Download Asynchronous Transfer Mode

Asynchronous Transfer Mode
Pradeep Kumar
CS 616
Prof. C. Tappert
ATM
Introduction
ATM Overview
Replacement of Traditional LAN technologies
Introduction of ATM specific API
Enhancement of Distributed Computing
Environment Structure
Future of ATM
Conclusion
Bibliography
Introduction
ATM was formulated in the early 80’s, as a result of
AT&T and French Telecompany’s research in Fast
Packets
ITU-T chose ATM for B-ISDN, in the mid 80’s
In 1989, ATM packet size of 48+5 was adopted as a
compromise between 64+5 (USA) and 32+4(Europe)
ATM Forum was founded in 1991
Originally intended for WANs, ATM is gaining
acceptance as a LAN technology, to reduce
bottlenecks in Distributed Computing
We will study the ATM technology and it’s role in
Distributed Computing
ATM Overview
BASIC CONCEPTS
Both a technology and a service (called Cell Relay)
Offers low cost transmission of various types of data – Voice,
Video and Data
Uses fixed, 53 bytes size packets with 5 bytes for header
Handles both Constant Bit Rate(Voice and Video) and
Variable Bit Rate(Data)
Key to ATM’ success is cell switching, which is faster than
traditional multiplexing techniques
Connection oriented
No guarantee of delivery, but does guarantee order of
delivery
Delivers speeds of 155 Mbps to 622 Mbps (compatible with
SONET and can transmit HD TV)
ATM Overview
B-ISDN ATM Model
5
Header
GFC/4
48
ATM Cell
53 bytes
User Data
VPI/8
VCI/16
PTI/3 CLP/1
HEC/8
Header
40 bits
VPI/12 for NNI
OSI
3/4
2/3
2
1
ATM
AAL
ATM
Sublayer
CS
SAR
ATM
TC
Physical
PMD
Functionality
Convergence - Provides standard interface
Segmentation and Reassembly
Flow Control
Cell Header generator / Extractor
Virtual Circuit / Path Management
Cell Multiplexing / De-multiplexing
Cell Rate decoupling
Header Checksum generation and verification
Cell and Frame generation
Packing / Unpacking Cells from enclosing envelope
Bit timing
Physical Network access
ATM overview
The Physical Layer
Deals with medium, voltages, bit timing
Medium is usually fiber optics but for runs less than 100m,
coaxial or CAT5 twisted pair is OK
Each link is unidirectional and two parallel lines are needed for
duplex operation
Connects a computer to an ATM switch or interconnects two
ATM switches
Multicasting is achieved by having the cell leave an ATM switch
on multiple lines
Cells are switched using either input or output queuing
5 bytes header – 4 bytes for Virtual Circuit and control info & 1
for HEC (Header Error Control)
HEC is limited to the first 4 bytes of the header, due to the high
reliability of fiber optics
ATM Overview
ATM Layer
Deals with cells and cell transport, congestion control and
global addressing
Has two interfaces – UNI (User to Network Interface) and
NNI (Network to Network Interface)
Uses Setup, Call proceeding, connect, connect ACK,
Release and Release complete for setup and release of
connections – Virtual Circuits
Offers service categories – QoS
Congestion control through



1. Admission control – fair denial of service to all classes
2. Resource reservation – bandwidth to ensure peak cell rate
3. Rate-based – sender is asked to slow down the traffic
ATM LAN uses LEServer to lookup network addresses
In IP over ATM LES is called ATMARP server
ATM Overview
ATM Adaptation Layer
Provides service to the application layer
Convergence sublayer deals with messages and interface to the
application
Segments varying length messages into 48 or 44 bytes(4 bytes
for AAL info. Optional) each and assembles them at the other end
Offers three types of services



1. Real-Time Vs Non Real-Time
2. Constant Bit Rate VS Variable Bit Rate
3. Connection-oriented VS Connectionless
AAL is further divided into four protocols




AAL1 – Real-Time, CBR and Connection oriented and no error
correction, for uncompressed data
AAL2 – same as AAL1 except for VBR, for compressed data
AAL3/4 – for transport of data sensitive to loss than time
AAL5 for high speed transfer of data
Replacement of Traditional
LAN technologies, with ATM
Comparison of ATM and the traditional LAN environments
such as Ethernet and FDDI, show that Ethernet and FDDI
suffer from high latency and low bandwidth and that only
one host can communicate at a time.
With multiprocessors FDDI speedup is capped at 8
processors and no speedup for Ethernet
ATM is fully bi-directional and offers simultaneous
communication
ATM offers high bandwidth 155Mbps to 622 Mbps and the
delay in the speed of ATM is at the host
Full potential of ATM is achievable, only by improvements
in hardware(bus design & controller implementation) and
software (Protocol stack exploiting the network)
ATM specific API
ATM’s high bandwidth moves the bottleneck from the
network to the Host and Protocols
To overcome the the overhead of the protocols,
applications should be able to access ATM directly.
Two approaches to the ATM specific API


1. Leave transport functionalities to the application
2. Modify message passing libraries
Compared to other APIs(PVM, RPC) ATM API offers best
performance
Modifying message passing libraries to support ATM API
has a drawback – requires a specific version of the library
for each vendor’s ATM API
Enhancement of Distributed
Computing Environment
ATM API level enhancements include very fast RPC models
 Occurs in user space, thus avoiding memory copying
buffering overhead
 Increases complexity for the programmer, to exploit the
interface, which is being hidden, more and more by
applications
Transport level enhancements are made possible by the high
reliability of current networks and include,
 Implementation of Protocol processing in a special
communication processor
 Modified Protocol to implement basic functionality of
message passing library
Application level enhancements, use multithreading to decrease
latency – concurrent thread execution and use of all available
resources allocated on a per task basis
Future of ATM
ATM is finally catching up with the hype due to falling prices and
wider acceptance
Wider acceptance, as a result of telephone and cable
companies’ response to the demands for higher bandwidth
 Sprint offers its ION network based on ATM
 Lucent Technologies bought out Yurie Systems for ATM
equipment
 Cisco acquired Stratacom and Lighstream
ATM is one of the fastest growing network industries - at more
than 60%, despite encroachments by the high powered Gigabit
Ethernet
Experts agree that ATM is here to stay and that, in the network
industry, there is room for both the ATM and Gigabit Ethernet
Conclusion
Advances in network performance have triggered
research in parallel computing using workstation
clusters
High speed networks such as ATM move the
bottleneck to the hosts
Full potential of ATM can only be achieved by
advances in both the hardware and software
Demands of Distributed / Parallel computing are
resulting in ATM technology, not only in WANs but
also in LANs
With increased acceptance by the telephone, cable
and network industries ATM is here to stay.
Bibliography
Regis j. Bate & Donald Gregory
 Voice and Data Communications Handbook, Signature edition, McGraw
Hill, 1998
Andrew S. Tenenbaum
 Computer Networks, Third Edition, Prentice Hall PTR, 1996
David E. McDysan / Darren L. Spohn
 ATM Theory and Application, McGraw Hill, 1995
Abhjit S. Pandya / Ercan Sen
 ATM Technology for Broadband Telecommunications Networks, CRC
Press, 1999
J. Vila-Sallent and J. Sole-Pareta. High Performance Distributed Computing
over ATM Networks: A Survey of Strategies
http://www.atmforum.com/
http://www.npac.syr.edu/users/mahesh/homepage/atm tutorial/p slide.htm
http://www.sic.ohio-state.edu/~jain/netsem/netsem2.htm
http://www.byte.com/art/9608/sec7/art4.htm
http://new.cnet.com/news/0,10000,0-1003-200-329815,00.htm