Below is a formal report, if you would prefer you can view a much more simple PowerPoint presentation of this topic here.
The ever increasing number of internetworked devices and data intensive applications creates great strain on global infrastructure and backbone networks. These functional demands require enormous amounts of data to move at near instantaneous speeds. Thus, the public network (which is the aggregate of backbone and service provider internetworks) requires continued evolution to keep pace with ever increasing flow of data. It is important to have an understanding of how data traverses the public network; the separation between organizational segments and individuals increases as globalization creates new long distance and international connections among people. The decisions made by the public network designers will dictate the speed of business and communication. This survey will analyze trends in design and protocols on the public network over time. By analyzing the strengths and weakness of individual technologies, rational behind emerging technologies can be better understood. This survey will help the student and IT administrator gain insight into the greater networking world, which will assist in effective decision making in terms of service acquisition and application deployment.
Over time, the nature of traffic traversing the public network has become increasingly diverse. Initially, traffic was uniform and highly predictable. Telephone conversations and television transmission generally have static bandwidth requirements; spikes and peak times (nights, holidays, announced events) are usually predictable. However, the emergence of the Internet and corporate site interconnects over public networks has greatly changed the nature of traffic flows. Internet traffic is bursty and often involves multiple conversations between distinct transmitters and receivers over the same paths. Shifts, spikes, and other changes are less predictable because they often are responses to news, recent events, or other factors that the provider cannot anticipate. Also, while most traffic years ago was local (80% local, 20% long haul), the inverse is true today were most traffic is long-haul and is directed between a few major cities (Greenfield, 2002, p. 9). As a result of these changes the public network requires new protocols to meet the changing traffic requirements. Not only has the type of traffic changed, but now there are increasing demands for quality of service (QoS) and new traffic intensive services are emerging.
One of the earliest popular technologies for the public network was Asynchronous Transfer Mode (ATM). ATM was initially developed by a French Telecommunications company in the 1970s as the foundation for their cable television infrastructure. ATM was well suited for both voice and video because it is connection-oriented and offers a higher QoS compared to a protocol such as Ethernet, which has no built-in QoS and is connectionless. Another distinction between ATM and Ethernet is that ATM has a small fixed-length packet; this design reduces delay and jitter, a problem associated with changes in delay (Greenfield, 2002, p. 34). While these characteristics are strengths for voice and video, they cause problems for data based traffic. The current TCP/IP protocol used on the layers above ATM does not map well to the fixed-size of ATM resulting in wasted bits being sent for every 50 bytes of data. In addition, “running IP over ATM requires an enormous amount of [expensive] additional software” (Greenfield, 2002, p. 35). In implementation, ATM achieves maximum performance when its virtual circuits (the logical connections that result from being connection-oriented) are fully meshed. As a result, there is significant routing overhead as the network scales upward (Greenfield, 2002, p. 192). ATM’s popularity is on the decline; however, the phasing out and replacement of this technology has been slow due to the large investment and support made by providers during its inception.
While ATM was designed to handle relatively uniform and simple traffic flows, synchronous optical networking (SONET) claims to be designed as a generic carrier for both smooth and burst traffic. SONET was developed and is used in the United States; it was later incorporated into Synchronous Digital Hierarchy (SDH) as a global standard. The difference between SONET and SDH deals with data rates used in the rest of the world, but for the most part the two are very similar. The overall technology will be discussed based on SONET in this survey. SONET is an optical multiplexing technology that is used for interconnecting multiple points in the public network. Unlike ATM, it is synchronous and based on time division multiplexing. The synchronous nature allows for higher speeds and does not require as much overhead, such as bit stuffing, used in asynchronous communications (Tektronix, p. 5). An important aspect of any technology used for the public network is management and traffic engineering abilities. A SONET frame can contain a great amount of overhead, which allows for “simpler multiplexing and greatly expanded operations, administration, maintenance, and provisioning (OAM&P) capabilities” (Tektronix, p. 16). SONET can carry very large payloads (50MB+) and was designed to easily scale the traditional hierarchal data streams used in telecommunications such as T1 and T3; this greatly simplifies the multiplexing and demultiplexing process because data streams are not divided in an irregular manner. Unlike other optical multiplexing technologies, SONET can exists in topologies besides point-to-point (hub, point-to-multipoint, ring).The ring design is the most common because it offers a high degree of protection and resiliency with a minimal investment in physical wiring (Greenfield, 2002, p. 134). SONET incurs a lower total cost of ownership (TCO) because it is a recognized standard, which increases vendor interoperability. In addition, SONET requires less equipment than other similar technologies which also lowers costs (Tektronix, p. 40). As a general backbone structure, SONET provides many benefits. However, there is limited flexibility in SONET. SONET equipment cannot easily provide lines and different speeds, forcing customers to purchase more than is sufficient for their needs. Another major problem is that when a link is established between two points two symmetric links are created; however, Internet traffic is inherently asymmetrical and bursty, which leads to wasted bandwidth for data traffic. (Greenfield, 2002, pp. 139-140). Despite its shortcomings, SONET has remained popular due to its relative low cost and management features.
