cc/td/doc/product/cable/svc_ctrl/scappsbb
hometocprevnextglossaryfeedbacksearchhelp
PDF

Table Of Contents

Overview of the Service Control Solution for MPLS/VPN Networks

Service Control in the MPLS/VPN Environment

Definitions and Acronyms

What are the Challenges for Service Control for MPLS/VPN Support?

How MPLS/VPN Support Works

Flow Detection

Subscriber Detection

How the Service Control MPLS/VPN Solution Works

Service Control MPLS/VPN Concepts

Non-VPN Subscribers

Bypassing Unknown VPNs

Additional MPLS Pattern Support

VPN Identifier (RD or RT)

Service Control MPLS/VPN Requirements

Topology

Capacity

Limitations


Overview of the Service Control Solution for MPLS/VPN Networks


Service Control in the MPLS/VPN Environment 

Definitions and Acronyms 

What are the Challenges for Service Control for MPLS/VPN Support? 

How MPLS/VPN Support Works 

Service Control MPLS/VPN Concepts 

Service Control MPLS/VPN Requirements 

Service Control in the MPLS/VPN Environment

MPLS/VPN networks are very complex and contain many routing protocols and many different levels of addressing and control. In addition, the various VPNs may use overlapping IP addresses (private IPs).

The SCE platform makes a distinction between identical IP addresses that come from different VPNs, and maps them into subscribers according to the MPLS labels attached to the packets. This involves various mechanisms in all levels of the system.

The following assumptions and requirements allow the SCE platform to operate in an MPLS/VPN environment:

The MPLS/VPN architecture is according to RFC-2547.

The specific type of encapsulation used is the MPLS shim header over Ethernet (described in RFC-3032).

There are two levels of MPLS labels.

External labels — Used for transport over the service provider MPLS core network.

Internal labels (BGP labels) — Used to identify the VPNs connected to each edge router, and typically controlled by the BGP protocol.

All IP addresses in one VPN are treated as a single subscriber.

The MPLS/VPN solution contains the SCE platform and the SM. The SM acts as a BGP peer for the PE routers in the service provider network, and communicates the BGP information to the SCE platform as subscriber information.


Note The MPLS/VPN solution supports the existence of non-VPN subscribers concurrently with the MPLS/VPN subscribers (see Non-VPN Subscribers ).


Definitions and Acronyms

The following table defines important terms and acronyms.

Table 2-1 MPLS/VPN Terms and Acronyms

Term or Acronym

Definition

PE (Provider Edge router)

A router at the edge of the service provider network. The PE routers are the ones that connect to the customers, and maintain the VPNs

P (Provider router)

A router in the core of the service provider network. P routers only forward MPLS packets, regardless of VPNs.

VPN (Virtual Private Network)

In the Service Control context, a VPN is the part of the VPN that resides in a specific site. This is the subscriber of the solution

BGP LEG

A software module that resides on the SM server and generates BGP-related login events. The BGP LEG communicates with the BGP routers (PEs) and passes the relevant updates to the SM software, which generates login events to the SCE platform for the updated VPN subscribers.

Upstream

Traffic coming from the PE router and going into the P router

Downstream

Traffic coming from the P router and going into the PE router

RD (Route Distinguisher)

Used to uniquely identify the same network/mask from different VRFs (such as, 10.0.0.0/8 from VPN A and 10.0.0.0/8 from VPN B)

RT (Route Target)

Used by the routing protocols to control import and export policies, to build arbitrary VPN topologies for customers

VRF (Virtual Routing and Forwarding instance)

Mechanism used to build per-interface routing tables. Each PE has several VRFs, one for each site it connects to. This is how the private IPs remain unique.


What are the Challenges for Service Control for MPLS/VPN Support?

Private IP addresses cause flows to look the same except for their MPLS labels.

The MPLS labels are different in each direction, and must be matched.

An entire VPN must be accounted as one subscriber. The problem is how to detect that a flow belongs to a certain VPN.

