- What Is MPLS?
- Why Is MPLS Needed?
- How Is MPLS Done?
- When and Where Is MPLS Used?
- Who Is Doing MPLS?
- The Label Switching Paradigm
- A Quick Introduction to MPLS
- Evolution of Internet Network Models
- Basics of the Internet
- Internetworking Technology Basics
- More Basics: Graph Theory and Modeling Language
- The Promise of MPLS
- The Promise of the Promise of MPLS
Internetworking Technology Basics
"Open" and "standards-based" are two related attributes that are highly desirable for the multivendor, heterogeneous device environment that is required for today's networks. These two factors are often cited as the prime reasons for the tremendous growth of the Internet and the evolution of its technologies, including MPLS.
Open Standards-Based Frameworks
The open and standards-based attributes are highly desirable for any communication framework. Open systems can be analyzed and expanded in ways that are not possible with closed, proprietary systems. Open systems allow for the interoperability of multivendor, multiversion devices. Standards-based means that the appropriate enterprise framework must be based on well-established and agreed-upon rules.7 Being open and standards-based allows an enterprise to field hardware and software from many vendors, creating a "best-in-breed" solution. The first efforts in the 1980s for device standardization were defined using the OSI seven-layer reference model (OSI-RM). The OSI-RM is universally used as the lowest common denominator for the basis of explaining how open systems based on standards communicate on a network and how networks can support communications with each other over an internetwork such as the Internet.
The OSI Reference Model
For any communication system to be effective, it must be able to interoperate with the largest number of networks, devices, applications, services, and whatever else it communicates with. Also, for the underlying network to support communications on an internetwork, it must support common and interoperable protocols that allow for this communication.
The OSI-RM is used to explain how open systems are architected. The model was standardized by the International Organization for Standardization (ISO) in 1978 and was approved as a standard in 1983 (OSI guideline IS #7498). It defines the communication between two end-systems in terms of seven layers. Each layer communicates with its corresponding peer-level layer via its protocol. A layer provides a well-defined set of services. Each layer (n) communicates and uses the services of the layer below it (n minus 1) through its well-defined interface. This layering allows the functionality to be modularized and understood in manageable, logical units. Figure 116 shows the layers of the model using two computers connected over an internetwork.
Figure 116 The OSI seven-layer reference model (OSI-RM).
The OSI-RM is often shown with intermediate devices that support just the lower three layers. These devices exist for aiding network traffic flow and do not support the higher-level user services. Examples of intermediate devices include routers, bridges, hubs, and other similar equipment. An MPLS LSR is also an example of an intermediate device.
The seven OSI layers are:
Physical layerThis is the lowest level in the model. This layer handles the interface to the physical medium and deals with the various physical characteristics of the medium such as voltages, data rates, and so on. Physical layers now include optical and wireless technologies.
Data Link layerThis layer provides error-free transmission for the network layer above. It handles flow control, error detection, and data delivery for the link between two connected points.
Network layerLayer 3 establishes, maintains, and terminates the connections between two communicating end-devices. It handles routing, congestion, and other internetworking issues.
Transport layerThis layer ties together the process-to-process communication of the upper three user levels. It guarantees error-free, end-to-end data transfer between communicating devices.
Session layerLayer 5 establishes and maintains the connection between different processes that are running on different machines. It handles connection establishment and data transfer between the sessions.
Presentation layerThis layer handles any data representation, translation, and presentation duties for communicating applications.
Application layerThe highest layer in the model provides user application access to the communication facilities provided by the lower six layers for exchanging data between applications that can be running on different machines.
Before we go deeper into the details of communication technology, it is important to fully understand the key concepts of this reference model. These include layers, services, protocols, and data encapsulation.
The layers of the model were created to handle complexity by abstracting how the various required services would be arranged. Each layer performs a well-defined set of functionality by providing services to the layer above and using the services of the layer directly below. With this in mind, Figure 117 on the next page is another look at the OSI reference model with the levels named by their responsibilities.
Figure 117 The OSI-RM functions.
This view of the model shows that the application layer at one computer knows what it wants to do, that is, what application it wishes to run. In a client/server model, the user at computer A would be the host client, initiating an application that may use a program running at the host server on computer B.
