MPLS-Part II_MPLS with Traffic Engineering.pdf

THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II MPLS with Traffic Engineering (MPLS-TE) recovery mechanisms TE Objectives “A major goal of Internet Traffic Engineering is to facilitate efficient and reliable network operations while simultaneously optimizing network resource utilizatio...

THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II MPLS with Traffic Engineering (MPLS-TE) recovery mechanisms TE Objectives “A major goal of Internet Traffic Engineering is to facilitate efficient and reliable network operations while simultaneously optimizing network resource utilization and performance.” RFC 2702, Requirements for Traffic Engineering over MPLS Defines how the flows should be routed to use the network resources efficiently (to avoid congestion), including reliability. - BW Optimization (traffic load balance, better network resource utilization). - Strict QoS guarantee. - Fast recovery in case of fault. Classic fish problem: - Routers always forward traffic along the least-cost route as discovered by intradomain routing protocol (IGP). - Network bandwidth may not be efficiently utilized: o The least-cost route may not be the only possible route. o The least-cost route may not have enough resources to carry all the traffic. o Some links are overused while others are underused. Traffic Engineering (TE) is used when the problems result from inefficient mapping of traffic streams onto the network resources. In such networks, one part of the network suffers from congestion during long periods of time, possibly continuously or due to changing traffic patterns, while other parts of the network have spare capacity. - Reduce the overall cost of operations by more efficient use of bandwidth resources. - Prevent a situation where some parts of a service provider network are over-utilized (congested), while other parts remain underutilized. Basically, TE helps you to optimize your network resources utilization, provide a better quality of service and enhance the network and services availability. Solution: - Tunneling techniques between source and destination such that the intermediate nodes do not interfere in the routing decision. Example at L2: ATM PVCs/SVCs -->Additional cost (equipment, net operation). Scalability problems. - MPLS-TE: Each LSP (label Switched Path) has its own constraints: BW, affinities, routing constraints, etc. The LSP is computed to satisfy the set of requirements. Components of MPLS-TE 1. LSP-TE configuration in the ingress LERs 1. Configure attributes 1. Destination address, BW, required protection and restoration, affinities, affinity relations, etc. 2. Distribution of resource and topology information (e.g. OSPF with TE extensions, that reflect link characteristics and reservation states). 3. Computation of LSP-TE 1. Each router uses its routing table and topology and resource database to, taking into account the LSP-TE constraints, compute the constrained shortest path (CSPF). 2. Define recovery mode of LSP in case of fault. THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II 4. LSP-TE setup 1. Once calculated, established with RSVP-TE. 5. (Labeled) packet forwarding (intermediate routers do not make any routing decision). OSPF-TE: Resource and Topology Information Traffic engineering information is carried in OSPF traffic engineering (OSPF-TE) link state advertisements. OSPF-TE LSAs are Type 10 Opaque LSAs, as defined in RFC 2370. Type 10 Opaque LSAs have area flooding scope. OSPF-TE LSAs have special extensions that contain information related to traffic engineering; these extensions are described in RFC 3630. The extensions consist of Type/Length/Value triplets (TLVs) containing the following information: - Link ID - IP address of the local and the remote interface for the link - TE metric for the link (by default, this is equal to the OSPF link cost) - Maximum bandwidth on the interface - Maximum reservable bandwidth on the interface - Unreserved bandwidth on the interface (BW not yet reserved) - Administrative group(s) to which the interface belongs: Affinity classes, Colors, Shared Risk Link Groups (SRLG) Resource and Topology Information Each interface may be assigned to one or multiple administrative groups Colors are often used to describe these groups (Gold, Silver, Bronze). User-friendly names can be used as well (Voice, Management, BestEffort). Names are locally significant to the router. When an ingress LSR calculates the path for an LSP-TE, it takes into account the administrative group to which a interface belongs. It is possible to include or exclude administrative groups. Resource and Topology Information Affinity - Value and mask - Can be used for: cost of the tunnel, physical types of links, physical distance, etc. Shared Risk Link Groups (SRLG) - SRLGs is a characteristic of a link that the network administrator assigns indicating that link share a common fiber or conduit. - The SRLG is flooded by the IGP and is used when backup tunnels are deployed. - Configuring SRLG membership enhances backup tunnel path selection so that a backup tunnel avoids using links that are in the same SRLG as interfaces the backup tunnel is protecting. THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II OSPF-TE: Resource and Topology Information The following events trigger the device to send out OSPF-TE LSAs: - Change in the interface’s administrative group membership - Change in the interface’s maximum available bandwidth or maximum reservable bandwidth - Significant change in unreserved bandwidth per priority level: o The triggering module decides when to advertise changes in link states. o Set a link's reserved bandwidth thresholds - In addition, OSPF-TE