|
top
&
A Comparison of
Technologies
View
the Credits:
ATM is a common network protocol intended to
standardize the way information is transmitted globally. ATM also promises to
provide much higher bandwidth to support evolving multimedia requirements such
as video and telephony over the Internet. An ATM network consists of a set of
ATM switches interconnecting point-to-point ATM links or interfaces. ATM
switches support two kinds of interfaces: user-network interfaces (UNI) and
network-node interfaces (NNI). UNIs connect ATM end-systems (hosts, routers,
etc.) to an ATM switch, while an NNI generally connects two ATM switches.
ATM does not provide redundant physical links
like FDDI, with its dual attached stations. Consequently, any end-system
requiring a redundant connection to an ATM network will need to support two
separate UNIs, and either operate one link in a standby mode, or perform local
connection-level load sharing between the links. For this reason, the connection
between a private ATM switch and a public ATM switch is a UNI -- known as a
Public UNI -- since these switches do not typically exchange NNI information.
ATM networks are fundamentally connection
oriented. This means that a virtual circuit needs to be set up across the ATM
network prior to any data transfer. ATM circuits consist of two types: virtual
paths and virtual channels. A virtual path is a bundle of virtual channels, all
of which are switched transparently across the ATM network.
The ATM network is not a standalone network.
Rather, it is interconnected with other networks such as the public switched
telephone network and maybe with mobile networks (cellular). When examining ATM
for quality of service, keep in mind this multi-network image.
The
basic operation of an ATM switch is very simple. First, it receives a cell
across a link (UNI or NNI). Next, the ATM switch looks up the connection value
in a local translation table to determine the outgoing port (or ports) of the
connection and the next link. Finally, the ATM switch retransmits the cell on
that outgoing link with the appropriate connection identifiers. With ATM,
information is divided into 53-byte fixed length cells, transported to their
destination, and re-assembled. (As will be described in a moment, this process
is similar to, but not to be confused with, a packet-switched network.) Being
fixed length allows the information to be transported in a predictable manner.
This predictability accommodates different traffic types on the same network.
The cell is broken into two main sections, the header and
the payload. The payload (48 bytes) is the portion which carries the actual
information (either voice, data, or video). The 5 byte header is the addressing
mechanism. It is the multi-channel nature of ATM that provides the dynamically
controlled bandwidth to which most people attribute its potential success.
Point-to-point connections, which connect two ATM
end-systems, can be unidirectional or bidirectional. Point-to-multi point
connections connect a single source end-system (known as the root node) to
multiple destination end-systems (known as leaves). Cell replication when the
connection splits into two or more branches. Such connections are unidirectional,
permitting the root to transmit to the leaves, but not the leaves to transmit to
the root, or to each other.
What is notably missing from these types of ATM
connections is a parallel to the multicasting or broadcasting capability found
in most LAN technologies such as Ethernet or Token Ring. In such technologies,
multicasting allows multiple end systems to both receive and transmit data
to/from other multiple systems. Such capabilities are easy to implement in
shared media technologies such as LANs, where all nodes on a LAN process all
packets. The parallel function in ATM to a multicast LAN group would be a
bidirectional multi point-to-multi point connection. Unfortunately, no such
capability exists in today's ATM environment.
Despite these early handicaps, there are several advantages that make ATM a
reasonable investment. First, ATM provides one network for all traffic
types-voice, data, video. Next, because of its high speed, ATM allows the
creation and expansion of new applications such as multimedia to the desktop.
Because ATM is not based on a specific standard, it is compatible with most of
today's network protocols. And ATM can be transported over twisted pair, coax
and fiber optics.
Efforts within the standards organizations continue to
assure that networks will be able to benefit from ATM incrementally by upgrading
only those portions of the network that are needed based on new application
requirements and business needs. ATM is evolving into a standard protocol for
local, public, and private wide area services. This standardization will make
network management easier by using the same technology for all levels of the
network. The information systems and telecommunications industries are not yet
standardized on ATM. However, these industries are beginning to focus on ATM as
possibly the networking standard of the near future. ATM has been designed from
the onset to be scalable and flexible in geographic distance, number of users,
access and trunk bandwidths ranging from Megabits to Gigabits.
