Understanding ATM networking and network layer switching, part one
Learn about the issues behind ATM and TCP/IP integration: How is it done? How will vendors do it in the future?
August 1997
![[Next story]](/sunworldonline/icons/tb-next.gif){kind=icon}
![[Table of Contents]](/sunworldonline/icons/tb-toc.gif){kind=icon}
![[Search]](/sunworldonline/icons/tb-search.gif){kind=icon}
|
|
Understanding ATM networking and network layer switching, part oneLearn about the issues behind ATM and TCP/IP integration: How is it done? How will vendors do it in the future? |
August 1997
|
Asynchronous transfer mode (ATM) has made heavy inroads into the data networking world of the Internet and TCP/IP. Network layer switching is a new concept with advanced technological developments behind it; and behind the technology is a lot of political debate. We examine the players and the prospects. (1,900 words)
 Mail this article to a friend |
he network layer provides logical addressing of machines on a large
network. This allows a machine to be recognized across various
networks independent of what the hardware interface may be and
allows the dynamic assignment or manual modification of a machine's
identity on the network. Rather than having to live with the
hardware interface address of the network card of each computer,
network administrators can assign a unique IP address so the machine
can be recognized anywhere in the intranet or even across the global
Internet. This means companies can also use different network
transmission technology within their networks and still have them
exchange traffic.
In our continuing coverage of the world of asynchronous transfer mode (ATM), we have now reached the question of network layer or Layer 3 switching. ATM is a switched circuit-based networking system. This means that each computer must establish a direct line to the computer it wishes to communicate with through however many mechanisms that are in place along the path. Unlike packet-based networks like the Internet, switched circuit networks, like your current analog or POTS telephone line, must be able to establish that clear line across the network between two systems.
The benefit of such a system is stricter control allowing greater security, reliability, and administration. The way ATM was designed also resulted in faster and larger networks. ATM cells of data are transferred through hardware-based switching systems that can work much faster because of the small, fixed-sized cells.
IP packets, on the other hand, very often have unpredictable packet lengths and come in the form of large packets. This results in the packet becoming fragmented across small bandwidth connections often requiring repetitious reassembly at different points on the network. Fragmentation is a well-known phenomenon of IP traffic and is supported in all IP network stacks. Furthermore, each packet has to be checked to determine its destination and any other optional parameters needed for transmission that takes up valuable CPU cycles in processing. Very often the processing that occurs in routers delivering the packets across networks is mostly unnecessary.
In an ATM switched circuit network, all the parameters defining where and how information is to be delivered is arranged ahead of time by the source and destination computers and by all the switches in between. This is akin to the process of dialing a number on your telephone to get to the destination. This set-up procedure takes a short time (in computer cycles that is) and once established, all the switches need to do is forward the packets along the path marked out on the network. They do not have to reprocess each cell as it comes through, and the destination point is already known.
Essentially the amount of time to deliver a cell in a switched network is magnitudinally smaller than that in a routed network. That's not to say that routed networks are worthless; in fact the same properties that make routed networks slower also make them more intelligent.
Advertisements
|
|
|
|
|
Data networking with ATM
ATM was developed to support many different kinds of digital
communications other than simple data transmission. It can also
deliver voice traffic or even video across a network -- and do it very
reliably and in perfect condition. One of the limits of the current
Internet is that the packet-based delivery of information makes it
unsuitable to deliver voice or video at any reasonable quality. Most
Internet telephony products provide low sound quality, and
reliability of audio transmission and video transmission over the
'Net is mostly limited to small fuzzy pictures of poor quality.
Since ATM can provide exactly timed delivery of cells, it can arrange voice or video cells to flow in a smooth pattern between the end points. Quality of sound depends upon the frequency and method whereby it is digitized. For example, CD quality audio is at 44 KHz, whereas AM radio comes in at 8 KHz; and as most would agree, this just barely satisfies a listener's ear. But several multiples of bandwidth are required for CD quality sound delivery in comparison to delivery of AM radio sound.
Data networking, however, is still the most popular use of ATM technology primarily because of vendor focus. ATM's close partnership with synchronous optical networking (SONET) technology has resulted in the development of multigigabit delivery systems. Of course, the greater the bandwidth, the more you can deliver. Although SONET can also work with other networking technologies, ATM/SONET is not only more readily available, it is also a proven environment. Hence, a lot of effort has been put into making data networking protocols such IP work with ATM.
The mapping of IP data to ATM data is mildly complex. Some details of how IP would work over ATM have been figured out during the past few years. However, the ATM standard itself is still undergoing a slow standardization process. Hence, there is still discussion on how to best take advantage of all the features that ATM can provide and incorporate them into IP.