In the downstream direction there is no external label. We must be able to understand the VPN information from the internal label + the MAC address of the PE.

How MPLS/VPN Support Works

Service Control supports two mechanisms that make MPLS/VPN support work:

Flow detection - This is the job of the SCE platform, to match upstream and downstream traffic to identify flows.

Subscriber detection - This is the job of the SM, to match downstream labels with the VPN to identify the subscriber entity.

Flow Detection 

Subscriber Detection 

How the Service Control MPLS/VPN Solution Works 

Flow Detection

Flow detection is the process of deciding which packets belong to the same flow. This relates to the first two challenges listed:

Private IP addresses cause flows to look the same except for their MPLS labels.

The MPLS labels are different in each direction, and must be matched.

Flow detection is based on the MPLS labels, extending the basic 5 tuple that SCOS uses to identify flows, and taking into account the fact that in MPLS, the packet is labeled differently in each direction.

Since MPLS traffic is unidirectional, each direction is classified separately by the SCE platform, using the following:

Downstream - the BGP label and the MAC address of the PE (only one label that is relevant to the classification)

Downstream labels are learned from the control plane (BGP).

Upstream - the combination of the external label, the BGP label, and the MAC address of the P router (two labels that are relevant to the classification)

Upstream labels are learned from the data plane.

Subscriber Detection

What is a VPN Subscriber? 

SM and Subscriber Detection 

What is a VPN Subscriber?

As in other modes of operation, in MPLS/VPN each flow belongs to a certain subscriber. A VPN subscriber is a customer of the Service Provider, who pays for the VPN service. All traffic of that VPN customer is aggregated into a single VPN subscriber for Service Control.

SM and Subscriber Detection

The network configuration that provides the division into VPN subscribers is controlled by the SM. The network-wide value that describes a VPN most closely is either the Route Target or the Route Distinguisher:

The administrator configures the SM to detect VPN subscribers, according to selected attribute (RT or RD) (see How to Configure the SM for MPLS/VPN Support.)

The network operator provides the SCE platform with a mapping between RT values and VPN subscriber names. (See How to Manage MPLS/VPN Support via SM CLU )

The relevant module in the Subscriber Manager server (SM) is the BGP-LEG. The BGP-LEG is added to the BGP neighborhood for obtaining the information on the MPLS labels. The local PEs are configured to add the BGP-LEG as a BGP peer.

BGP-LEG gets MP-BGP messages from the PEs with the allocated labels per VPN and forwards them to the SM module.

The SM updates each SCE platform with the mapping of MPLS labels to VPN subscribers.

How the Service Control MPLS/VPN Solution Works

How the Service Control MPLS/VPN Solution Works: A Summary 

SCE Platform Tasks in the MPLS/VPN Solution 

BGP LEG Tasks in the MPLS/VPN Solution 

SM Tasks in the MPLS/VPN Solution 

How the Service Control MPLS/VPN Solution Works: A Summary

The SM is configured with the VPNs that should be managed.

A VPN is identified by the RD / RT and the PE.

The BGP-LEG updates the SM with the MPLS labels.

The SM pushes the VPN subscriber to the SCE platform with the downstream MPLS labels of the VPN.

The SCE platform resolves the PE MAC addresses and updates its tables with the new information.

The SCE platform learns the upstream labels, including the P MAC address.

The SCE platform provides the regular services to the VPN subscriber (BW management, reports, etc.)

SCE Platform Tasks in the MPLS/VPN Solution

Matching upstream to downstream labels

Mappings of downstream labels to VPN subscribers are received from the SM

Upstream labels are learned from the data

The MAC addresses of the PEs are used to distinguish downstream labels of different PEs

After the learning period, each flow is classified as belonging to one of the VPN subscribers

The SCE platform runs the SCA-BB application for the network flows, which are classified to VPN subscribers, thus providing subscriber aware service control and reporting

BGP LEG Tasks in the MPLS/VPN Solution

The BGP LEG is a software module that runs on the SM server

The LEG maintains a BGP session with a list of PEs

After the sessions establishment, the LEG propagates MP-BGP route-updates from the PEs to the SM module

SM Tasks in the MPLS/VPN Solution

The VPNs are stored in the SM database as VPN subscribers.