LSAs can be triggered by OSPF; for example, when an interface’s link state is changed. When an interface is no longer enabled for MPLS, the device stops sending out OSPF-TE LSAs for the interface. TE Link State database LSRs use IGP extensions to create and maintain a TE Link State database (TE-LSDB or TED) that contains the TE network topology that is updated by IGP flooding whenever a change occurs (establishment of new LSP, change of available bandwidth). LSP Attributes and Requirements Used for TE In addition to the topology information in the TED, the following userspecified parameters are considered when the ingress LSR calculates a TE path for a signalled LSP: - Destination address of the egress LER - Explicit path to be used by the LSP - Class of Service (CoS) value assigned to the LSP - Bandwidth required by the LSP - Setup and Hold priority for the LSP. o The hold priority specifies how likely an established LSP is to give up its resources to another LSP. To be preempted, an LSP must have a lower hold priority than the preempting LSP's setup priority. 0 (highest priority) 7 (lowest priority). o An LSP's setup priority must be lower than or equal to its hold priority. By default, an LSP's setup priority is 7 and its hold priority is 0. - Metric for the LSP - Including or excluding links belonging to specified administrative groups CSPF algorithm Using information in the TED, as well as the attributes and requirements of the LSP, CSPF calculates a traffic-engineered path - If more than one LSP needs to be enabled, select the LSP to calculate based on its setup priority and bandwidth requirement When the ingress router invokes the CSPF algorithm, it creates a subset of the TED information based on the provided constraints 1. Prune all links which don’t have enough reservable BW 2. Prune all links which don’t contain an included administrative group color 3. Prune all links which do contain an excluded administrative group color 4. Calculate a shortest path from the ingress to egress using the subset of information THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II LSP-TE setup Once the path is computed - Need to establish forwarding state along the path - Reserve resources along the path Two approaches - RSVP extensions - CR-LDP (Constraint-Routing LDP) RSVP extensions - How to send PATH messages on explicit routes? Introduce new object ERO (Explicit route object) similar to source routing - PATH messages (from head to tail) carries LABEL_REQUEST - RESV message for label binding CR-LDP - In addition to using label_request and label_mapping messages, use ER message similar to ERO LSP-TE setup: RSVP 1. The ingress LER sends an RSVP Path message towards the egress LER. - The Path message contains the traffic engineered path calculated by the CSPF process, specified as an EXPLICIT_ROUTE object (ERO). The Path message travels to the egress LER along the route specified in the ERO. 2. The Path message requests resource reservations on the LSRs along the path specified in the ERO. RSVP vs CR-LDP THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II LSP Admission control On receipt of PATH message. - Router will check if there is bandwidth available to make the reservation. - If bandwidth available then RSVP accepted. - PATH message is sent to next hop (downstream). On receipt of a RESV message. - Router actually reserves the bandwidth for the TE LSP. - Label allocated and RESV message is sent to upstream node. - If pre-emption is required lower priority LSP are torn down. OSPF/ISIS updates are triggered. Recovery techniques in MPLS-TE To ensure the reliability of LSP-TE it is important to define how the LSPs will be recovered in case of faults. Global recovery - Restoration: Default recovery mode in MPLS-TE, in which the failure is notified to the ingress LER through RSVP and a routing protocol (IGP) which is used to recalculate a new path. - Protection: The ingress LER establishes two LSP-TEs, one primary and one backup. Local recovery - Protection (Fast Reroute): The LSP affected by the fault is locally rerouted by the node immediately before the failure (node or link) in the upstream direction. Recovery Time Fault Detection Time: depends on the fault detection mechanism in use and the underlying layer 1 and layer 2. Hold-Off Time: permits that MPLS waits before trying to reroute. Useful if the inferior layers have a recovery scheme. Fault Notification Time: time for the FIS (Fault Indication Signal) to be received by the node in charge of the traffic recovery. Recovery Operation Time: synchronization between network elements to coordinate. Traffic Recovery Time: time from the last recovery action until the traffic is completely recovered. Global MPLS-TE Restoration 1. Configuration of a LSP-TE in the ingress LER (destination, BW, priority, protection and restoration requirements, etc.) 2. Node R3 detects the fault in the link R3-R4 and sends a FIS (Fault Indication Signal) (RSVP message or IGP update) to the ingress LER (R1). 3. Node R1 has to find another path that complies to the constraint requirements using a CSPF (Constrained Shortest Path First) algorithm. 