Table Of Contents
Ethernet Access Security Overview
Hierarchically Structured EWAN Design
Hierarchy and Ethernet Service Selection
Containing VLAN, Subnet, and Spanning Tree Domains over the EWAN
Comparing Access Switches and Routers
Flexible Policing and Traffic Shaping
Address Structuring and Traffic Segmentation
Fault Isolation and Traffic Control
SP Value-Added, Service-Friendly Platform
Routers Are a Futureproof Solution
When Is It Acceptable to Use an Access Switch?
Attaching to an EMS with Switches
Attaching to an EMS with Routers
Attaching to an EWS with Switches and Routers
Network Element Level Security
VLAN and Other Layer 2 Security
Design Guide
Enterprise Connections to
Layer 2 Ethernet ServicesAbstract
The purpose of this document is to aid in the design of enterprise WANs that connect via Layer 2 Ethernet services. Topics include architecture, security, and switching versus routing as they pertain to the different Ethernet service types. The appendix contains configuration examples for switches and routers connecting to different Ethernet services. The term "routing" will be used to indicate both Layer 3 routing and Layer 3 switching.
Note:
Although Layer 3 design issues are discussed in this document, they are in relation to Layer 2 Ethernet service, not Ethernet service networks that operate on Layer 3. Cisco Systems already provides many design documents pertaining Layer 3 service offerings; see http://www.cisco.com/univercd/home/home.htm.
Introduction
In the last several years, significant advances in technology—including bandwidth, quality of service (QoS), multicast, and availability improvements—have taken place in the Ethernet LAN. In the LAN, Ethernet has clearly emerged as the dominant technology—due not only to its simplicity, cost advantages, and ubiquity but also because of its incremental speed advances. In the last four years, the industry has seen a jump from a shared 10-Mbps segment to switched 100 Mbps to switched 1 Gbps to multiple lambdas of 1 Gbps and now to 10 Gbps and multilambda 10 Gbps. Service providers looking to provide higher bandwidth, as well as enhanced services such as QoS, are now looking to Ethernet to allow them to scale the bandwidth offered to enterprise customers for WAN and MAN (metropolitan-area network) applications.
Until now, however, Ethernet MANs have remained relatively dormant. Metro service providers have relied mostly on their SONET/SDH infrastructures to provide data services. Although SONET/SDH is clearly well understood and works as specified, it is not optimized for data traffic. As the bandwidth demands of the WAN and LAN have increased, it has become necessary to match LAN and WAN capacity and transmission speeds in the MAN.
Because of the availability, cost, and speed advances in Ethernet, many service providers are looking to offer their customers metro Ethernet as a connectivity option. Many newly established Ethernet service providers (ESPs), are already offering their customers Ethernet connectivity. Many other incumbent local exchange carriers (ILECs); post, telegram, and telegraph operators (PTTs); and InterLATA exchange carriers (IXCs) are considering or offering Ethernet as a Layer 1 private line service, as a pure Layer 2 transport mechanism, or to provide Internet Protocol (IP) and Multiprotocol Label Switching (MPLS) VPN services to complement their existing SONET/SDH, Frame Relay, or ATM services.
Ethernet offers a viable alternative for increasing the capacity of the ATM and Frame Relay services. Many service providers are looking to limit their spending on their aging ATM and Frame Relay services while offering like services with superior functionality, scalability, and lifetime cost ownership.
Ethernet Services
Cisco and the Metro Ethernet Forum (MEF) endorse three main Layer 2 Ethernet service types. The names of the services differ, but their functionality is the same. They are as follows:
•
•
•
When discussing an Ethernet WAN (EWAN), the following terminology should be used (Figure 1):
•
•
•
•
•
Figure 1
EWAN Terms
Ethernet Wire Service
An Ethernet Wire Service is a service that emulates a point-to-point Ethernet segment (Figure 2). This is similar to Ethernet private line (EPL), a Layer 1 point-to-point service, except the provider edge operates at Layer 2 and typically runs over a Layer 2+ network. The EWS encapsulates all frames that are received on a particular UNI and transports these frames to a single-egress UNI without reference to the contents contained within the frame. The operation of this service means that an EWS can be used with VLAN-tagged frames. The VLAN tags are transparent to the EWS (bridge protocol data units [BPDUs])—with some exceptions. These exceptions include IEEE 802.1x, IEEE 802.2ad, and IEEE 802.3x, because these frames have local significance and it benefits both the customer and SP to terminate them locally.
Figure 2
EWS Example
Since the service provider simply accepts frames on an interface and transmits these without reference to the actual frame (other than verifying that the format and length are legal for the particular interface) the EWS is indifferent to VLAN tags that may be present within the customer Ethernet frames.
EWS subscribes to the concept of "all-to-one" bundling. That is, an EWS maps a port on one end to a point-to-point circuit and to a port on another end. EWS is a port-to-port service (Figure 3). Therefore, if a customer needs to connect a switch or router to n switches or routers it will need n ports and n pseudowires or logical circuits.
Figure 3
Nonservice Multiplexing Example: Each Destination (Left) Needs Its Own Port (Right)
One important point to consider is that, although the EWS broadly emulates an Ethernet Layer 1 connection, the service is provided across a shared infrastructure, and therefore it is unlikely that the full interface bandwidth will be, or needs to be, available at all times. EWS will typically be a sub-line rate service, where many users share a circuit somewhere in their transmission path. As a result, the cost will most likely be less than that of EPL. Unlike a Layer 1 EPL, the SP will need to implement QoS and traffic engineering to meet the specific objectives of a particular contract. However, if the customer's application requires a true wire rate transparent service, then an EPL service—delivered using optical transmission devices such as DWDM (dense wavelength division multiplexing), CDWM (coarse wavelength division multiplexing), or SONET/SDH—should be considered.
Ethernet Relay Service
Ethernet Relay Service is similar to EWS in that it offers point-to-point connectivity. The key differentiation between EWS and ERS is that an ERS uses a VLAN tag to multiplex several, non-same-destination pseudowires to one port. That is, unlike EPL and EWS, ERS is a "one-to-many," multiplexed service. Service multiplexing simply means that multiple pseudowires utilize a single access interface or UNI. These circuits can terminate within an L2VPN or on, for example, an Internet gateway. From the service user's perspective, this service multiplexing capability offers more efficient interface utilization, simplification of cable plant, and reduced maintenance costs associated with additional interfaces.