Each layer, in turn, presents the data to each successive layer below it to perform the duties that are required to deliver the information from one application to the application running at the remote computer. The data is successively transformed into the proper format, prepared for the right session, packed for the correct transport end location, routed to the appropriate network, framed to the link format, and then (finally!) transformed into the physical bits that are moved across the communication medium.
At the receiving computer, the process is reversed and the message eventually reaches the corresponding application. It becomes quite evident that this process involves a large number of related and cooperating processes. Each process is modeled as a service or set of services.
With MPLS and conventional routing, the seven OSI-RM layers can be condensed into a simplified model as shown in Figure 118 on the next page. Often, when dealing with layered communications models, it is important to understand that a layer represents a "bundle," or collection of functionality. Because communications models were developed at different times for different purposes (protocols, services, etc.), the number of layers varies, but the idea of what a layer represents remains the same. The best example is that the OSI-RM is seven layers, while the TCP/IP reference model only contains four layers, yet both are used in successful network implementations.
Figure 118 Simplified communication layer model.
ferent times for different purposes (protocols, services, etc.), the number of layers varies, but the idea of what a layer represents remains the same. The best example is that the OSI-RM is seven layers, while the TCP/IP reference model only contains four layers, yet both are used in successful network implementations.
With MPLS, the routing and switching layers contain the key functionality: the control plane where the paths are set up and maintained, and the data forwarding plane where the label manipulation occurs.
Services are the set of well-defined functions (also called "primitives," "operations," or "methods") provided and used by the layers in the model. A layer provides services to the service user in the layer above. The service communicates in an established way through an unambiguously defined network service access point (NSAP).
Figure 119 on the next page shows the general OSI service model. The line between the layer N service provider and its corresponding service user at layer N + 1 is called the "interface." Interfaces separate the layers. Figure 120 further refines the service model by showing the intra-layer relationships.
Figure 119 General OSI service model.
Figure 120 Layer services.
Both the OSI and TCP/IP models include the concepts of layers, services, and protocols. Before these two models and their protocols can be compared, the seven layers of the OSI model need to be further divided into three major parts: the end-to-end services (Layers 1, 2, 3), transport layer (Layer 4), and application services (Layers 5, 6, and 7).
In this view, the lowest layer, end-to-end services, focuses on the data transmission among end-systems across the internetwork communications facility. The upper layer, application services, focuses on the user requirements and applications. The transport layer and its interfaces separate these two types of services. The transport layer's main purpose is to shield the application services from the internetwork details of the end-to-end services below. The relationships between services within the OSI model are shown in Figure 121 on the next page.
Figure 121 Network services model.
Protocols are the sets of rules that control the information flow between two cooperating peer layers. Within each protocol, there is a definition of the data that is passed during communication. The exact definition of the protocol data unit (PDU) is dependent on the protocol. The PDUs include header information and the data portion. Each layer encapsulates the layer above's PDU with its own header. This header provides the information that is required for the communication services at that level. The PDU of the layer above becomes the data portion for the layer below.
Closely related to the topics of layers, services, and protocols is data encapsulation. Figure 122 shows the basic model of data encapsulation when used with layered communications protocols. Encapsulation places the header and data into the data "capsule" of the protocol in the layer below it. There are four protocols in this example. Typically, the only minor exception to pure encapsulation in actual implementations is at Layer 2, where a data link trailer field is maintained for a checksumming feature.
Figure 122 Data encapsulation.
Models can be used to reduce the complexity of a system by decomposing that system into simpler, more understandable representations. Models can be created to group the many components of a system into larger, more abstract building blocks that can be more easily understood by the designers, architects, and analysts who will flesh out these models into designs, and eventually into products. Models can be used to simulate the behavior of systems and their subsystems. In an abstract fashion, models can also represent other functional requirements of a system, such as how data will be defined and stored, how the design will accommodate security, which protocols will be used, and others.
The main idea behind modeling is to make things easier to understand, but there is a danger in complex systemssuch as communication protocols and technologiesto oversimplify. We are kept honest by Albert Einstein's famous adage: "Everything should be simple as possible, but not simpler."C It should be noted that many of the topics presented in this primer are very complex and have filled volumes by themselves. Models, however, have consistently shown themselves to be an excellent starting point when discussing the Internet and its technologies.
The simplest model of the Internet is shown in Figure 123.
Figure 123 A top-level model of the Internet.