As new technologies continue to replace legacy ATM and SONET networks, the business case dictates that their financial costs should not outweigh their functional benefit. One such technology that has garnered great attention is Dense Wave Division Multiplexing (DWDM) for optical networks. Its popularity can be attributed to its ability to provide improved performance using elements of the current public network infrastructure. In a DWDM system, multiple carriers (different wavelengths/colors) transmit independently of each other and can have different traffic characteristics. A system is considered DWDM if it has eight or more carriers in a single fiber. The goal is to maximize the number of available carriers in the existing installed fiber (Horak, 2001, p. 2). Because each carrier is separate, optical switching allows for the signal to be rerouted without electro-optical conversion. DWDM enables very high data rates, easily hundreds of gigabits per second, over one physical medium. While DWDM equipment is significantly less expensive then SONET, there are of course some disadvantages. Each carrier can only accept one type (e.g. PCM Voice or TCP/IP data) at a given time, different streams cannot be mixed. This is ineffective because it leads to underutilization and undercuts QoS because all traffic on a carrier is given the same consideration en route (Horak, 2001, p. 3). While DWDM presents many technical advantages, the costs of replacing highly functional SONET networks may serve as a deterrent. The reader should note that networks can be designed that use SONET over DWDM.
Another technology which hopes to build upon existing infrastructure is Resilient Packet Ring (RPR). RPR was designed to incorporate the best aspects of SONET and Ethernet. One the critical goals of RPR was to create a technology that could serve all types of communications, not solely data or voice. To meet QoS demands there are three traffic classes: Synchronous (high quality voice), guaranteed, and best effort traffic (Greenfield, 2002, p. 157). RPR can easily integrate into the ring architecture that exists with SONET networks. RPR uses at least two rings, operating in opposite directions, which allows for faster end-to-end delivery by taking the shortest path. Another interesting ability of RPR, compared to SONET, is spatial reuse, which is a process where the bandwidth used by one transmission can be reclaimed by another after the first transmission is stripped from the ring. Compared to Ethernet, there is much less intermediary processing of traffic (Green & Schlicht, 2002). RPR’s greatest strength is that it can easily and inexpensively interface with SONET and Ethernet infrastructures and substitute for one or the other depending on the situation.
While many new technologies are being developed for the public network, there are many who would like to see ubiquitous Ethernet deployed end-to-end. Ethernet’s popularity and low cost make it popular; however, it is poorly suited for the public network. As stated earlier, traffic engineering and QoS are important for a mixed multimedia public network, and are not standard features of Ethernet. In addition, Ethernet works best in a meshed environment; however, it presents little advantage in the current networks which are physically rings (Greenfield, 2002, p. 156). Ethernet could possibly be used at the WAN level, but would require extra provisions and overhead to be added for traffic of a connection-oriented nature. However, Ethernet is enticing because it would reduce the number of layers and conversion for end-to-end transmission (Kerner, 2008). For the time being providers will likely continue to use multiple technologies to build out their networks.
The issues concerning the build out and evolution of the public network are numerous. Realistically, there will be no decisive “winner” because providers will make purchasing and implementation decisions based on many factors including cost, user demands, and predictions of the future nature of traffic. TCP/IP continues to grow in importance. If VoIP and IPTV can completely replace traditional telephone and cable television traffic may become more uniform. This could lead to simplified decisions for protocols in the future. With the physical medium being capable of ever higher speeds, overhead and latency become less important. RPR to this point has not been highly adopted, but shows the greatest potential for success. It easily integrates with the public network’s physical infrastructure and can handle multiple traffic flows. Its low price and relative ease of connection to Ethernet make it a strong contender. While it would be desirable, even at high speeds Ethernet lacks the robustness and management capabilities that the public network requires. A successful technology would easily interface with Ethernet and must provide a high level of management and yield a low TCO.
Eogogics Inc. (2005). Tutorial on Optical Networking. Retrieved September 5, 2008, from Eogogics Knowledge Center: http://www.eogogics.com/talkgogics/tutorials/optical-networking
Green, M., & Schlicht, L. (2002, September 3). Maximize the Metro With Resilient Packet Ring. Retrieved September 16, 2008, from CommsDesign: http://www.commsdesign.com/showArticle.jhtml;jsessionid=OBV1GC3HCJ2AAQSNDLPSKH0CJUNN2JVN?articleID=16505799
Greenfield, D. (2002). The Essential Guide to Optical Networks. Upper Saddle River, NJ: Prentice Hall PTR.
Horak, R. (2001, April 3). SONET vs. DWDM. Retrieved September 16, 2008, from Call Center Magazine: http://www.callcentermagazine.com/GLOBAL/stg/commweb_shared/shared/article/showArticle.jhtml?articleId=8704557&pgno=1
Kerner, S. M. (2008, January 11). Ethernet Traffic Doubles, While ATM and SONET/SDH Dip. Retrieved September 16, 2008, from Optically Networked: http://www.opticallynetworked.com/news/article.php/3721101
Optical Networks: DWDM and SONET. (2008). Retrieved September 16, 2008, from The Insight Research Corporation: http://www.insight-corp.com/reports/opticalnetworks.asp
Tektronix. (n.d.). Synchronous Optical Network. Retrieved September 5, 2008, from www.noc.garr.it/docum/Pos/sonet-textronix.pdf