Using the same example as above, where a router connects to n other routers, the source router only needs one port for the service instead of n, as is the case with an EWS. The service need not be port-to-port, but can be logical-pseudowire-to-logical-pseudowire. In the case of an ERS, each circuit can terminate at a different remote location (Figure 4), whereas using EWS, all frames are mapped to a single circuit and therefore a single egress point.
Figure 4
ERS Service Multiplexing Example: One Port (Left) Can Be Used for All Destinations (Right)
Like Frame Relay, ERS allows a customer device to access multiple connections through a single physical port attached to the service provider network. The service offered by ERS can be thought of as being similar in concept to Frame Relay, in that a VLAN number is used as a virtual circuit identifier in a similar fashion to Frame Relay data link connection identifier (DLCI) or an ATM permanent virtual circuit (PVC). Unlike EWS, ERS does not forward BPDUs, because IEEE 802.1Q (VLAN tagging) only sends BPDUs on a default VLAN. In a hub-and-spoke network, only one spoke at most would receive BPDUs, thus breaking the spanning tree in the rest of the network. Therefore, an ERS does not transmit any BPDUs and runs routing protocols instead of Ethernet Spanning Tree. The routing protocols give the customer and provider greater flexibility, traffic determination characteristics, and value-added services.
Ethernet Multipoint Service
An Ethernet Multipoint Service (EMS) differs from EWS and ERS in that an EMS provides a multipoint connectivity model. It should be noted that an EMS service definition is still under review within the IETF Virtual Private LAN Service (VPLS) working group. Although EMS uses a multipoint model, it can forward unicast packets to single destinations; that is, it also supports point-to-point connections. To the end user, the network looks like a giant Ethernet switch where each customer has their own VLAN or broadcast domain, rather than end-to-end pseudowire link(s) (Figure 5).
Figure 5
EMS Example
An EMS does not map an interface or VLAN to a specific point-to-point pseudowire. Instead, it models the operation of a virtual Ethernet switch: EMS uses the customer's MAC address to forward frames to the correct egress UNI within the service provider's network. An EMS emulates the service attributes of an Ethernet switch and learns source MAC to interface associations, floods unknown broadcast and multicast frames, and (optionally) monitors the service user's spanning tree protocol. One important point to note is that although the service provider may utilize spanning tree within the transport network, there is no interaction with the service user's spanning tree.
This service works similar to an MPLS VPN, except it functions at Layer 2 instead of Layer 3. While a VPLS EMS is a viable solution, its scalability and QoS control are suspect compared to that of MPLS VPNs. In addition, it is much more difficult, and may be impossible, for the service provider to offer value-added Layer 3 services (this is discussed later in the document).
Finally, emulating LANs in the metro requires a lot of overhead. EMS and protocols run the risk of turning into ATM LAN Emulations (LANE), which have shown their overcomplexity and inability to scale.
Ethernet Access Security Overview
This section discusses security as it pertains to Layer 2 Ethernet as well as the myth that because an access device operates at Layer 2 it is much more secure than one that operates at Layer 3. This discussion assumes that the customer premises equipment (CPE) is an untrusted network element (NE). The CPE can be a switch or a router.
It is further assumed that the service provider network has been protected against "inside" attack, using procedures to secure access to network devices and that a number of measures have been taken to protect against external attacks against the service provider network. These precautions include hiding the IP core topology and deploying security measures such as packet filtering.
The two main threats to the access network are denial-of-service (DoS) attacks and spoofing attacks.
Denial-of-Service Attacks
DoS attacks are intended to bring a network to a state in which it can no longer carry legitimate users' data. Such attacks commonly take one of two forms: attacking network components or flooding the network with extraneous traffic. An attack is designed to cause a component to stop forwarding packets or to forward them improperly. Network attacks can take the form of a misconfiguration or the injection of a spurious update. A flood attack bombards a device with unroutable or unswitchable packets, causing its performance to degrade. A flood attack on a network is similar to a flood attack on an individual device, except that the flooded packets are usually broadcast.
DoS attacks include the following types:
•
•
•
•
Spoofing Attacks
The objective of DoS attacks is to make a device or network unusable—a state that will be quickly detected by a network's users and administrators. In contrast, spoofing uses a spurious update to cause packets to be routed to a host, from which an intruder may monitor the data in the packets. After examination, these packets are usually reswitched (routed) to their correct destinations. This is known as a "man in the middle" attack. The intruder may or may not have altered the contents of the packets, so these attacks are not always perceived by other network users. In an Ethernet environment different types of spoofing attacks are possible. Attackers can take advantage of the ability to spoof both IP and MAC addresses to carry out DoS attacks and avoid traceability, to hijack a service and avoid billing, or to eavesdrop on traffic.
Types of spoofing attacks include the following:
•
•
•
Hierarchically Structured EWAN Design
The development of Layer 2 switching in hardware several years ago led to network designs that emphasized Layer 2 switching. These designs are often characterized as "flat," because instead of relying on the logical, hierarchical structure and summarization provided by routers, they are most often based on the campuswide VLANs model, where a set of VLANs span the entirety of the network. This type of architecture favored the "departmental segmentation approach," in which, for example, Marketing and Engineering needed to exist on the same broadcast domain to avoid crossing "slow" routers. Since these departments could exist anywhere within the network, VLANs had to span the entire network. However, Layer 3+ switching provides exactly the same advantages as routing with the added performance boost from packet forwarding handled by specialized hardware. Adding Layer 3 switching in the distribution layer and backbone of the campus network segments the campus into smaller, more manageable pieces, as defined in several different ways. This approach also eliminates the need for networkwide VLANs, allowing for the design and implementation of a far more scalable architecture. This design strategy is directly applicable to large EWAN networks.
The foundation of the multilayer design is the building block, or module (Figure 6). Smaller networks will likely need only one module, while larger scale networks can use several. The basic building block comprises two layers, the access layer and the distribution layer". When scaling from a smaller to a larger network, a core layer is included as well. The access layer collects the users or sites and presents them to the network. The distribution layer collects the access links on a per-region basis. Intraregional traffic can be forwarded directly by the distribution layer. In the case where access sites are small branch or SOHO sites, the distribution layer provides services such as firewalling and policy management. The core layer is used for high-speed forwarding between regions or distribution points. Services such as firewalling, policy management, server farms, and Internet POP handoff occurs at core sites.