The route that communications packets actually traverse involves going through several distinct collections of network devices. In the general model, the hosts are located in the "outer" shell called the local domain, or often, "the last mile." The local domain is connected to an access layer that contains, among other things, the devices to which the home and enterprise computers attach to access the Internet.
The access layer includes customer premise equipment (CPE), routers, digital subscriber line access multiplexers (DSLAMs), Data Over Cable Service (DOCS) terminations, and other technologies that offer these connections. The access layer often uses local area networks (LANs) because the devices are usually in close proximity.
Access networks are often connected to metropolitan area networks (MANs). MANs typically span distances of up to several hundred kilometers and serve large, concentrated urban areas. MANs bridge the service requirements between the wide area, long-haul, regional network carriers and the access networks. MANs are responsible for interconnecting a wide variety of enterprise host traffic in the form of all the protocols that are running in the Internet today such as TCP/IP. Transport technology within the metropolitan area is evolving from T1/T3 TDM to high-speed routers and optical switches for the next generation of multiservice information needs that is currently including the use of MPLS. There is also a new type of MAN called a metropolitan optical network (MON) that uses optical technologies. New directions in the metropolitan area also include the use of gigabit Ethernet and the integration of new optical technologies.
Metropolitan traffic is often sent to regional areas that consist of long-haul carriers and equipment that carries the packet traffic over longer distances. All regional areas connect to the Internet core.
Finally, the heart of the Internet is the core. Here, large devices shunt vast amounts of data as quickly as possible. The backbone of the Internet comprises companies and organizations known as ISPs, which are often ranked in "tiers" depending on their size; a tier one ISP would be the largest. The largest ISPs actually have overlapping areas where their equipment is deployed and must be connected at junctures called network access points (NAPs). Each ISP comprises a set of devices called points of presence (POPs). These POPs are where access layer routers can connect to the ISP. Within each POP, there are actually several types of routers that are used for various purposes. These include access routers, border routers, hosting routers, and core routers. Access routers are used for connecting to remote customers, border routers connect various ISPs, hosting routers connect to various Web servers, and core routers provide inter-POP connections.
The diagram that best sums up these relationships is often called the "onion-skin" model. This view of the Internet is shown in Figure 124 on the next page. The Internet "cloud" can be viewed as a set of concentric circles, with each circle containing a vast array of different host and network devices. This model includes a global communications infrastructure that offers nearly universal access to the services and applications available on the Internet.
Figure 124 The "onion-skin" model of the Internet.
work devices. This model includes a global communications infrastructure that offers nearly universal access to the services and applications available on the Internet.
This top-level model can be further divided into logical models and physical models to delve deeper into the complexities of the devices, protocols, and technologiessuch as MPLSthat are used.
A sample logical model that includes three hosts and seven routers is shown in Figure 125. This model contains two types of network nodes: hosts and network devices. The hosts initiate and terminate applications that use the internetworking communications infrastructure to exchange information with any other hosts that they know the addresses for. The network devices, that is, the routers, connect the various networks that comprise this internetwork example. The routers know how to send packets between other routers to deliver the information from source to destination for any hosts that wish to communicate with each other.
Figure 125 A logical model of the Internet.
information with any other hosts that they know the addresses for. The network devices, that is, the routers, connect the various networks that comprise this internetwork example. The routers know how to send packets between other routers to deliver the information from source to destination for any hosts that wish to communicate with each other.
This logical model can be used to introduce MPLS and how it is beginning to be deployed in the Internet as a new transport method. The routers can become LSRs, and various LSPs can be set up to deliver packet traffic between the three hosts if the network administrators wish to take advantage of the new benefits that MPLS offers. These benefits include the major MPLS applications of TE, QoS, VPNs, and path restoral.
The logical model can be refined to depict a physical model that is closer to an actual internetwork. The physical model based on the logical model of Figure 125 is shown in Figure 126.
Figure 126 A physical model of the Internet.
This physical model contains several new network devices that are included in an MPLS data flow path, but are not MPLS-enabled devices. As the new Internet integrates optical technologies, devices such as DXCs, ADMs, SONET rings, and DWDMs will deliver more packet traffic over longer distances in much less time. In this physical model, for example, MPLS would only be operating in the three routers. These would be LSRs.
To more fully understand the various types of network models, it is important to study some additional basic background information. Two such basic topics are graph theory and a standardized modeling language.