A VPN subscriber is a group of VPN sites.

Each VPN site is defined by:

The IP address of the loopback interface of the PE router.

The RD or RT that identifies the VPN within the PE router.

The SM receives updates from the BGP LEG, and updates the VPN subscriber information with the new MPLS labels.

The relevant SCE platforms that will get the MPLS updates are defined by the VPN subscriber domain

Service Control MPLS/VPN Concepts

Non-VPN Subscribers 

Bypassing Unknown VPNs 

Additional MPLS Pattern Support 

VPN Identifier (RD or RT) 

Non-VPN Subscribers

The MPLS/VPN solution supports the existence of non-VPN (regular IP) subscribers concurrently with the MPLS/VPN subscribers, with the following limitations and requirements:

The SM must work in "push" mode.

Non-VPN subscribers cannot have MPLS/VPN mappings.

VLAN subscribers are NOT supported at the same time as MPLS/VPN subscribers.

In typical MPLS/VPN networks, traffic that does not belong to any VPN is labeled with a single MPLS label in the upstream direction, which is used for routing. The downstream direction of such flows typically contains no label, due to penultimate hop popping.

The SCE platform uses the one or more labels upstream and no label downstream definition to identify non-VPN flows. Classification and traffic processor load balancing on these flows is performed according to the IP header, rather than the label.

This process requires learning of the upstream labels in use for such flows, and is done using the flow detection mechanism described above (see Flow Detection ).

Bypassing Unknown VPNs

In an MPLS network, there may be many VPNs crossing the SCE platform, only a small number of which require service control functionality. It is necessary for the SCE platform to recognize which VPNs are not managed.

The SCE platform automatically bypasses any VPN that is not configured in the SM

The VPNs are bypassed by the SCE platform without any service

Note that the label limit of 57,344 different labels includes labels from the bypassed VPNs.

Each bypassed VPN entry, both upstream and downstream, is removed from the database after a set period of time (10 minutes). If the entry is still used in the traffic, it will be re-learnt. This allows the database to remain clean, even if the labels are reused by the routers for different VPNs.

In the show bypassed VPNs command, the age is indicated with each label - the length of time since it was learned.

Additional MPLS Pattern Support

The MPLS/VPN solution was designed to provide DPI services in MPLS/VPN network. These networks use BGP protocol as the control plane for the VPNs and LDP protocol for routing. There are complex networks where the MPLS infrastructure is used not only for VPN and routing, but also for other features such as traffic engineering (TE) and better fail-over. These features are usually enabled per VRF in the PE.

The Service Control MPLS/VPN solution does not support VPNs that use other MPLS-related features. Features such as MPLS-TE or MPLS-FRR (Fast Reroute) are not supported. VPNs for which these features are enabled can be automatically bypassed in the system, but are not allowed to be configured in the SM as serviced VPNs. Configuration of these VPNs in the SM might cause misclassification due to label aliasing.

The following list describes the labels combinations that are supported by the SCE platform and how each combination is interpreted by the platform:

One or more labels upstream, no labels downstream:

Assumed to be non-VPN (see Non-VPN Subscribers ).

The SCE platform treats the following IP flows as non-VPN flows, and ignores their labels.

One label upstream, one label downstream:

Assumed to be VPN traffic, in which the P router happens to be the last hop in the upstream.

The label in the downstream is treated as a BGP label, like the regular case. If the BGP label is known from the SM, then the flow is assigned to the correct subscriber, otherwise, it is treated as a bypassed VPN.

Two labels upstream, one label downstream:

This is the typical configuration of the system. Of the two upstream labels, one is for BGP and one for LDP. The downstream label is for BGP only

More than two labels upstream, or more than one label downstream:

These combinations occur when other MPLS-related features are enabled for the VPN. Such VPNs are not supported and should not be configured in the SM. However, they can be bypassed in the SCE platform without any service and without harming the service for other VPNs.