Figure 6
Hierarchical EWAN Network
Figure 6 shows seven sites: three access, two distribution, and two core network elements. In actuality, such an network would probably have 20 or more access nodes to support the two layers shown. In the case of smaller networks, the core layer and distribution layer can be collapsed into a single layer. Still smaller networks can be collapsed to the access layer only; here each access node would make up its own domain. Regardless of the size of the network and number of layers, the design principles are the same, even if the number of layers differs.
Access links are usually single links to the EWAN. However, multiple links running IP load balancing, Cisco EtherChannel® technology, or spanning tree can be used for redundancy. In addition, access links are usually not multihomed to more than one distribution point. However, they may be multihomed for high-availability sites. Core links are usually meshed based upon traffic load, and there should always be more than one path to and from each node, with no single point of failure.
The multilayer design model is highly deterministic and fault tolerant. It is also easier to troubleshoot than a flat network, and its scalability advantages are unrivaled. The modular building-block approach scales easily as new buildings or server farms are added to the enterprise. Intelligent Layer 3 routing protocols such as OSPF can handle load balancing and fast convergence across the backbone. Many value-added services in Cisco IOS® software, such as route summarization, DHCP relay, and intelligent multicast handling, are implemented in the Cisco Catalyst® multilayer switches at the distribution layer. Access policies are also implemented with access lists at the access, distribution, or distribution switches.
Hierarchy and Ethernet Service Selection
Simply stated, if the enterprise network requires a dual hierarchy (access, distribution, and core), then an ERS should always be used. If an EMS is used (Figure 7), the network turns into one large broadcast domain, where the access switches and routers will bypass the distribution routers and peer directly with the core routers. This will wreak havoc with traffic patterns, security, and any special services run within the enterprise.
Figure 7
EMS Will Cause Routers to Form Meshed Adjacencies, Ruining a Hierarchical Network
One way around this problem would be to have multiple instances of EMS between core and distribution routers and distribution and access routers. However, this is not recommended, because it would cost extra to support multiple instances of a single service. Also, the number of ports needed on the distribution routers will at least double, because they have to attach to both layers in the hierarchy. Therefore, if the network is large enough to need a dual-layer hierarchy, ERS is the service of choice, because it allows the circuits to connect between two specified and distinct points.
Containing VLAN, Subnet, and Spanning Tree Domains over the EWAN
A Layer 2 switched domain can be considered a failure domain because a misconfigured or malfunctioning workstation, server, or switch can negatively affect or disable the entire domain by flooding it with broadcasts or undesirable frames. A protocol malfunction (spanning-tree error or misconfiguration) can inhibit a large part of a network. Problems of this nature can be very difficult to localize, especially in a flat network.
The scope of a failure domain should therefore be reduced as much as possible. The best way to achieve this is by restricting its scope to a single EWAN link. In other words, only one unique VLAN should exist per EWAN link. Over the WAN, instead of traffic types being differentiated with VLAN IDs, traffic will be differentiated by class of service (CoS), type of service (ToS), or differentiated services code point (DSCP).
IP subnets come into play if routers are used. If all the network elements attaching to the EWAN are switches (not recommended) the IP structure of the enterprise is simply superimposed over the WAN. In a network where all the WAN network elements are routers, each access link should have a unique IP domain (prefix). At the distribution layer, these prefixes should be summarizable to a single prefix and in turn summarizable to the core. In a hybrid network where one or more layers are switches, unique and summarizable IP domains should be configured at the lowest layer where routers are present. By implementing a sensible IP addressing scheme, Layer 3 switches gain the ability to exchange summarized routing information, rather than having to learn a path to every host in the network. Summarization is the key to the scalability of routing protocols such as Open Shortest Path First (OSPF) and Border Gateway Protocol (BGP).
Containing VLANs and summarizing IP addresses is a good start. However, if the two are not synchronized, most of their benefits may be lost. All the EWAN IP subnets should map to the Layer 2 VLANs; therefore, if the IP subnet is the logical Layer 3 equivalent of the VLAN, at Layer 2 one VLAN = one subnet. The IP network address is defined at the Layer 3 switch where the Layer 2 switch domain terminates. This one-to-one mapping contains MAC-layer broadcasts, multicasts, and unknown unicasts flooded throughout the Layer 2 domain, such as those that could flood the EWAN and cripple an entire organization. In addition, intelligent, protocol-aware features of routers will further contain broadcast packets. Flooding of multicast traffic can be constrained to a community of interested ports by using Internet Group Management Protocol (IGMP), IGMP snooping, and Protocol Independent Multicast (PIM).
Another advantage of the one-to-one mapping of EWAN link-IP subnet-VLAN is that if spanning tree is used, the spanning tree domain is contained between each EWAN network element. Although Hot Standby Router Protocol (HSRP), IP load balancing, and EtherChannel technology provide better link utilization and failover times than spanning tree, when spanning tree is used for redundancy, the spanning tree domain is limited to the link group, not the entire network. Thus, when a link outage occurs it does not cause a topology change throughout the network and degrade service.
Finally, access policies are should be defined on the routers over the EWAN. A convenient way to define policy is with ACLs that apply to an IP subnet. Thus, a group of servers with similar access policies can be conveniently grouped together in the same IP subnet and the same VLAN. Other important services, such as DHCP are also defined on an IP-subnet basis.
Comparing Access Switches and Routers
An important question regarding Layer 2 EWAN services is, "Is it better to attach to the service with a switch or a router?" Unfortunately, there is no one correct answer. Clearly there are come CAPEX advantages when one looks at the price tag of a switch versus the price tag of a router. Switches are almost always less expensive than routers, and for some networks, cost is the main or only issue. In that case a switch will be used, regardless of any networking issues. However, in most cases a router is the better choice and will save the enterprise money in the long run, because they offer the following advantages:
•
•
•
•
•
Flexible Policing and Traffic Shaping
"Policing" is the ability to look at packets, compare them to a traffic contract, and either pass them, drop them, or mark them as nonconforming. It is a common misconception that policing alone provides complete traffic engineering—that is, that if the flow of packets is restricted into a network cloud, congestion will not occur. Although this may be true for a grossly underbooked, inefficient networks, it is not true for most "real" networks. When you simply restrict traffic into a cloud, important aspects—including traffic patterns, application-specific QoS issues, and time-of-day usage—must be considered. Even under the most thorough traffic analysis, many nondeterministic traffic patterns can still occur—especially with an EMS—any of which can cause a network element, or port, to congest and drop critical traffic. Since an EMS is a broadcast domain, its QoS characteristics are very unpredictable and can easily congest upon egress (Figure 8).
Figure 8
Policed, but Congested, Network
Even though policing in its own right does not constitute robust traffic engineering, it still plays a vital role in maintaining a congestion-free network. Although many switches can police, they do not have the same policing capabilities as routers. Many switches can police on a per-port basis, on IEEE 802.1P priority, and some can police on an IEEE 802.1Q VLAN. However, most routers can also do this, as well as police on IP ToS, DSCP, TCP port, UDP port, and IP address. Thus, with routers the granularity of policing can be based on IP level priorities, applications such as voice over IP (VoIP), and internal web applications, or even end stations, such as file servers or storage devices. This breadth of service enables the enterprise network to get the best use out of their expensive and critical wide-area infrastructure.
As stated earlier, policing at the edge cannot solve every problem, especially when you are trying to get every bit out of expensive WAN links. Traffic shaping adds another dimension to congestion avoidance and control. "Traffic shaping" is the ability of a router, under congestion or under traffic contract violation, to buffer and smooth traffic to an acceptable rate until the congestion or violation has abated. This feature is common on routers but seldom found on switches. Even if a switch can shape, it has the same limitations, compared to routers, as policing does. Figure 9 shows a 50-Mbps contract over a 100-Mbps link. Traffic is first policed to conform to 50 Mbps, and then the excess bandwidth is throttled, or shaped, so it does not have to be dropped. Most switches would simply drop the extra traffic at the policer.
Figure 9
Policing and Shaping
Address Structuring and Traffic Segmentation
With IP, each end station and router has a configurable address. Although some switches and network adapters allow you to customize the MAC address, this has only a few uses, because these addresses cannot be structured or summarized like IP addresses. Summarization allows large multiples of IP addresses, in a structured system, to be stored in memory as a single-entry summary, rather than individually. This reduces memory sizes, reduces address lookup times, aids in debugging, and reduces failure and recovery time because the device needs only to relearn a summary or a group of summaries rather than a complete list of addresses. Switches do not enjoy this luxury, and since many switches cannot learn addresses at line rate, after network failure, traffic is flooded while a switch tries to relearn its forwarding table. This only exacerbates the congestion and QoS problems associated with the failure.
Summaries can also be hierarchical. For example, a common scheme is to hierarchically vary summarizations as they relate to the access nodes, distribution layer, and core of a network. However, VLANs, like MAC addresses, cannot be hierarchically summarized. Even a VLAN tag-stacking scheme is a one-to-one mapping of customer to tag and is only used by an service provider. Thus, Ethernet cannot duplicating an IP or MPLS hierarchy. By using hierarchies, you can segment broadcast domains. This means that broadcast storms, intentional and unintentional, can be contained to small communities of interest that, under strain, do not affect the rest of the network.
Many protocols also help make Layer 3 multicasting even more efficient. Host-to-router protocols, such as IGMP, and router-to-router protocols such as PIM allow routers to create minimal tree multicast structures, ensuring that the multicast packets traverse only those links destined for valid destinations, rather than being broadcast (Figure 10). If this were done over a Layer 2 network, the multicast would be flooded throughout the Layer 2 domain (Figure 11). Furthermore, even when a switch performs IGMP and PIM snooping, there are issues regarding failure-recovery behavior and, more importantly, QoS, which limits these features' potential and predictability in large networks.)
Figure 10
Desired Effect When Multicasting
Figure 11
Adverse Effects of Multicasting over a Layer 2 Ethernet Switched Network
Fault Isolation and Traffic Control
Another important quality of traffic segmentation is fault isolation. Since traffic can be highly segmented, when issues arise they are constrained to smaller areas, allowing them to be located, and thus fixed, more quickly—lessening the mean time to repair (MTTR). Also aiding MTTR is the structured nature of IP and the vast array of tools that take advantage of this structure. Two common tools often taken for granted are ping and traceroute. These simple tools allow you to determine if a host or router is reachable at the network level, pointing to potential application-layer issues. Traceroute also lists the path that a packet takes as it traverses the network layer, pinpointing the beginning of the failure. In addition to these basic yet effective commands are the vast array of proprietary or management-based software tools available on management systems and protocol analyzers. In contrast, even simple tools like ping and traceroute have no counterparts in the switching world. In addition, as packets traverse from one Layer 2 boundary to another, they pass through routers, and at that point any Layer 2 packet trace loses its end-to-end significance, rendering it useless. The availability of these Layer 3 tools makes it easier to debug a network-not to mention that it reduces the number of expensive ($50,000 and up) protocol analyzers you need to buy.
Finally, routing protocols offer greater flexibility and control over to the path a packet takes through the network and how the topology reacts to change (Figure 12). Spanning Tree Protocol, the most commonly used Layer 2 topology protocol, has several shortcomings when it comes to large-scale networking. It is slow to react, needs to block links (and in most cases many links) rendering them useless, has trunking support protocols (IEEE 802.17AD) that cannot share loads over multiple-nonparallel links, and it cannot forward based on policies. On the other hand, link-state-based routing protocols can quickly react to and repair large-scale networks and can forward and balance loads over any number of links based on policies such as source, destination, route priority, and congested transit network. All of this allows you to use your network efficiently. Another shortcoming of large-scale flat Spanning Tree Protocol networks is that, to keep the network properly utilized and maintaining proper QoS the service provider has to reengineer the network every time a new customer is added. In contrast, routing protocols self-adapt as new users come online.
Figure 12
Inadequacies of Spanning Tree Protocol Versus Routing Protocols
SP Value-Added, Service-Friendly Platform
Another big advantage gained when choosing a router as an access device is that the service provider's ability to deploy value-added services grows. Simply put, routers are feature rich. They are so because all the important features reside above Layer 2, allowing the enterprise to outsource high-touch services and save money. These services range from security to voice management to storage integration. This allows an enterprise to streamline its network needs by combining extranet resources and sending them over one service and one WAN port.
A common example of such a service is outsourced firewalling—an enterprise's frontline defense, which authenticates and controls outside access. IP Security (IPSec) VPNs offer additional security by authenticating sources and encrypting data before it passes over WAN links. IPSec is commonly used by financial institutions and government authorities to protect their data. In addition, a routed access network allows the service provider or enterprise to deploy intrusion-detection software to detect and locate hackers.
Voice-related services include Survivable Remote Site Telephony (SRST) and IP Centrex. SRST detects failures (unreachable destinations) in the network, and then takes IP telephony calls and reroutes them to the public telephone network, rather than dropping the calls because the destination is no longer reachable. For those who do not wish to manage their own IP telephony system but still want the cost savings associated with an integrated voice and data network, service providers offer IP Centrex, a remote service that offers call-management features including voice mail.
Another exciting area that can be integrated is IP storage. IP enhances traditional storage networks by allowing storage area network (SAN) traffic to be multiplexed along with voice and data traffic. In addition, IP storage can apply IP structuring and VLAN concepts creating virtual storage communities that allow an enterprise to better utilize its facilities. These virtual storage communities can either be managed by the enterprise or outsourced to the service provider.
Routers Are a Futureproof Solution
Routers are the right technology for the future, not just today. Whereas switches cannot act as routers, most access enterprise routers can also act as full Layer 2 switches. More importantly, these routers are based on IP and they run IP protocols. The primary purpose of an IP routing protocol is to scale—IP routing protocols were developed because networks became too large to run without them.
In addition, for service providers to continue to operate, they must continue to develop value-added features that provide real benefits to their customers. This requires routers. In addition, many routers now are application aware—and even when they are not, application developers bind their applications to TCP and UDP ports and make it possible for their applications to write to ToS and DSCP priority fields. These hooks are not readily available to switches, and without them a service provider's ability to provide today's or future value-added services becomes difficult and expensive. Thus, today's routers are "futureproof" in that they will continue to scale and provide new and beneficial services to meet tomorrow's needs.
When Is It Acceptable to Use an Access Switch?
Given that switches typically cost less than a router, using a switch as an access node merits consideration, even though there are very inexpensive routers on the market. There are two cases in which it is acceptable to deploy a switch as an access node. The first occurs over a point-to-point link using EPL. In this case, the EWAN link appears to be another segment in a LAN, and EPL is secure in that it does not terminate on any extranet services (Figure 13). The second case occurs when dark fiber when is used, for the same reasons as with EPL. If you use an access switch with EPL, or EWS a hub-and-spoke topology should be implemented; the spoke nodes are switches that hub back to a router. Furthermore, hub-to-spoke traffic is switched, while spoke-to-spoke traffic is routed, for security and traffic segmentation reasons. This network design is commonly found in school districts and remote medical clinics or offices. You should never use a switch with an EMS because QoS, security, and traffic patterns are unpredictable.
Figure 13
Using Switches over EPL, Mapping Each Circuit to a Unique VLAN
Appendix
Configurations
This section has two main subsections: "Network Configurations" and "Security Configurations." The network section covers the configurations for the different network solutions. The security section discusses general network element and Layer 2 security configurations that should be in place when attaching to an EWAN. These configurations include general QoS and OSPF configurations. For an exhaustive list of configurations, consult the QoS and OSPF (Open Shortest Path First) design guides or the appropriate manual.
Network Configurations
This section first describes the solution parameters and then moves on to configure details. The following six solutions are covered:
•
•
•
•
•
•
Attaching to an EMS with Switches
Figure 14 shows a network of three switches connected to an EMS. There are two service types: a high-priority gold service, and a low-priority best effort service. The switches will be configured to see the network as one common Layer 2 broadcast domain/VLAN.
Figure 14
Switches Connecting to an EMS
Keys to this implementation are as follows:
1.
2.
Comments:
•
•
•
switch1> enableswitch1# config tCreate VLAN 1, the WAN VLAN, and give it an IP address.
switch1(config)#int vlan 1switch1(config-if)#ip address 10.1.1.1 255.255.0.0Make Interface Gigabit Ethernet 0/11 an IEEE 802.1Q trunk.
switch1(config-if)#int gi0/11switch1(config-if)# switchport trunk encapsulation dot1qswitch1(config-if)# switchport mode trunkswitch1(config-if)# exitCreate VLAN 1, the LAN VLAN, and give it the same ID as the WAN VLAN ID. Or, if a different VLAN ID is configured, the two must be bridged together.
switch1(config)# int vlan 1Make Interface Gigabit Ethernet 0/1 an untagged access VLAN. Alternatively, it can be configured to be an IEEE 802.1Q trunked VLAN. This depends on the configuration of the attached LAN.
switch1(config-if)#int gi0/1switch1(config-if)#switchport access vlan 1Map CoS bits to DSCP bits.
switch1(config)# mls qos map cos-dscpCreate a class map for best effort traffic.
switch1(config)# class-map beMake all packets with a DSCP value of 0 be matched to the best effort class.
switch1(config-cmap)# match-all beswitch1(config-cmap)# match ip dscp 0switch1(config-cmap)# exitCreate a class map for gold service.
switch1(config)# class-map goldMake all packets with a DSCP value of 3 be matched to the gold service class.
switch1(config-cmap)# match-all goldswitch1(config-cmap)#match ip dscp af31 af32 af33switch1(config-cmap)# exitCreate a policy map for the WAN port, and police packets matched to class maps "be" and "gold".
switch1(config)# policy-map ewanpolicyswitch1(config-pmap)# class beswitch1(config-pmap-c)# trust dscpCreate a policer to police packets that match to class "be".
Police the flow at a sustained rate of 20 Mbps, with a maximum burst size of 500 kBps, and drop all nonconforming traffic.
switch1(config-pmap-c)# police 20000000 500000 exceed-action dropswitch1(config-pmap-c)# exitCreate a policer to police packets that match to class "gold".
Police the flow at a sustained rate of 10 Mbps, with a maximum burst size of 250 kBps, and downgrade all nonconforming traffic.
switch1(config-pmap)# class goldswitch1(config-pmap-c)# trust dscpswitch1(config-pmap-c)# police 10000000 250000 exceed-action policed-dsp-transmitswitch1(config-pmap-c)# exitswitch1(config-pmap)# exitMap the policy "ewanpolicy" to the outgoing traffic on the WAN port.
switch1(config)# interface gi0/11switch1(config-if)# service-policy output ewanpolicyswitch1(config-if)# exitTurn on IGMP snooping on the WAN interface.
switch1(config)#int gi0/11switch1(config-if)#ip igmp snoopingSave the configuration.
switch1(config)# write tAttaching to an EMS with Routers
Figure 15 shows a network of three routers connected to an EMS. There are two service types: a high-priority gold service and a low-priority best effort service. The switches will be configured to see the network as one common Layer 2 broadcast domain/VLAN. However, the traffic flows between the two access switches will be forced to go through the distribution router.
Figure 15
Routers Connecting to an EMS
Keys to this implementation are as follows:
1.
2.
3.
4.
Comments:
1.
2.
3.
4.
router4> enablerouter4# config tCreate VLAN 2, the WAN VLAN, and give it an IP address.
router4(config)#int vlan 2router4(config-if)#ip address 10.1.1.4 255.255.0.0Give the Interface Gigabit Ethernet 0/11 an untagged VLAN.
router4(config-if)#int gi0/11router4(config-if)# switchport access vlan 2router4(config-if)# exitCreate VLAN 101, the LAN VLAN, and give it a VLAN ID different from the WAN VLAN ID.
router4(config)# int vlan 101Make Interface Gigabit Ethernet 0/1 an IEEE 802.1Q trunked VLAN. Alternatively, it can be configured to be an untagged access VLAN. This depends on the configuration of the attached LAN.
router4(config-if)#int gi0/1router4(config-if)# switchport trunk encapsulation dot1qrouter4(config-if)# switchport mode trunkMap CoS bits to DSCP bits.
router4(config)# mls qos map cos-dscpCreate a class map for best effort traffic.
router4(config)# class-map beMake all packets with a DSCP value of 0 be matched to the "best effort" class.
router4(config-cmap)# match-all berouter4(config-cmap)# match ip dscp 0router4(config-cmap)# exitCreate a class map for "gold" service.
router4(config)# class-map goldMake all packets with a DSCP value of 3 be matched to the "gold" service class.
router4(config-cmap)# match-all goldrouter4(config-cmap)#match ip dscp af31 af32 af33router4(config-cmap)# exitCreate a policy map for the WAN port, and police packets matched to class maps "be" and "gold".
router4(config)# policy-map ewanpolicyrouter4(config-pmap)# class berouter4(config-pmap-c)# trust dscpCreate a policer to police packets that match to class "be".
Police the flow at a sustained rate of 20 Mbps, with a maximum burst size of 500 kBps, and drop all nonconforming traffic.
router4(config-pmap-c)# police 20000000 500000 exceed-action droprouter4(config-pmap-c)# exitCreate a policer to police packets that match to class "gold".
Police the flow at a sustained rate of 10 Mbps, with a maximum burst size of 250 kBps, and downgrade all nonconforming traffic.
router4(config-pmap)# class goldrouter4(config-pmap-c)# trust dscprouter4(config-pmap-c)# police 10000000 250000 exceed-action policed-dsp-transmitrouter4(config-pmap-c)# exitrouter4(config-pmap)# exitMap the policy "ewanpolicy" to the outgoing traffic on the WAN port.
router4(config)# interface gi0/11router4(config-if)# service-policy output ewanpolicyrouter4(config-if)# exitCreate an OSPF instance, and give it an arbitrary process ID.
router4(config)# router ospf 10Place the OSPF instance in Area 0, unless a backbone area 0 already exists somewhere else in the network. In that case, use some other, unused area.
router4(config-router)# network 10.1.0.0 0.0.255.255 area 0Create the static route from router 4 (10.1.1.4) to router 5 (10.1.2.5) via router 6(10.1.31.6). If desired, a default route may be used instead.
router4(config)# ip route 10.1.2.0 10.1.3.6Enable IP routing.
Router4(config)# ip routingEnable multicast routing, PIM.
router4(config)# ip multicast-routingrouter4(config)int gi0/11router4(config-if)#ip pim sparse-denserouter4(config-if)# exitSave the configuration.
router4(config)# write tSimple ERS with Routers
Figure 16 shows a network of three routers connected to an ERS. Two point-to-point circuits run from the access routers to the distribution router. There are three service types: a voice service, a high-priority gold service, and a low-priority best effort service. The routers will be configured to see the network as a series of circuits, each with its own VLAN and IP subnet.
Figure 16
Routers Connecting to an ERS
Keys to this implementation are the following:
1.
2.
3.
4.
Comments:
•
•
•
Router7> enablerouter7# config tCreate VLAN 7, the WAN VLAN on router 7. Give it an IP address, and make it an 802.1Q trunked port.
router7(config)#int vlan 7router7(config-if)#ip address 10.1.2.7 255.255.255.0router7(config)#int gi 0/11router7(config-if)# switchport trunk encapsulation dot1qCreate VLAN 8, the WAN VLAN on router 8. Give it an IP address, and make it an 802.1Q trunked port.
router8(config)#int vlan 8router8(config-if)#ip address 10.1.3.8 255.255.255.0router8(config)#int gi 0/11router8(config-if)# switchport trunk encapsulation dot1qCreate VLANs 7 and 8, the WAN VLAN on router 9, and give them IP addresses that map the proper subnet to the proper VLAN.
router9(config)#int vlan 7router9(config-if)#ip address 10.1.2.9 255.255.255.0router9(config-if)#ip exitrouter9(config)#int vlan 8router9(config-if)#ip address 10.1.3.9 255.255.255.0router9(config-if)#ip exitrouter9(config)#int gi 0/11router9(config-if)# switchport trunk encapsulation dot1qrouter9(config)#siwtchport trunk allowed vlan 7 8router7(config-if)# exitCreate VLAN 102, the LAN VLAN, and give it a VLAN ID different from the WAN VLAN IDs.
router7(config)# int vlan 102Make Interface Gigabit Ethernet 0/1 an IEEE 802.1Q trunked VLAN. Alternatively, it can be configured to be an untagged access VLAN. This depends on the configuration of the attached LAN.
router7(config-if)#int gi0/1router7(config-if)# switchport trunk encapsulation dot1qrouter7(config-if)# switchport mode trunkMap CoS bits to DSCP bits.
router7(config)# mls qos map cos-dscpCreate a class map for best effort traffic.
router7(config)# class-map beMake all packets with a DSCP value of 0 be matched to the "best effort" class.
router7(config-cmap)# match-all berouter7(config-cmap)# match ip dscp 0router7(config-cmap)# exitCreate a class map for "gold" service.
router7(config)# class-map goldMake all packets with a DSCP value of 3 be matched to the "voice" services class.
router7(config-cmap)# match-all goldrouter7(config-cmap)#match ip dscp af31 af32 af33router7(config-cmap)# exitrouter7(config)# class-map voiceMake all packets with a DSCP value of "ef" be matched to the "voice" services class ("ef" is the DSCP value for IP telephony; you could also match on RTP values).
router7(config-cmap)# match-all voicerouter7(config-cmap)#match ip dscp efrouter7(config-cmap)# exitCreate a policy map for the WAN port, and police packets matched to class maps "be" and "gold".
router7(config)# policy-map ewanpolicyrouter7(config-pmap)# class berouter7(config-pmap-c)# trust dscpCreate a policer to police packets that match to class "be".
Police the flow at a sustained rate of 20 Mbps, with a maximum burst size of 500 kBps, and drop all nonconforming traffic.
router7(config-pmap-c)# police 20000000 500000 exceed-action droprouter7(config-pmap-c)# exitCreate a policer to police packets that match to class "gold".
Police the flow at a sustained rate of 10 Mbps, with a maximum burst size of 250 kBps, and downgrade all nonconforming traffic.
router7(config-pmap)# class goldrouter7(config-pmap-c)# trust dscprouter7(config-pmap-c)# police 10000000 250000 exceed-action policed-dsp-transmitrouter7(config-pmap-c)# exitCreate a policer to police packets that match to class "voice".
Police the flow at a sustained rate of 5 Mbps, with a maximum burst size of 50 kBps, and drop all nonconforming traffic.
router7(config-pmap)# class voicerouter7(config-pmap-c)# trust dscprouter7(config-pmap-c)# police 5000000 50000 exceed-action droprouter7(config-pmap-c)# exitrouter7(config-pmap)# exitMap the policy "ewanpolicy" to the outgoing traffic on the WAN port.
router7(config)# interface gi0/11router7(config-if)# service-policy output ewanpolicyrouter7(config-if)# exitCreate an OSPF instance, and give it an arbitrary process ID.
router7(config)# router ospf 10Place the OSPF instance in area 0, unless a backbone area 0 already exists somewhere else in the network. In that case, use some other, unused area.
router7(config-router)# network 10.1.0.0 0.0.255.255 area 0Enable IP routing.
Router7(config)# ip routingEnable multicast routing, PIM.
router7(config)# ip multicast-routingrouter7(config)int gi0/11router7(config-if)#ip pim sparse-denserouter7(config-if)# exitSave the configuration.
router7(config)# write tLarge-Scale ERS with Routers
Figure 17 shows a hierarchical network of four routers connected to an ERS. In the case of a large network that needs to be configured hierarchically, there would probably be two or more core routers, six or more distribution routers, and many access routers; this example is a simplified version. Two point-to-point circuits run from the access routers to the distribution router, and one runs from the distribution router to the core router. There are three service types: a voice service, a high-priority gold service, and a low-priority best effort service. The routers will be configured to see the network as a series of circuits, each with its own VLAN and IP subnet.
Figure 17
Hierarchically Routed Network Connecting to an ERS
Keys to this implementation are as follows:
1.
2.
3.
4.
Comments:
•
•
•
Router10> enablerouter10# config tCreate VLAN 9, the WAN VLAN on router 10. Give it an IP address and make it an 802.1Q trunked port.
router10(config)#int vlan 9router10(config-if)#ip address 10.1.2.10 255.255.255.0router10(config)#int gi 0/11router10(config-if)# switchport trunk encapsulation dot1qCreate VLAN 10, the WAN VLAN on router 11. Give it an IP address, and make it an 802.1Q trunked port.
router11(config)#int vlan 10router11(config-if)#ip address 10.1.3.11 255.255.255.0router11(config)#int gi 0/11router11(config-if)# switchport trunk encapsulation dot1qCreate VLAN 11, the WAN VLAN on router 13. Give it an IP address, and make it an 802.1Q trunked port.
router13(config)#int vlan 11router13(config-if)#ip address 10.1.1.13 255.255.255.0router13(config)#int gi 0/11router13(config-if)# switchport trunk encapsulation dot1qCreate VLANs 9, 10, and 11, the WAN VLAN on router 12. Give them IP addresses that map the proper subnet to the proper VLAN.
router12(config)#int vlan 9router12(config-if)#ip address 10.1.1.13 255.255.255.0router12(config-if)#ip exitrouter12(config)#int vlan 10router12(config-if)#ip address 10.1.1.13 255.255.255.0router12(config-if)#ip exitrouter12(config)#int vlan 11router12(config-if)#ip address 10.1.1.13 255.255.255.0router9(config)#int gi 0/11router9(config-if)# switchport trunk encapsulation dot1qrouter9(config)#siwtchport allowed vlan 9 10 11router10(config-if)# exitCreate VLAN 103, the LAN VLAN, and give it a VLAN ID different from the WAN VLAN IDs.
router10(config)# int vlan 103Make Interface Gigabit Ethernet 0/1 an IEEE 802.1Q trunked VLAN. Alternatively, it can be configured to be an untagged access VLAN. This depends on the configuration of the attached LAN.
router10(config-if)#int gi0/1router10(config-if)# switchport trunk encapsulation dot1qrouter10(config-if)# switchport mode trunkMap CoS bits to DSCP bits.
router10(config)# mls qos map cos-dscpCreate a class map for "best effort" traffic.
router10(config)# class-map beMake all packets with a DSCP value of 0 be matched to the "best effort" class.
router10(config-cmap)# match-all berouter10(config-cmap)# match ip dscp 0router10(config-cmap)# exitCreate a class map for "gold" service.
router10(config)# class-map goldMake all packets with a DSCP value of 3 be matched to the voice service class.
router10(config-cmap)# match-all goldrouter10(config-cmap)#match ip dscp af31 af32 af33router10(config-cmap)# exitrouter10(config)# class-map voiceMake all packets with a DSCP value of "ef" be matched to the voice services class ("ef" is the DSCP value for IP telephony; you could also match on RTP values.).
router10(config-cmap)# match-all voicerouter10(config-cmap)#match ip dscp efrouter10(config-cmap)# exitCreate a policy map for the WAN port, and police packets matched to class maps "be" and "gold".
router10(config)# policy-map ewanpolicyrouter10(config-pmap)# class berouter10(config-pmap-c)# trust dscpCreate a policer to police packets that match to class "be".
Police the flow at a sustained rate of 20 Mbps, with a maximum burst size of 500 kBps, and drop all nonconforming traffic.
router10(config-pmap-c)# police 20000000 500000 exceed-action droprouter10(config-pmap-c)# exitCreate a policer to police packets that match to class "gold".
Police the flow at a sustained rate of 10 Mbps, with a maximum burst size of 250 kBps, and downgrade all nonconforming traffic.
router10(config-pmap)# class goldrouter10(config-pmap-c)# trust dscprouter10(config-pmap-c)# police 10000000 250000 exceed-action policed-dsp-transmitrouter10(config-pmap-c)# exitCreate a policer to police packets that match to class "voice".