The benefits of ATM are the following:
- high performance via hardware switching
- dynamic bandwidth for bursty traffic
- class-of-service support for multimedia
- scalability in speed and network size
- common LAN/WAN architecture
- opportunities for simplification via VC architecture
- international standards compliance
The high-level benefits delivered through ATM services
deployed on ATM technology using international ATM standards can be summarized
as follows:
- high performance via hardware switching with terabit
switches on the horizon
- dynamic bandwidth for bursty traffic meeting
application needs and delivering high utilization of networking resources;
most applications are or can be viewed as inherently bursty; data
applications are LAN–based and are very bursty, voice is bursty, as both
parties are neither speaking at once nor all the time; video is bursty, as
the amount of motion and required resolution varies over time
- class-of-service support for multimedia traffic
allowing applications with varying throughput and latency requirements to be
met on a single network
- scalability in speed and network size supporting link
speeds of T1/E1 to OC–12 (622 Mbps) today and into the multi–Gbps range
before the end of the decade; networks that scale to the size of the
telephone network (i.e., as required for residential applications) are
envisaged
- common LAN/WAN architecture allowing ATM to be used
consistently from one desktop to another; traditionally, LAN and WAN
technologies have been very different, with implications for performance and
interoperability
- opportunities for simplification via switched VC
architecture; this is particularly for LAN–based traffic that today is
connectionless in nature; the simplification possible through ATM VCs could
be in areas such as billing, traffic management, security, and configuration
management
- international standards compliance in central-office
and customer-premises environments allowing for multivendor operation
ATM technologies, standards, and services are being
applied in a wide range of networking environments, as described briefly below:
- ATM services—Service providers
globally are introducing or already offering ATM services to their business
users.
- ATM workgroup and campus networks—Enterprise
users are deploying ATM campus networks based on the ATM LANE standards.
Workgroup ATM is more of a niche market with the wide acceptance of
switched-Ethernet desktop technologies.
- ATM enterprise network consolidation—A
new class of product has evolved as an ATM multimedia network-consolidation
vehicle. It is called an ATM enterprise network switch. A full-featured ATM
ENS offers a broad range of in-building (e.g., voice, video, LAN, and ATM)
and wide-area interfaces (e.g., leased line, circuit switched, frame relay,
and ATM at narrowband and broadband speeds) and supports ATM switching,
voice networking, frame-relay SVCs, and integrated multiprotocol routing.
- multimedia virtual private networks and managed
services—Service providers are building on their ATM networks to
offer a broad range of services. Examples include managed ATM, LAN, voice
and video services (these being provided on a per-application basis,
typically including customer-located equipment and offered on an end-to-end
basis), and full-service virtual private-networking capabilities (these
including integrated multimedia access and network management).
- frame-relay backbones—Frame-relay
service providers are deploying ATM backbones to meet the rapid growth of
their frame-relay services to use as a networking infrastructure for a range
of data services and to enable frame relay to ATM service interworking
services.
- Internet backbones—Internet service
providers are likewise deploying ATM backbones to meet the rapid growth of
their frame-relay services, to use as a networking infrastructure for a
range of data services, and to enable Internet class-of-service offerings
and virtual private intranet services.
- residential broadband networks—ATM
is the networking infrastructure of choice for carriers establishing
residential broadband services, driven by the need for highly scalable
solutions.
- carrier infrastructures for the telephone and
private-line networks—Some carriers have identified opportunities
to make more-effective use of their SONET/SDH fiber infrastructures by
building an ATM infrastructure to carry their telephony and private-line
traffic.
ATM
technology will play a central role in the evolution of current workgroup,
campus and enterprise networks. ATM delivers important advantages over existing
LAN and WAN technologies, including the promise of scalable bandwidths at
unprecedented price and performance points for bulk data transfer.
However, these benefits come at a price. Contrary to
common opinion, ATM is a very complex technology. While the structure of ATM
cells and cell switching do facilitate the development of hardware intensive,
high performance ATM switches, the deployment of ATM networks requires
the development of a highly complex, software intensive, protocol
infrastructure. This infrastructure is required to both allow individual ATM
switches to be linked into a network, and for such networks to internetwork with
the huge installed base of existing local and wide area networks.
The ATM Forum (standards group) recently released new
signaling capabilities as part of its UNI 4.0 specification. UNI 4.0 will add
support for, among other things, leaf-initiated joins to a multi point
connection (similar to current LAN capabilities). While some would like to use
this to allow for true multi point-to-multi point connections, note that
signaling support for such connections does not imply the availability of
required equipment. At the time of this writing, it is not clear that UNI 4.0
will have any better solution for multicast with ATM than what exists today.
Addressing schemes, routing formats, and protocol
interfacing pose challenges for the ATM Forum as efforts continue to integrate
this standard across all networks. Security presents another challenge within an
ATM network. It is not at all clear, for example, just how, or whether, it might
be possible to implement firewalls in an ATM environment. The problem is that
once an ATM connection is set up, no intermediate devices interpret or process
the information sent down that connection. Once a connection is set up between
two end nodes, any data can be transmitted through that connection without
visibility to network administrators. While firewalls or other security
mechanisms could be implemented in the end systems, it is not likely to be a
practical solution.
Whether the current state of ATM standardization meets
your particular needs should be the topic of further research. Point-to-point
communications for established media formats is supported today. Broad
distribution of multiple formats through tiered networks is not yet available.
Frame relay is an interface used for wide-area networking.
It is used to reduce the cost of connecting remote sites, in any application
that would typically use expensive leased circuits. The more locations you have,
the greater your savings. With the right carrier service, frame relay network
links may be quickly adjusted to meet changes to applications and network
topology.
Traditional high-speed WANs are built on high-capacity
leased lines, which often take months to order and install. Because network
managers must react quickly when a new application, organizational unit,
business partner or location is needed, leased-line networks often present a
roadblock. Using frame relay, backbone changes can be quickly programmed by the
carrier, saving the long installation delays and high costs associated with
running physical circuits.
WAN designers also have more flexibility when using frame
relay. Whereas physical circuits are typically sized in fractions of 56 or 64
kilobits, virtual circuits may be defined with finer granularity, often in
fractions of 4 kilobits. Virtual circuits are directional, allowing the send and
receive path to be of different sizes if necessary. Frame relay is a form of
statistical multiplexing, a method of dynamically allocating transmission
bandwidth to more efficiently share high-cost links. In this way, a single
access line can support connections to many remote sites.
Frame relay is first and foremost an interface , a
method of multiplexing traffic to be submitted to a WAN. Carriers build frame
relay networks using switches from vendors such as Cascade Communications, Cisco
Systems/StrataCom, Northern Telecom, Newbridge Networks or Bay Networks. As a
customer, your devices see only the switch interface, and are blind to the inner
workings of the carrier network, which may be built on very high-speed
technologies such as T1, T3, Sonet and/or ATM.
As of this writing, all major carrier networks implement permanent
virtual circuits (PVCs). These circuits are established via contract with
the carrier and typically are build on a flat-rate basis. Although switched
virtual circuits (SVCs) have standards support and are provided by the major
frame relay backbone switch vendors, they have not been widely implemented in
customer equipment or carrier networks.
Your carrier programs its frame relay switches to allow
your traffic to pass through the network. These are the essential parameters you
must understand:
- Access rate is the rate at which your access
circuits join the frame relay network. As discussed
earlier , these are typically at 56 Kbps, T1 (1.536 Mbps) or Fractional
T1 (a multiple of 56 Kbps or 64 Kbps). It is not possible to send data at
higher than access rate.
- Port speed is the rate at which the port on the
frame relay switch is clocked. It is not possible to send data at higher
than port speed.
- Committed information rate (CIR) is the amount
of data that the network will receive from the access circuit. All frames
received at this level will be passed.
- Committed burst size (Bc) is the maximum amount
of data that the network will transfer in a burst, defined over a (short)
interval. All frames received at this level will be passed.
- Excess burst size (Be) is the amount of data
above Bc that the network will try to deliver. Frames submitted at this
level may be marked as "Discard Eligible," indicating that they
may be dropped if there is not capacity in the cloud.
- A Data-Link Connection Identifier (DLCI)
identifies a virtual circuit to your equipment. This is an end point for a
PVC. Based on the standard, this number doesn't describe anything more than
that, and has no significance beyond the single link. In practice, however,
many designers assign some significance to it: It often represents either a
site or a circuit.
