERT charter is to work with the Internet community to facilitate its
response to computer security events involving Internet hosts, to take
proactive steps to raise the community's awareness of computer security
issues and to conduct research targeted at improving the security of existing
systems. CERT products and services include 24-hour technical assistance
for responding to computer security incidents, product vulnerability assistance,
technical documents and tutorials.
Educom
is a nonprofit consortium of higher education institutions that facilitates
the introduction, use, and access to and management of information resources
in teaching, learning, scholarship, and research. Educom believes that
education and information technology (IT) will provide the most significant
enhancements for human capability over the coming decade and that IT will
have a fundamental impact upon education's ability to fulfill its mission.
http://www.educom.edu
MBONE
Virtual Internet Backbone for Multicast IP. IP-Multicast is the class-D
addressing scheme in IP implemented by Steve Deering at Xerox PARC. It
was adopted at the IETF March 1992 meeting and acquired the name MBONE
after the July 1992 IETF meeting. IP Multicast-based routing allows distributed
applications to achieve real-time communication over IP wide area networks
through a lightweight, highly threaded model of communication.
MUD
Multi-User Dimension or Multi-User Domain. Originally "Multi-User Dungeon".
MOO
Mud, Object Oriented. One of several kinds of multi-user role-playing environments,
so far only text-based.
NSFnet
National Science Foundation Network, A high speed hierarchical "network
of networks" in the US, funded by the National Science Foundation. At the
highest level, it is a backbone network comprising 16 nodes connected to
a 45Mb/s facility which spans the continental United States. Attached to
that are mid-level networks and attached to the mid-levels are campus and
local networks. NSFNET also has connections out of the US to Canada, Mexico,
Europe, and the Pacific Rim. The NSFNET is part of the Internet.
NGI, Next Generation Internet
The Next Generation Internet (NGI) initiative is a U.S. multi-agency Federal
research and development program that is developing advanced networking
technologies, developing revolutionary applications that require advanced
networking, and demonstrating these capabilities on testbeds that are 100
to 1,000 times faster end-to-end than today's Internet.
http://www.ngi.gov
UCAID
The University Corporation for Advanced Internet Development (UCAID) is
a non-profit consortium, led by university members working in partnership
with corporate and affiliate members, to provide leadership and direction
for advanced networking development within the university community. http://www.ucaid.edu
vBNS
very high performance Backbone Network Service (vBNS): a network that will
connect up around 100 research institutions -- and already links five NSF
supercomputer centers at -- 2.4 gigabits per second by the year 2000. Begun
in 1995, the vBNS is an investment of up to $50 million in a 5-year National
Science Foundation project with MCI. http://www.vbns.net/
References
-
[1] UCAiD, Internet2 Frequently Asked Questions
<http://www.internet2.edu/html/faqs.html>
-
-
[2] UCAiD, Internet2 Project Mission & Goals
<http://www.internet2.edu/html/mission_and_goals.html>
-
-
[3] UCAiD, Internet2 Preliminary Engineering
Report, Chapter 3. Connectivity specifications and sources, Jan 1997, <http://www.internet2.edu/html/connectivity.html>
-
-
[4] UCAiD, Internet2 Preliminary Engineering
Report, Chapter 2. Gigapops, Jan 1997, <http://www.internet2.edu/html/gigapops.html>
-
-
[5] FloridaNet, High Performance Connections
Grant proposal, Aug 1997, <http://www.internet2.ufl.edu/prop/>
-
-
[6] Berger Peter, Quality of Service Administration
issues, Nov 1997, <http://www.internet2.edu/presentations/QOS-SECU/sld001.htm>
-
-
[7] Mayer David, IP multicast issues, Nov 1997,
<http://www.antc.uoregon.edu/I2/GO97/>
-
-
[8] UCAiD, Internet2 Applications working document,
Architectural concepts, Jan 1997, <http://www.internet2.edu/html/architectural_concepts.html>
-
-
[9] UCAiD, Internet2 Applications working document,
Application Examples: Learningware and Instructional management system,
Jan 1997, <http://www.internet2.edu/html/learningware.html>
-
-
[10] UCAiD, Internet2 Applications working document,
Application Examples: Digital Libraries and Information Access and Distribution,
Jan 1997, <http://www.internet2.edu/html/digital_libraries.html>
-
-
[11] UCAiD, Internet2 Applications working document,
The Virtual Laboratory: An Application Environment for Computational Science
and Engineering, Jan 1997, <http://www.internet2.edu/html/virtual_laboratory.html>
-
-
[12] Hanss Ted, I2 applications priorities,
May 1997 <http://www.internet2.edu/presentations/tedhanss_net97/>
-
-
[13] UCAiD, Internet2 application demonstration,
Oct 1997, <http://www.internet2.edu/oct97/html/provably_secure_video-conferencing.html>
-
-
[14] University of Michigan, Internet2 application
demonstrations, Oct 1997, <http://www.citi.umich.edu/projects/secure_video/Internet2_demo.html>
-
-
[15] vBNS, Gigapop FAQ, <http://www.vbns.net/Gigpop4.htm>
-
-
[16] GartnerGroup, Internet2 Is the First Step
Toward Converged Data Networks Research Note, Jan 1998