ATM and the Internet

By Sylvester Chen Staff Writer

ntroduction

In the 1992 Presidential Campaign, Al Gore popularized the term "Information Super Highway". The Internet was just about to catch on in the mainstream. In 1993, Mosaic had came out and was dubbed the "killer application." It had transposed computers to the forefront of pop culture. The World Wide Web captured a wide audience because it is simple to use and it allows an easy mixture of text and graphics.

The Information Super High or Internet was first started by Defense Department. They did not develop a network so that million of people can download pictures from web pages or play games over a network. They motives were much more practical. Each branches of the armed forces had contracted different companies to develop their computers and their machines would not talk to each other. They needed network for all the computers they already have.

Beneath all networks are large amount of electrons traveling across physical wires at dazzling speed. You can think of them as electron cars driving across a wire-like-highway. However, all these traffic must follow some traffic rules. We all know to drive in the in the right lane but someone must tell the electrons to travel the correct way, too. Otherwise, accidents and collisions will soon happen. Networks do follow some predefine set of rules and we call them protocols.

ATM is such a network protocol. It has been viewed as a technically superior protocol poising to takeover the current ones. Why hasn't it happen? To examine these issues we must take a closer look at both the Internet and ATM.

World Wide Web & Internet

Although World Wide Web is only one of the services on the Internet, it has become the most popular one. Mosaic lead the charge in 1993 and since then had many spin offs. One of the most publicized decedent is the Netscape Browser. Microsoft has also entered the market in a relatively short period with a competitive browser called Internet Explorer.

The world wide web operates in a client server paradigm. The browsers are front end applications running on a client machine. They interface to servers where the information resides. There are three key types of contents provided on the WWW, Hyper Text Markup Language, Common Gateway Interface and Java. The HTML are files that formats the text when interpreted by a browser and makes it easy to incorporate and integrate other types of contents. The CGI are programs executed on the server to output dynamic information. Many visitor counters on the web are constructed with CGI scripts. Java is newest edition to family. It provides half compiled codes to be executed by the browser.

Conceptually, browsers can even be considered as operating systems. They provide file management like operating systems do. They are equipped to direct access to files residing on remote servers. Modern browsers execute applets and launch local applications to read downloaded data. Microsoft even has the ambition of merging future releases of Windows with their browser.

The web has been able to provide so much content in the most lively and exciting fashion. The high expectations have provoked companies to go online. Some news agencies have been providing news free of charge. There are radio broadcasts carried over RealAudio. Movie clips to an upcoming movie can be downloaded from studio home pages. ESPN even airs baseball games live with animated players. Businesses have begun selling products over the WWW. These materials and the potential for even more, is enticing many manufactures to enter into the market of Web TV and Internet appliances.

With growing amounts of information to be transmitted and the growing number of users who demand it, the network is hitting bottlenecks. Sites that are normally pretty fast slow down dramatically during peak hours. In addition, with a growing trend toward multi-media, larger files will find their way onto the network. With a growing network, a better network architecture has become more of a necessity.

A Network Conversation

Imagine a conversation at a dinner party, the more people are talking the harder it is to hear the conversation. At this point, you are probably speaking at the top of your voice to the person across the table asking her to repeat herself. Computers count on reliable communications and users demand data integrity, so we cannot allow communications to go on like this. To make sure that your friend can hear you loud and clear, as a rule of thumb, only one person is allowed to talk at a time.

The faster a person can finish his piece, the faster the next person gets to speak. When people use the word "bandwidth," they actually mean the speed of communication. Since the electrons move quickly, everyone gets a bit in before the next person talks. A large chunk of information is simply broken down and transmitted in small pieces. This looks as if everyone is talking at the same time. It is as if the information is traveling down a wide highway.

Network Overview

Establishing communications between computers is an inherently complicated task. Network transmission can potentially be slow and error-prone. To simplify the problem, we need to divide and conquer. The standard approach is to partition the responsibility into a layered stack.

The International Standard Organization (ISO) that standardizes goods and services established a seven layer model for networks. Each layer is either its own protocols, or a logical segmentation. When data needs to be transmitted, an application passes it to the network stack. Each layer in the stack processes the information and passes it down to the layer below. Eventually, the information travels across the wire and arrives at the stack of another system. The information travels up the layers where each layer of this system only understands the equivalent layer of the another system.

The application layer interacts with the user. File transfer and remote-login protocols work at this layer. The presentation layer is responsible for setting up socket connections and resolving differences in formats among various sites in a network. Then the session layer implements process-to-process communications. The transport layer maintains packet order, flow control and generating of address. TCP is an example of transport layer. The network layer is the IP layer and also where routers work. It provides routing and decoding of addresses from packets. The data-link layer is where the Ethernet protocol operates. It performs error checking for its lower layer. The physical layer handles the mechanical and electrical details of the physical transmission. These layers however are strictly conceptual. They do not always exist in a particular protocol.

