16-06-2014, 12:11 PM
Interconnection of FDDI LANs
[attachment=64922]
Abstract
Local area networks (LANs) have emerged as the
technology of choice for inter-computer communication. As
computers get faster, so must the LAN environment that
supports them. Today there are several successful
competing LANs in the 100 Mbps arena, and we are now
starting to look forward to the next generation of network to
take over when the current generation cannot provide the
necessary bandwidth and connectivity. This paper will
discuss some of the architectural trade-offs available to
designers of the next generation LAN, and proposes an
architecture which provides for convenient interconnection
of multiple FDDI networks (and network segments). The
underlying network architecture is based on FDDI
protocols, an attractive feature for those who have invested
into FDDI. Scaleability appears to be a required feature for
any future network design and is addressed in detail. The
ultimate goal of this research is to present an alternative
next generation networking solution that could be adopted
by ANSI as the FDDI follow on standard.
Introduction
LANs have successfully penetrated most computing
environments today and are largely considered to be
commodity products at this point. Ethemet, with over thirty
million installed nodes, has been the most successful LAN
of the 80’s and will continue to play an dominant role well
into the 90’s. Unfortunately, Ethernet has limited
bandwidth, and many applications, especially video, cannot
operate in a satisfactory manner on this kind of network.
This has led to several emerging new LANs competing in
the 100 Mbps range. Fiber distributed data interface
(FDDI), FDDI-11, dual queue dual bus (DQDB),
asynchronous transfer mode (ATM) and two variants of
FastEthernet are the main players in this 100 Mbps market,
with standards and products slowly becoming available.
The network traffic is continually increasing over time,
and customers are demanding more and more connectivity
from the LAN providers. This has led to much more effort
being put into intemetworking. One way to grow network
throughput is to scale the operating speed of the network.
ATM and DQDB are scalable, and as the push for speed
continues, they will probably benefit from this fact. FastEthemet
is not realistically scalable beyond a few hundred
Mbps and has a very limited geographical reach at that
point. The FDDI community has not chosen an expansion
path beyond the current 100 Mbps standards and is soliciting
new ideas at this point.
It is important to consider the types of services that a
network must provide. Multimedia is poised to become a
major user of advanced LAN services, and will significantly
contribute to LAN traffic levels. Typical multimedia applications
are high definition television (HDTV), medical
imaging, image-mail and voice-mail. The real success of
multimedia will come when it is possible to seamlessly
make all these information exchanges with any of a large
number of users, similar to today’s telephone network. At
the same time the network must also be efficient with packet
type data services, possibly requiring multiple priority
classes.
Multiple priority classes have long been used to provide
services at different quality levels to different users. By
using priorities it is possible to control access to the network
during times of peak load, and guarantee that service
is provided to users in the intended sequence. In cases of
severe congestion, service may be denied altogether to users
with low priority. Many of todays networks, including token
ring and FDDI, provide prioritized access control.
As network capacity demands grow, there are two basic
directions to expand in, increase the operating speed or
interconnect multiple networks. The choice is not obvious,
and involves trade-offs and limitations on the range of
scaleability and it’s applicability to different traffic pattems.
In most designs the increase in operating speed approach is
taken. A major drawback to this approach is, that all of the
equipment on the network has to be upgraded to the higher
operating speed. The main drawback of the interconnected
or parallel network schemes is that more routing and circuit
67
0-8186-6680-3/9$40 4.00 0 1994 IEEE
switching is needed. Most of the work in this research
project will focus on the parallel network scheme.
There are several major benefits to an architecture of
interconnected parallel LANs. First, The network can be
grown incrementally without having to upgrade all the
nodes. Second, in many cases much of the traffic can be isolated
to a single network; this lowers the overall network
traffic. Also, a network failure does not have to cripple nor
even affect network traffic, if the fault can be circumvented
or isolated. Together these features increase the performance
and reliability of LAN environment as a whole. The
main challenges are going to be the design of the intelligent
network interface devices to handle the internetwork routing,
and efficiently managing the different service classes.
A next-generation high-speed network architecture is
proposed here, that will allow for scalable and flexible network
designs at aggregate speeds of up to Gpbs data rates.
The use of an industry standard network, FDDI, as an
underlying network protocol would give the project a realistic
operating environment and a chance at wide spread
acceptance.
This is an extension of the work on FDDI
networks[l, 2, 3,8,9, 11, IO, 12, 17,23,24]. The better
defined networks, such as FDDI [7, 12,201, DQDB [22]
and ATM [6, 16,211, have been extensively covered in the
literature and a few general overview documents are listed
for each. FastEthernet is in the project approval status
within IEEE 802 committee, and has splintered into two
separate subgroups at this point. There has been some architectural
work printed on FDDI follow on network(FF0L)
[18], but the project is too early in the development stage to
be able to do any performance analysis or such, the subcommittee
is dormant at this point and there are no corporate
sponsors. Some interesting work on characteristics of FDDI
protocol at gigabit speeds has been done in [ 131. The thrust
of FFOL seems to be pushing to drop the timed token protocol
in favor of a cell based protocol, although this is not
final.
Some studies have been done one intemetworking
[4,5, 14, 141 and this is an important research field that is
not yet fully understood. Not everybody agrees what multimedia
is, but they agree that it will be an important factor in
the future. High speed networks will be needed to support
the increasing demand presented by multimedia applications
[3, 191. These will include, but not be limited to,
voice, video, graphics, animation and data.