- Oversubscription An instance where the sum of
CIRs from multiple PVC's to a given location are higher than the port or
access channel rate.
What do you pay for? Port speed is typically the
most costly parameter to increase, though access rates can jump dramatically if
new local loops are involved, such as a move from 56 Kbps to T1. Individual PVC
fees are next costly. Once a PVC is established, the additional cost to increase
CIR is typically small and can be done in small (4 Kbps) increments.
It's important to know that the carrier's backbone is
shared by many users and possibly multiple services. To keep you (and everybody
else) from sending more data than the network can hold, frames sent above your
contracted rate may be marked as Discard Eligible (DE). DE bits are set
by the carrier network, not your equipment. If your equipment receives DE-marked
frames, this indicates that data sent at this rate in the future may get
dropped. This may be an early indicator of traffic rates that you didn't plan
for in the design of your frame relay WAN.
Frame relay equipment notices congestion when it sees
frames marked with the Forward Error Correction Notification (FECN) and Backward
Error Correction (BECN) bits. These merely indicate an overload within the
carrier network, and are only of value in monitoring the carrier's health.
You might expect your equipment to notify end stations to
stop sending data to keep additional frames from being discarded or hitting a
congested network. In practice, however, this doesn't happen: Most routers,
bridges and frame relay access devices (FRADs) do nothing when these bits get
set. Instead, they expect higher layer protocols, such as TCP/IP, to know how to
react implicitly to the packet loss.
Broadcasts over PVC-based networks such as frame relay
create special problems. If the broadcast must go to multiple remote sites
through PVCs as signed to a single access channel, your equipment is forced to
pump it out over each DLCI in turn. As noted elsewhere, you'll want to minimize
the use of broadcasts whenever you can.
Frame relay was initially designed to provide
transport for the delay of insensitive data sent by higher level applications
capable of recovering lost or dropped frames. Frame relay has no inherent frame
correction mechanism; errored frames are simply discarded and higher layer
protocols determine what frames need to be re-sent. Since the frames vary in
length and traverse the buffers and matrices of the switches, delay across the
network is of a non-determinative nature. While prioritization mechanisms are
sometimes employed to provide differential services, to date there are no true
quality of service (QoS) levels available from frame relay networks based on
standard metrics. The Frame Relay Forum and ITU-T (International
Telecommunications Union-Telecommunications Standardization Sector) translation
are attempting to address this issue during the coming year.
In contrast, ATM was built from the ground up to
provide differential QoS levels (e.g., constant bit rate, variable bit rate,
available bit rate, unspecified bit rate) with consistent characteristics (at
least for the constant bit rate and real-time variable bit rate service levels).
Utilizing a fixed-length payload of 48 bytes and a 5-byte header, ATM switches
are able to provide highly determinative service classes each optimized for
specific types of data and applications.
The question then is which technology to use for a
given network?
Frame relay is historically very good at
transporting data which is not highly dependent upon precise delivery intervals.
Examples of this are typical client-server database queries, e-mail and file
transfers, and broadcast video applications. ATM has typically been tagged as
the transport of choice for delay-sensitive information such as interactive
video and voice. However, recent enhancements to the basic frame relay service
have tended to blur some of the distinctions between the two. FRF.11 has defined
standards for carrying voice over frame relay. FRF.12 has defined fragmentation
issues to create a more determinative delay pattern to the frames by chopping up
the large data frames into smaller pieces to better match the size of the voice
frames; thus it is able to minimize frame delay variation through the network.
So, the choice of which technology to use where
depends on the character of the predominant traffic. If your network will be
transporting a high degree of voice traffic and/or near-broadcast quality video,
ATM is most likely the better choice due to its ability to support constant bit
rate (CBR) traffic as well as inherent broader bandwidth (currently up to OC-12,
or 622 Mbps, with plans to extend to OC-48).
If the dominant traffic type is
non-delay-sensitive data (which could be voice, video, fax, imaging, multimedia,
file transfer, or e-mail), frame relay is a better choice even if bandwidth in
the range of 45 Mbps is required. Frame has less overhead than ATM, is more
readily understood by most networking professionals, and is easier to install,
and frame equipment and services are generally less expensive than ATM devices
and services.
Beyond providing a means for frame-based user
equipment to access ATM networks at the user-to-network interface, an important
consideration is how to interwork frame relay networks with ATM networks, and
thereby interwork the network users. This leads to two frame relay/ATM
interworking scenarios; Network Interworking and Service Interworking. These two
interworking functions provide a means by which the two technologies can
interoperate.
Simply stated, Network Interworking provides a
transport between two frame relay devices (or entities).
Service Interworking enables an ATM user to
transparently interwork with a frame relay user, and neither one knows that the
far end uses a different technology.
The Network Interworking function facilitates the
transparent transport of frame relay user traffic and frame relay PVC signaling
(sometimes called LMI protocol) traffic over ATM. This is sometimes referred to
as tunneling. This means that multiprotocol encapsulation (and other higher
layer procedures) are transported transparently as they would over leased lines.
An important application for this interworking function (IWF) is connecting two
frame relay networks over an ATM backbone network.
As shown in the figure above, the ATM network is
used in place of a transmission facility (leased line) to connect the two frame
relay networks. The Network IWF can be external to the networks as shown, but is
more likely to be integrated into the ATM network switch or frame relay switch.
Each frame relay PVC can be carried over an ATM PVC, or all of the frame relay
PVCs can be multiplexed onto a single ATM PVC. This method of connecting frame
relay networks may provide economic savings when compared to leased lines. This
is especially true when the frame relay NNI is operating at a low percentage
utilization.
Network Interworking also includes a scenario in
which an ATM host computer emulates frame relay in the service specific
convergence sublayer.
The Service IWF does not transport traffic transparently. It functions more
like a protocol converter in that it facilitates communication between
dissimilar equipment.
As shown in the figure above, a frame relay user sends traffic on a PVC
through the frame relay network to the Service IWF which then maps it to an ATM
PVC. The frame relay PVC address-to-ATM PVC address mapping and other options
are configured by the network management system associated with the IWF. Again,
the Service IWF can be external to the networks as shown, but is more likely to
be integrated into the ATM network switch or frame relay switch. Note that in
the case of Service Interworking, there is always one ATM PVC per frame relay
PVC.
Frame relay PVC status signaling is converted to ATM OAM cells. Likewise, OAM
cells are converted to frame relay status signaling. Therefore, if a failure
occurs in one network, the user of the other network will be notified. Other
indications such as congestion indication and discard eligible/cell loss
priority are also mapped between networks per PVC.
The Service IWF Maps the frame relay DLCI to the ATM VPI/VCI, the FECN bit
maps to the PT field in which congestion indication is encoded, and the DE bit
maps to the CLP bit.
The frame relay multiprotocol encapsulation procedures (RFC-1490[10]) are not
identical to the ATM multiprotocol encapsulation procedures (RFC-1483[8] and
RFC-1577[9]). When providing frame relay/ATM Service Interworking for
multiprotocol routers, it is necessary for the IWF to convert the multiprotocol
protocol data unit headers from frame relay to ATM and vice versa. This header
processing can be turned on or off per PVC as some applications do not require
it.
Applications of frame relay and ATM technologies overlap. Making a choice
between the technologies must be driven by business considerations. This
includes consideration of:
- availability of equipment and service from multiple suppliers,
- cost of network and user equipment,
- existing installed base of user and network equipment,
- recurring cost of service,
- cost of managing the network (including complexity and required manpower),
- efficiency of bandwidth utilization,
- service classes provided by the network,
- performance (interface speed, delay, throughput),
- interoperability between vendors.
The above considerations are key elements to the selection of network
technology. An informed business decision can be made when the above
considerations are properly weighed and evaluated.
Frame relay and ATM each have fundamentally unique characteristics. One
cannot provide all the features of the other. Therefore, the use of both
technologies will continue to grow to keep pace with the applications for which
they are best suited.
ATM
Internetworking Access at the Customer Premises
International
Engineering Consortium
Telecommunications
Online
Epanorama.net
Saint
Roch Tree
George C. Sackett & Christopher Y. Metz, (1996) ATM and Multiprotocol
Networking, p. cm.--McGraw-Hill series on computer communications
| 