VPN Identifier (RD or RT)

Either the Route Distinguisher (RD) attribute or the Route Target (RT) attribute can be used to identify the VPN subscriber. It is required to decide which attribute best reflects the VPN subscriber partitioning, and configure the system accordingly. Note that the configuration is global for all the subscribers, that is, all subscribers must be identified by the same attribute.

The Route Distinguisher (RD) is generally used to distinguish the distinct VPN routes of separate customers who connect to the provider, so in most cases the RD is a good partition for the subscribers in the network. Since the RD is an identifier of the local VRF, and not the target VRF, it can be used to distinguish between VPN sites that transfer information to a common central entity (for example a central bank, IRS, Port Authority, etc.).

The Route Target (RT) is used to define the destination VPN site. Though it is not intuitive to define the VPN subscriber based on its destination route, it might be easier in some cases. For example, if all the VPN sites that communicate to a central bank should be treated as a single subscriber, consider using the RT as the VPN identifier.

It is important to note that this configuration is global. Therefore, if at some point in time, any VPN subscriber would have to be defined by RD, then all the other VPN subscribers must be defined by RD as well. This is a point to consider when designing the initial deployment

Service Control MPLS/VPN Requirements

Topology 

Capacity 

Limitations 

Topology

Following are the general topology requirements for MPLS/VPN support:

The SCE platform is placed in the network between the P routers (Provider MPLS core) and the PE (Provider Edge) routers.

The subscriber side of the SCE platform is connected toward the PE router.

The network side of the SCE platform is connected toward the P router.

The BGP LEG is installed on the SM, and is placed somewhere in the network.

It speaks with the SCE platform through the management IP.

In a cascade installation:

The two SCE platforms are connected to each other via the cascade interfaces.

The data link between the P and the PE is connected via the other interfaces on each SCE platform, as described above:

Subscriber side of each SCE platform connected toward the PE router

Network side of each SCE platform connected toward the P router

The following drawing depicts a typical cascade installation.

Figure 2-1

Capacity

The system supports:

2015 MPLS/VPN subscribers

57,344 different labels (including upstream and downstream, and including the bypassed VPNs)

256 PEs per SCE platform

4 interfaces per PE

Limitations

Mutuallyexclusive system modes

When the system works in MPLS/VPN mode, the following modes are not supported:

Other tunneling modes (MPLS/TE, L2TP, VLAN classify, etc...).

TCP Bypass-establishment

DDoS

Flow Filter TOS rules - When the MPLS/VPN feature is activated, the flow filter mode is automatically switched to tunnel-id. When the feature is de-activated, the flow filter mode remains tunnel-id.

This provides easy configuration of MPLS/VPN. To assure correct and consistent configuration of the TOS/Tunnel-ID mode, the system does not allow configuration of TOS based rules when in tunnel-ID and vice versa

Number of MPLS labels

The choice of the unique VPN site must be based on the BGP label only. The BGP label must be the innermost label.

The MPLS/VPN solution supports various combinations of labels. See Additional MPLS Pattern Support .

The system does not support VPNs for which other MPLS-related features, such as MPLS-TE or MPLS-FRR, are enabled.

Subscriber-related limitations

The following subscriber-related limitations exist in the current solution:

The SM must be configured to operate in Push mode.

VLAN subscribers cannot be used.

Two sites of the same VPN must be aggregated into one subscriber if the following conditions are both true:

They are both connected to the same SCE platform

They both communicate with a common remote site using the same upstream labels and P router.

TCP related Requirements

Number of Upstream TCP Flows - There must be enough TCP flows opening from the subscriber side on each PE-PE route in each period of time. The higher the rate of TCP flows from the subscriber side, the higher the accuracy of the mechanism can be.


hometocprevnextglossaryfeedbacksearchhelp

Posted: Wed May 30 08:40:04 PDT 2007
All contents are Copyright © 1992--2007 Cisco Systems, Inc. All rights reserved.
Important Notices and Privacy Statement.