[attachment=64922]
Abstract
Local area networks (LANs) have emerged as the
technology of choice for inter-computer communication. As
computers get faster, so must the LAN environment that
supports them. Today there are several successful
competing LANs in the 100 Mbps arena, and we are now
starting to look forward to the next generation of network to
take over when the current generation cannot provide the
necessary bandwidth and connectivity. This paper will
discuss some of the architectural trade-offs available to
designers of the next generation LAN, and proposes an
architecture which provides for convenient interconnection
of multiple FDDI networks (and network segments). The
underlying network architecture is based on FDDI
protocols, an attractive feature for those who have invested
into FDDI. Scaleability appears to be a required feature for
any future network design and is addressed in detail. The
ultimate goal of this research is to present an alternative
next generation networking solution that could be adopted
by ANSI as the FDDI follow on standard.
Introduction
LANs have successfully penetrated most computing
environments today and are largely considered to be
commodity products at this point. Ethemet, with over thirty
million installed nodes, has been the most successful LAN
of the 80’s and will continue to play an dominant role well
into the 90’s. Unfortunately, Ethernet has limited
bandwidth, and many applications, especially video, cannot
operate in a satisfactory manner on this kind of network.
This has led to several emerging new LANs competing in
the 100 Mbps range. Fiber distributed data interface
(FDDI), FDDI-11, dual queue dual bus (DQDB),
asynchronous transfer mode (ATM) and two variants of
FastEthernet are the main players in this 100 Mbps market,
with standards and products slowly becoming available.
The network traffic is continually increasing over time,
and customers are demanding more and more connectivity
from the LAN providers. This has led to much more effort
being put into intemetworking. One way to grow network
throughput is to scale the operating speed of the network.
ATM and DQDB are scalable, and as the push for speed
continues, they will probably benefit from this fact. FastEthemet
is not realistically scalable beyond a few hundred
Mbps and has a very limited geographical reach at that
point. The FDDI community has not chosen an expansion
path beyond the current 100 Mbps standards and is soliciting
new ideas at this point.
It is important to consider the types of services that a
network must provide. Multimedia is poised to become a
major user of advanced LAN services, and will significantly
contribute to LAN traffic levels. Typical multimedia applications
are high definition television (HDTV), medical
imaging, image-mail and voice-mail. The real success of
multimedia will come when it is possible to seamlessly
make all these information exchanges with any of a large
number of users, similar to today’s telephone network. At
the same time the network must also be efficient with packet
type data services, possibly requiring multiple priority
classes.
Multiple priority classes have long been used to provide
services at different quality levels to different users. By
using priorities it is possible to control access to the network
during times of peak load, and guarantee that service
is provided to users in the intended sequence. In cases of
severe congestion, service may be denied altogether to users
with low priority. Many of todays networks, including token
ring and FDDI, provide prioritized access control.
As network capacity demands grow, there are two basic
directions to expand in, increase the operating speed or
interconnect multiple networks. The choice is not obvious,
and involves trade-offs and limitations on the range of
scaleability and it’s applicability to different traffic pattems.
In most designs the increase in operating speed approach is
taken. A major drawback to this approach is, that all of the
equipment on the network has to be upgraded to the higher
operating speed. The main drawback of the interconnected
or parallel network schemes is that more routing and circuit
67
0-8186-6680-3/9$40 4.00 0 1994 IEEE
switching is needed. Most of the work in this research
project will focus on the parallel network scheme.
There are several major benefits to an architecture of
interconnected parallel LANs. First, The network can be
grown incrementally without having to upgrade all the
nodes. Second, in many cases much of the traffic can be isolated
to a single network; this lowers the overall network
traffic. Also, a network failure does not have to cripple nor
even affect network traffic, if the fault can be circumvented
or isolated. Together these features increase the performance
and reliability of LAN environment as a whole. The
main challenges are going to be the design of the intelligent
network interface devices to handle the internetwork routing,
and efficiently managing the different service classes.
A next-generation high-speed network architecture is
proposed here, that will allow for scalable and flexible network
designs at aggregate speeds of up to Gpbs data rates.
The use of an industry standard network, FDDI, as an
underlying network protocol would give the project a realistic
operating environment and a chance at wide spread
acceptance.
This is an extension of the work on FDDI
networks[l, 2, 3,8,9, 11, IO, 12, 17,23,24]. The better
defined networks, such as FDDI [7, 12,201, DQDB [22]
and ATM [6, 16,211, have been extensively covered in the
literature and a few general overview documents are listed
for each. FastEthernet is in the project approval status
within IEEE 802 committee, and has splintered into two
separate subgroups at this point. There has been some architectural
work printed on FDDI follow on network(FF0L)
[18], but the project is too early in the development stage to
be able to do any performance analysis or such, the subcommittee
is dormant at this point and there are no corporate
sponsors. Some interesting work on characteristics of FDDI
protocol at gigabit speeds has been done in [ 131. The thrust
of FFOL seems to be pushing to drop the timed token protocol
in favor of a cell based protocol, although this is not
final.
Some studies have been done one intemetworking
[4,5, 14, 141 and this is an important research field that is
not yet fully understood. Not everybody agrees what multimedia
is, but they agree that it will be an important factor in
the future. High speed networks will be needed to support
the increasing demand presented by multimedia applications
[3, 191. These will include, but not be limited to,
voice, video, graphics, animation and data.