Vendors have decided that IP version 4 can work over ATM by just using that transmission system as a "fat pipe." Simply said, the control available with ATM is for the most part ignored, and basic connections are established between two points with traffic going through large bandwidth pipes of 45 Mbps, 155 Mbps, etc. An IP packet from the source computer goes into a router, then through an ATM switch, through the ATM network to a switch near the other point, out into a router on the destination end, and finally into the destination computer.
As you can see, you still need both routers and switches in this situation. The router decides where to deliver the information and then uses the switches as a fat pipe to a router on the other end. The router converts IP over Ethernet, for example, into IP over ATM before sending it out through an interface to the switch.
Routing and switching
From the scenario we saw above, the integration is fairly simple.
However, this is where economics come into play. The basic question
for most is why we need both routers and switches; why not just
combine them into a single unit? This is exactly what most vendors
are trying to accomplish with Layer 3 switching.
Some have done it in a very basic form by building a switch and adding a router card into the framework or vice versa. However, that only reduces cost by physically combining the units. It would be more beneficial if a vendor could also figure out how to make things go faster through the interfaces. The issues with routing versus switching come into play here.
Although IP packets are different sizes and can go to any number of destinations, many also follow a pattern. If one could figure out this pattern and switch those packets/cells that are deterministic rather than repeatedly route each and every one, then the data would flow much quicker and smoothly. There is some degree of statistical averages and probability to be worked out but none that cannot be overcome.
Now that we know the problem and the theory behind how to fix it, let's look at the solutions that vendors are touting. This whole idea of router/switch integration for IP traffic has been given a name: IP switching. Although this is also the name of a specific technique by one vendor (Ipsilon) it also accurately describes the concept in general. Since ATM fat pipes are already taking over WANs for the Internet and corporations, vendors are paying close attention to this multibillion dollar market. The standard that is approved may come forward to take a large share of this market. It may turn out that several standards are approved. In fact this is such a problem that the whole question of standardization is being brought to its knees.
To give you the heads up let's simply talk vendor economics here and move on to technology implementation later. Cisco is a very large player in the networking world, if not number one. It has developed a technology known as tag switching. Ipsilon, a relatively small company, started the whole movement with its IP switching technology. Newbridge is a strong supporter of the ATM Forum's Multiprotocol Over ATM, a more generic system that allows any number of different network layer systems to function over ATM while getting the full benefits. Still other vendors like 3Com and Cascade have partially sided with Ipsilon's technology. 3Com also has a separate development for the LAN or small WAN known as Fast IP. Cabletron also has its own system.
These are some of the biggest vendors in the networking market. Considering that the future of the market is in hand, none really want to give in to their competitors. Furthermore, there are at least two standardization committees involved: the ATM Forum which drives most standardization efforts for ATM technology today and the Internet Engineering Task Force (IETF), which controls the direction of standardization for the Internet and TCP/IP. Practically all vendors are in both standards bodies as passive attendees, or in other cases, as active participants. It is getting so complicated that egos are now playing heavily in all the action. For example, Cisco representatives are leaders in the IETF working group on IP switching technology which has resulted in political conflict within the group; most fear that Cisco will make an offer the rest cannot refuse -- in the familiar tradition of Don Corleone of the Mario Puzo's movies.
The David and Goliath scenario of Ipsilon versus Cisco is the most interesting of all. Aside from the mythic implications, Cisco's overbearing personality and desire to consume the market has many at odds against them. Ipsilon has gained good ground for such a relatively tiny company. It has products, but it does not hold a significant portion of the market even though its technology does implement and focus on other companies' technologies.
In any case, the possibility of IP switching on a global scale or even a WAN scale is becoming one of political, logistical, and administrative issues rather than one of technological ones. We will take a more careful look at specifics behind IP switching technology in the next issue and at vendor implementations and competing protocols.
|
|
Rawn Shah is VP of RTD Systems & Networking, Inc. He has worked with many different aspects of the LAN world and is currently strongly investigating the world of ATM and DSL and their implementation and implication on the future of voice, video, Internet, and data networking. Reach Rawn at rawn.shah@sunworld.com.
If you have technical problems with this magazine, contact webmaster@sunworld.com
URL: http://www.sunworld.com/swol-08-1997/swol-08-connectivity.html
Last modified:
|
|
Resources
About the author
Rawn Shah is VP of RTD Systems & Networking, Inc. He has worked with many different aspects of the LAN world and is currently strongly investigating the world of ATM and DSL and their implementation and implication on the future of voice, video, Internet, and data networking.
Reach Rawn at rawn.shah@sunworld.com.
|
|
|
|
![[(c) Copyright Web Publishing Inc., and IDG Communication company]](/sunworldonline/icons/b-copyright97.gif){kind=icon}
If you have technical problems with this magazine, contact webmaster@sunworld.com
URL: http://www.sunworld.com/swol-08-1997/swol-08-connectivity.html
Last modified: