Section 11.2. Traffic Engineering: An Overview


11.2. Traffic Engineering: An Overview

Key to the usefulness of MPLS is that after the LSP is established, your EGP or IGP or both can view the LSP as a traffic path when calculating best paths to a destination. This is again similar to an ATM or Frame Relay VC: Even though the path actually traverses multiple switching nodes, a routing protocol can view it as a single link between the ingress node and the egress node.

The example LSP you have seen so far in Figures 11.2 and 11.3 is too simple to be of practical interest. There is only one physical route from the ingress LSR to the egress LSR, so routing protocols are going to choose it whether an LSP exists or not. But consider the ingress and egress nodes in Figure 11.6. This network contains multiple paths between the ingress and the egress. The IGP in this network will do just what it is designed to do and select the shortest path between ingress and egress based on its given metrics. Assuming that all routers in the network are ingress for some traffic flows and egress for other traffic flows, the traffic might be fairly evenly distributed throughout the network. But suppose much more traffic flows into the one ingress point shown and out of the egress point shown, than anywhere else in the network. The single path chosen by the IGP might become congested while available bandwidth on other paths is underutilized.

Figure 11.6. The IGP will select only one of the multiple paths between the ingress and egress router in this topology.


Consider also a case in which the network in Figure 11.6 is a multiservice network. You might want to route best-effort traffic between the ingress and egress over longer paths, reserving the bandwidth on the shortest path for delay-sensitive traffic such as voice.

Such requirements are the basis of MPLS traffic engineering. Using MPLS LSPs, you can engineer your traffic loads across the network in such a way that available bandwidth is more efficiently utilized; you can establish different paths for different traffic classes for better multiservice performance; and you can route traffic around trouble spots such as congested links and nodescongestion that IGPs cannot detect.

Traffic engineering capability has long been a part of ATM and Frame Relay networks. However, before MPLS TE, the only traffic engineering that could be done in IP networks without an ATM or Frame Relay overlay was a crude manipulation of link metrics. Changing a link metric is an all-or-nothing action; the IGP still chooses the shortest path. MPLS TE enables you to track a number of interface parameters throughout your network and then use these parameters to specify how a path is selected and what packets use what path, permitting much more granularity in regulating traffic flows.

11.2.1. TE Link Parameters

As you certainly know, an IGP selects a shortest path based on a metricsome numeric valueassigned to the router interfaces throughout the network. The foundation of traffic engineering is also an assignment of values to interfaces. But because a useful and flexible TE application requires a variety of parameters on which you can base path selection, there must be a variety of values that can be assigned to interfaces that reflect these parameters. Those parameters are:

  • Maximum Bandwidth

  • Maximum Reservable Bandwidth

  • Unreserved Bandwidth

  • Traffic Engineering Metric

  • Administrative Group

The first three parameters enable a mechanism by which you can specify the bandwidth an LSP can use. For example, an LSP might be required to have 10M available to it. When the LSP is being set up, it can traverse only links between the ingress and egress on which at least 10M of bandwidth is available. That 10M is then reserved on the links and becomes unavailable for use by another LSP. So if a 10M LSP is set up across a link on which there is a total of 15M of available bandwidth, only 5M of reservable bandwidth is left after the LSP is established. If a second LSP is to be set up and also requires 10M, it cannot use this link and must be set up on an alternate path that provides enough bandwidth. If no other path with sufficient reservable bandwidth is available, the second LSP cannot be established.

Maximum Bandwidth is the bandwidth of the interface. It might be the actual bandwidth of the interface or it might be a configured number.

Maximum Reservable Bandwidth specifies how much of the link bandwidth can be reserved by LSPs.

Unreserved Bandwidth is the amount of maximum reservable bandwidth that has not yet been used by LSPs.

Traffic Engineering Metric is a 24-bit value that can be assigned to an interface and is used the same as an IGP metric. The TE metric allows you to set up a metric-based LSP topology that is different from the metric-based IGP shortest-path topology.

Administrative Group, also known as affinity, enables you to make an interface a member of one or more of 32 possible administrative groups. Administrative groups are often called link colors because you can associate names with each of the 32 administrative groups, and traditionally those names have been the names of colors. For example, you might "color" all of your highest-speed links gold, your medium-speed links silver, and your low-speed links bronze. You could then specify that certain LSPs can only use gold or silver links and other LSPs can only use silver or bronze links. Or instead of specifying what links an LSP can use, you might specify what links an LSP cannot use: For instance, an LSP might use any link except platinum links.

Figure 11.7 shows a Cisco Systems IOS output displaying traffic engineering parameters for an interface. You can observe the TE metric, the maximum bandwidth, the maximum reservable bandwidth, and the administrative groups (affinity bits) to which the interface belongs. Notice that the maximum reservable bandwidth is greater than the maximum bandwidth. Specifying a maximum reservable bandwidth greater than the maximum bandwidth permits oversubscription of the interface.

Figure 11.7. An IOS output showing the TE parameters associated with an interface.

[View full width]

Cisco7# show ip ospf mpls traffic-eng link OSPF Router with ID (10.1.1.1) (Process ID 1) Area 0 has 1 MPLS TE links. Area instance is 14. Links in hash bucket 8. Link is associated with fragment 1. Link instance is 14 Link connected to Point-to-Point network Link ID :192.168.5.4 Interface Address :10.5.0.1 Neighbor Address :10.5.0.2 Admin Metric :84 Maximum bandwidth :150000 Maximum reservable bandwidth :250000 Number of Priority :8 Priority 0 :250000 Priority 1 :250000 Priority 2 :250000 Priority 3 :250000 Priority 4 :250000 Priority 5 :250000 Priority 6 :250000 Priority 7 :212500 Affinity Bit :0x3


Of interest in Figure 11.7 are the eight entries labeled "Priority" 0 through 7. The value associated with each of these eight priorities is the unreserved bandwidth. When an LSP is being configured for TE, is can be given a setup and a hold priority, and each of these priorities is a value between 0 and 7. Setup priority is the "strength" the LSP has to preempt another LSP, and the hold priority is the "strength" an LSP has to resist being preempted. If a new LSP has a setup priority higher than the hold priority of an existing LSP, and there are not enough link resources such as bandwidth to support both, the stronger LSP can replace the weaker LSP, and the weaker LSP must find a new path to its egress. So the unreserved bandwidth in Figure 11.7 is allocated separately for each of the eight setup priority levels; 0 is the highest or "strongest," and 7 is the lowest.

11.2.2. Constrained Shortest Path First

The calculation of a traffic engineered path takes place only in the ingress router. That means the ingress router must have some way to learn all of the TE parameters assigned to all MPLS interfaces in the network, and it must have a place to store that information. This is where OSPF and IS-IS come in: Both protocols have extensions that allow them to carry the TE interface parameters along with the normal interface parameters such as OSPF or IS-IS metrics and link state. Those extensions, the real topic of this chapter, are detailed in Sections 11.3 and 11.4.

Just as OSPF LSAs and IS-IS LSPs are stored in a link state database, the traffic engineering parameters carried by the extensions to these protocols are stored in a special database called the traffic engineering database (TED). Figure 11.8 shows an example of a TED from a Juniper Networks LSR. For each entry, you can observe the administrative groups (called color in this display), the metric, and the bandwidth parameters.

Figure 11.8. A JUNOS output showing a traffic engineering database.

[View full width]

jeff@Juniper3> show ted database extensive TED database: 0 ISIS nodes 6 INET nodes NodeID: 172.16.229.7 Type: Rtr, Age: 72166 secs, LinkIn: 1, LinkOut: 1 Protocol: OSPF(0.0.0.0) To: 172.16.229.190-1, Local: 172.16.229.191, Remote: 0.0.0.0 Color: 0 <none> Metric: 100 Static BW: 1000Mbps Reservable BW: 1000Mbps Available BW [priority] bps: [0] 1000Mbps [1] 1000Mbps [2] 1000Mbps [3] 1000Mbps [4] 1000Mbps [5] 1000Mbps [6] 1000Mbps [7] 1000Mbps Interface Switching Capability Descriptor(1): Switching type: Packet Encoding type: Packet Maximum LSP BW [priority] bps: [0] 1000Mbps [1] 1000Mbps [2] 1000Mbps [3] 1000Mbps [4] 1000Mbps [5] 1000Mbps [6] 1000Mbps [7] 1000Mbps NodeID: 172.16.229.8 Type: Rtr, Age: 72161 secs, LinkIn: 1, LinkOut: 1 Protocol: OSPF(0.0.0.0) To: 172.16.229.189-1, Local: 172.16.229.188, Remote: 0.0.0.0 Color: 0 <none> Metric: 100 Static BW: 1000Mbps Reservable BW: 1000Mbps Available BW [priority] bps: [0] 1000Mbps [1] 1000Mbps [2] 1000Mbps [3] 1000Mbps [4] 1000Mbps [5] 1000Mbps [6] 1000Mbps [7] 1000Mbps Interface Switching Capability Descriptor(1): Switching type: Packet Encoding type: Packet Maximum LSP BW [priority] bps: [0] 1000Mbps [1] 1000Mbps [2] 1000Mbps [3] 1000Mbps [4] 1000Mbps [5] 1000Mbps [6] 1000Mbps [7] 1000Mbps NodeID: 172.16.229.9 Type: Rtr, Age: 10924 secs, LinkIn: 3, LinkOut: 3 Protocol: OSPF(0.0.0.0) To: 172.16.229.190-1, Local: 172.16.229.190, Remote: 0.0.0.0 Color: 0 <none> Metric: 100 Static BW: 1000Mbps Reservable BW: 1000Mbps Available BW [priority] bps: [0] 1000Mbps [1] 1000Mbps [2] 1000Mbps [3] 1000Mbps [4] 1000Mbps [5] 1000Mbps [6] 1000Mbps [7] 1000Mbps Interface Switching Capability Descriptor(1): Switching type: Packet Encoding type: Packet Maximum LSP BW [priority] bps: [0] 1000Mbps [1] 1000Mbps [2] 1000Mbps [3] 1000Mbps [4] 1000Mbps [5] 1000Mbps [6] 1000Mbps [7] 1000Mbps To: 172.16.229.10, Local: 172.16.229.193, Remote: 172.16.229.192 Color: 0 <none> Metric: 100 Static BW: 155.52Mbps Reservable BW: 155.52Mbps Available BW [priority] bps: [0] 155.52Mbps [1] 155.52Mbps [2] 155.52Mbps [3] 155.52Mbps [4] 155.52Mbps [5] 155.52Mbps [6] 155.52Mbps [7] 155.52Mbps Interface Switching Capability Descriptor(1): Switching type: Packet Encoding type: Packet Maximum LSP BW [priority] bps: [0] 155.52Mbps [1] 155.52Mbps [2] 155.52Mbps [3] 155.52Mbps [4] 155.52Mbps [5] 155.52Mbps [6] 155.52Mbps [7] 155.52Mbps To: 172.16.229.10, Local: 172.16.229.195, Remote: 172.16.229.194 Color: 0 <none> Metric: 100 Static BW: 155.52Mbps Reservable BW: 155.52Mbps Available BW [priority] bps: [0] 155.52Mbps [1] 155.52Mbps [2] 155.52Mbps [3] 155.52Mbps [4] 155.52Mbps [5] 155.52Mbps [6] 155.52Mbps [7] 155.52Mbps Interface Switching Capability Descriptor(1): Switching type: Packet Encoding type: Packet Maximum LSP BW [priority] bps: [0] 155.52Mbps [1] 155.52Mbps [2] 155.52Mbps [3] 155.52Mbps [4] 155.52Mbps [5] 155.52Mbps [6] 155.52Mbps [7] 155.52Mbps


When you configure an LSP at an ingress LSR, you specify constraints on the LSP: What link colors it can or cannot use, its bandwidth, the maximum number of LSR hops it can traverse, and so on. Using these constraints and the information in the TED, the LSR runs a modified SPF algorithm called constrained shortest path first (CSPF), which calculates the shortest path to the egress within the constraints you specified. A specification of the resulting shortest path is then fed to the signaling protocolRSVP or CR-LDPwhich sets up the LSP.




OSPF and IS-IS(c) Choosing an IGP for Large-Scale Networks
OSPF and IS-IS: Choosing an IGP for Large-Scale Networks: Choosing an IGP for Large-Scale Networks
ISBN: 0321168798
EAN: 2147483647
Year: 2006
Pages: 111
Authors: Jeff Doyle

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net