Police the flow at a sustained rate of 5 Mbps, with a maximum burst size of 50 kBps, and drop all nonconforming traffic.
router10(config-pmap)# class voicerouter10(config-pmap-c)# trust dscprouter10(config-pmap-c)# police 5000000 50000 exceed-action droprouter10(config-pmap-c)# exitrouter10(config-pmap)# exitMap the policy "ewanpolicy" to the outgoing traffic on the WAN port.
router10(config)# interface gi0/11router10(config-if)# service-policy output ewanpolicyrouter10(config-if)# exitCreate an OSPF instance, and give it an arbitrary process ID.
router10(config)# router ospf 10Place the OSPF instance in area 1.
router10(config-router)# network 10.1.0.0 0.0.255.255 area 1Create an OSPF instance, and give it an arbitrary process ID.
router11(config)# router ospf 10router11(config-router)# network 10.1.0.0 0.0.255.255 area 1router13(config)# router ospf 10router13(config-router)# network 10.1.0.0 0.0.255.255 area 1Create an OSPF instance, and give it an arbitrary process ID.
router12(config)# router ospf 1Place the OSPF instances in areas 1 and 0, unless a backbone area 0 already exists somewhere else in the network. In that case, make sure that area 0 in this example is connected to the real-world area 0. If you will not be connecting example area 0 to real-world Area 0, then put all the routers in this example into area 1.
router12(config-router)# network 10.1.0.0 0.0.255.255 area 1router12(config-router)#exitrouter12(config)# router ospf 10router12(config-router)# network 10.1.0.0 0.0.255.255 area 0Enable IP routing.
Router10(config)# ip routingEnable multicast routing, PIM.
router10(config)# ip multicast-routingrouter10(config)int gi0/11router10(config-if)#ip pim sparse-denserouter10(config-if)# exitSave the configuration.
router10(config)# write tHybrid EMS- ERS Configuration
Figure 18 shows a network of two switches connected to a router via an EMS. The router has an additional connection via an ERS to the Internet. The router and switches see the EMS as a broadcast domain. The router sees the ERS as a point-to-point circuit with a different VLAN and IP subnet. The switches do not see the ERS and will send traffic to the router to get to the Internet. The router attaches to both the ERS and EMS through one physical port, although if you prefer, separate ports can be used. There are two service types: a high-priority gold service and a low-priority best effort service.
Figure 18
Combined EMS-ERS Solution
Keys to this implementation are as follows:
1.
2.
3.
4.
5.
Comments:
•
•
•
•
b> enableswitch13# config tCreate VLAN 1, the WAN VLAN, and give it an IP address.
switch13(config)#int vlan 12switch13(config-if)#ip address 10.1.2.13 255.255.255.0Make Interface Gigabit Ethernet 0/11 an IEEE 802.1Q trunk.
switch13(config-if)#int gi0/11switch13(config-if)# switchport trunk encapsulation dot1qswitch13(config-if)# switchport mode trunkswitch13(config-if)# exitCreate VLAN 104, the LAN VLAN, and give it a VLAN ID different from the WAN VLAN ID.
switch13(config)# int vlan 104Make Interface Gigabit Ethernet 0/1 an untagged access VLAN. Alternatively, it can be configured to be an IEEE 802.1Q trunked VLAN. This depends on the configuration of the attached LAN.
switch13(config-if)#int gi0/1switch13(config-if)#switchport access vlan 104Map CoS bits to DSCP bits.
switch13(config)# mls qos map cos-dscpCreate a class map for best effort traffic.
switch13(config)# class-map beMake all packets with a DSCP value of 0 be matched to the "best effort" class.
switch13(config-cmap)# match-all beswitch13(config-cmap)# match ip dscp 0switch13(config-cmap)# exitCreate a class map for gold service.
switch13(config)# class-map goldMake all packets with a DSCP value of 3 be matched to the "gold" service class.
switch13(config-cmap)# match-all goldswitch13(config-cmap)#match ip dscp af31 af32 af33switch13(config-cmap)# exitCreate a policy map for the WAN port, and police packets matched to class maps "be" and "gold".
switch13(config)# policy-map ewanpolicyswitch13(config-pmap)# class beswitch13(config-pmap-c)# trust dscpCreate a policer to police packets that match to class "be".
Police the flow at a sustained rate of 20 Mbps, with a maximum burst size of 500 kBps, and drop all nonconforming traffic.
switch13(config-pmap-c)# police 20000000 500000 exceed-action dropswitch13(config-pmap-c)# exitCreate a policer to police packets that match to class "gold".
Police the flow at a sustained rate of 10 Mbps, with a maximum burst size of 250 kBps, and downgrade all nonconforming traffic.
switch13(config-pmap)# class goldswitch13(config-pmap-c)# trust dscpswitch13(config-pmap-c)# police 10000000 250000 exceed-action policed-dsp-transmitswitch13(config-pmap-c)# exitswitch13(config-pmap)# exitMap the policy "ewanpolicy" to the outgoing traffic on the WAN port.
switch13(config)# interface gi0/11switch13(config-if)# service-policy output ewanpolicyswitch13(config-if)# exitTurn on IGMP snooping on the WAN interface.
switch13(config)#int gi0/11switch13(config-if)#ip igmp snoopingSave the configuration.switch13(config)# write t\*******************\router15> enablerouter15# config tCreate VLANs 12 and 13 , the WAN VLANs on router 15, and give them the appropriate IP addresses. VLAN 12 attaches to the EMS, and VLAN 13 attaches to the ERS. Also create an 802.1Q tagged trunk.
router15(config)#int vlan 12router15(config-if)#ip address 10.1.1.15 255.255.255.0router15(config-if)#ip exitrouter15(config)#int vlan 13router15(config-if)#ip address 10.1.2.15 255.255.255.0router15(config-if)#ip exitrouter15(config)#int gi 0/11router15(config-if)# switchport trunk encapsulation dot1qrouter15(config)#siwtchport trunk allowed vlan 12 13router15(config-if)# exitCreate VLAN 104, the LAN VLAN, and give it a VLAN ID different from the WAN VLAN ID
router15(config)# int vlan 104Make Interface Gigabit Ethernet 0/1 an IEEE 802.1Q trunked VLAN. Alternatively, it can be configured to be an untagged access VLAN. This depends on the configuration of the attached LAN.
router15(config-if)#int gi0/1router15(config-if)# switchport trunk encapsulation dot1qrouter15(config-if)# switchport mode trunkMap CoS bits to DSCP bits.
router15(config)# mls qos map cos-dscpCreate a class map for