Packet Switching & TCP/IP

In packet switching, data is broken down in small fragments called packets. Their length can vary. Headers are placed around the packet so that the information is reassembled on the end of the network. The switching of the packets are distributing across multiple switches. A piece of information can be sent over different paths. The advantage is that in the event of a failure the information can be routed around it.

The most prominent example of packet switching is TCP/IP. It is also the most popular protocol for the Internet. In typical environment TCP/IP is implemented on top of Ethernet because Ethernet is one of the least expansive LAN alternatives.

Currently Ethernet travels at 10 Mbps and even 100 Mbps have entered the market now. Ethernet implements "Carrier Sense, Multiple Access, Collision Detect." It means that all the machines can be connect along the same piece of wire and each listens to make sure that the line is clear before sending anything. When systems do detect a collision of signals, both parties back off and wait awhile before trying again.

IP consists of higher level functions. The IP number defines the address of a host with a four bit number. Then IP moves a packet from node to node and routes packets from source to destination accordingly. TCP is responsible for reliable in-sequence delivery of packets. In another words, it performs error checking on the data. If some data lost is detected, it triggers the retransmission.

ATM

Asynchronous transfer mode or ATM is a hybrid circuit-switched and packet switched network scheme. The Telephone system is the most prominent example of circuit switching. When two people communicate, a permanent physical link is established. Nobody else can utilize the resources. This, however, is a disadvantage in data transmission because the duration tends to be in bursts. The hybrid scheme is suppose to combine the advantages of both circuit switching and packet switching.

ATM travels at 53 Mbps and up to 650 Mbps over fiber. The potential advantage of ATM is that the most popular Internet protocol (IP) can be implemented over it because ATM can operate within the link layer. However, this standard is still being worked out by the Internet Engineering Task Force.

Cells are the data units of ATM. Cells are not too different from packets except for a set length. ATM standardized the cell size because of the advantage of switching data over hardware. This results in faster traffic flow. ATM stipulates each cell to be 53 bytes in length. The first 5 bytes are the header. It contains address, error control and six other fields for accounting and error checking purposes. The remaining 58 bytes are all data.

A virtual circuit is created between two machines for the duration of a session. The two machines conduct a hand shake, negotiates, and then draws a contract. The contract specifies the bandwidth that is allocated. The path between the two machine is fixed for time period. This creates the appearance of dedicated resources and eliminates time consuming table look up for every packet.

In the negotiation process data is categorized into service classes. In another word, network traffic can be prioritized. There are three traffic types. The highest priority is given to constant bit rate. The network acts as a dedicated circuit to allow a constant flow of traffic. This type traffic is exemplified by voice & video. With variable bit rate, ATM guarantees availability of bandwidth but actual usage may vary. It is intended for video-conferencing in which images do not need to be constantly updated. Available bit rate has no set bandwidth. It uses whatever is left. It is for general traffic such as E-mail that is not time critical.

In order to guarantee satisfactory performances during the negotiation, certain criterions must be considered. They are maximum cell rate (peak cell rate), available cell rate, cell-transfer delay, cell-loss ratio. The other party checks to make sure it can honor this commitment before granting it.

In order to counter contract breaches, ATM has traffic policing. Each cell has a cell loss priority. This bit marks a cell property. If more cells were delivered than agreed upon, then the cell loss priority bit is flipped on for the extra cell. When the network becomes really busy, cells with the CLP bit are discarded. ATM also has plans to implement congestion control by allowing a switch to tell applications how much bandwidth is available. This prevents an application from dumping data unknowingly.

The advantages of ATM is that is has low latency, high throughput. This means that the traffic spends less time waiting with dedicated links and allocated bandwidths. The bandwidth is also used far more efficiently than Ethernet. There is a consistent standard across LANs and WANs which servess corporations trying to build multi-site networks. ATM also allows administrators to adjust for lad patterns. It even has the potential for creating a nationwide bandwidth on demand network. In addition, TCP/IP can be implemented on top of ATM in order to soften the impact of network upgrades and maintain comparability. Although, the standards are still in the works under IETF.

However, not all ATM specifications have been standardized and it seems some crucial components are missing. There is not an API for ATM so that there can be common interface for ATM applications. Currently, it does not support multi-casting so that videos cannot be simultaneously broadcasted. It also lacks TCP/IP's software base. The downside of priorities base ordering is starvation. Users can potentially be locked out when all bandwidth has been given away. These problems still need to be resolved.

Conclusion

Since Telecommunication Deregulation in August of '95, the telecommunication industry has shifted into over-drive. Time Warner merged into Turner Broadcasts. AT&T split up. MCI tried to merge with British Telecommunications. Cable companies are trying to become Internet service providers. The industry is in motion and the competition has been fierce. Eventually, to deliver movies and multimedia and all the materials consumers demand, they will begin to adopt ATM. The high price of ATM will be driven down by its popularity and competition. Nonetheless, the lack of finished high-standard products has created discrepancies among the manufacture of ATM equipment. This crucial weakness needs to be addressed soon. The market has demonstrated that technically superior products are not automatically the guaranteed winner.