4. The new LSP is signalled and the traffic is rerouted. Advantages - Does not require any additional configuration of backup paths (even though these paths could have been precalculated). THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II Drawbacks - Slowest recovery mechanism compared to other protection mechanisms, given that it implies the transmission of the Fault Indication Signal (FIS) to the ingress LER, a dynamic path computation and the signalling of the new LSP-TE. - Lack of predictability. These is no guarantee that the LSP-TE could be rerouted upon failure. A last-resort option is to relax the TE LSP contraints. Global MPLS-TE Protection The ingress LER calculates and signals a backup LSP-TE for each primary LSP before any fault is produced (both LSPs need to be disjoint paths in the elements against which the protection is defined). The constraints for the primary and backup LSP can be equal or different. When the ingress LER receives the FIS, it switches the traffic to the backup LSP. Advantages - In networks with many links and nodes and limited number of TE LSPs to protect, this mechanism is easy to deploy and requires a limited amount of provisioning. On the contrary, the use of local protection would require the protection by tunnels of every network element. - Given that the backup LSP is signalled before the failure, the path is deterministic and provides a strict control of the backup path. Drawbacks - Requires doubling the number of LSP-TE, which has an impact on scalability in full mesh networks. - In many cases cannot provide a recovery time in ms, given that it has to inform the ingress LER. - Given that the primary and backup LSP need to be disjoint, it can be that the primary does not follow the optimum path. Local MPLS-TE Protection Local protection techniques - Local: the LSP is rerouted by the node immediately upstream to the node or link that has failed. The rerouting is produced in the place closest to the fault, instead of in the ingress LER. Eliminates the need to send a FIS to the ingress LER (faster recovery time). - Protection: A backup resource is pre-assigned and pre-signalled before the fault. No need to calculate this in case of a fault. Techniques called Fast-Reroute - One-to-One backup (or detour) - Facility Backup (or bypass) LSP R1-R2-R3-R4-R5 is “fast reroutable” if it is signalled with a set of specific attributes in the RSVP Path message. - Indicates the wish to benefit from local recovery in case of a fault. - The LSP affected by a fault is locally rerouted by the node immediately upstream to the fault ->PLR (Point of Local Repair), (R2, if fault in R3 or in link R2-R3) THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II Local MPLS-TE Protection: Fast Reroute Fast-Reroute uses backup tunnels to reroute affected net elements - When the backup tunnel ends in the next hop to the PLR, it is called a NHOP (NextHOP) backup tunnel. - If it ends in a neighbour of the next node to the PLR, it is called a NNHOP (Non NextHOP) backup tunnel. In this case, the node is called a Merge Point. Fast-Reroute: One-to-One backup At each hop, one backup LSP (called detour) is created for each fast-reroutable LSP-TE. If it only protects the LSP-TE T1, T2 and T3 against faults in the link R3- R4 (Link Prot.) and in the node R4 (Node Prot.). Each node within the fast-reroutable LSP-TE has to realize the same operation. Each detour must avoid the resources against which it protects. Operation mode - Assignment of labels for the primary LSP-TE, T1, and for the detour, D1. - When a fault occurs in R4, PLR R3 detects the fault and reroutes the traffic to the detour with its labels. Fast-Reroute: Facility backup A NHOP tunnel is required to protect against a link fault: Link Protection A NNHOP (bypass) tunnel is required to protect against a (bypassed) node fault (Node Protection) - This also protects against the link between this node and the one immediately upstream (Link Protection). Backup tunnels can be shared. Advantages: - Requires a smaller set of backup tunnels if the BW protection must be guaranteed. - The number of backup tunnels is not a function of the number of LSP-TE in the MPLS network, which preserves scalability. Requires a NHOP tunnel to protect against a link fault (Link Protection). - When the PLR detects a fault, it realizes the same label exchange as before the fault but also stacks the backup path label. Requires a NNHOP (bypass) tunnel to protect against a (bypassed) node fault (Node Protection). - The PLR puts the label that the MP (Merge Point) is waiting to receive and adds the label for the backup path. This way, the MP receives the same packet but through another interface. o The PLR needs to get to know the NNHOP LSR label through RSVP extensions. Stacking of labels in MPLS to reroute different LSP-TE over the same path. - The highest label is the backup tunnel. THE EVOLUTION FROM CIRCUITS TO PACKETS: MPLS II Path Reoptimization Once the fault detection is realized and backup tunnels are being used, it is possible that not optimum paths are used - The PLR (R1) sends a RSVP Path Error message to the ingress LER to indicate that a local reroute has occured. This provokes the search for an alternative path Fast Reroute using SRLG Configure routers that automatically create backup tunnels to avoid MPLS TE SRLGs of their protected interfaces. Backup tunnels provide link protection by rerouting traffic to the next hop bypassing failed links or avoiding SRLGs. Local MPLS-TE Protection Advantages - Local protection mechanism that can provide very fast recovery times, equivalent to SONET/SDH. o Facility and one-to-one backup are equivalent in terms of recovery time. - Can provide guaranteed BW, propagation delay and jitter in case of a fault. In the facility method, the backup capacity can be reduced drastically because backup tunnels can be shared. - Offers high granularity in the concept of Class of Restoration (CoR). - The facility method has a high scalability, given that the number of backup tunnels is a function of the number of network elements to be protected and it is not a function of the number of fast-reroutable LSP-TEs. Drawbacks - Requires configuration and establishment of a number of backup LSP-TEs which in large networks cannot be negligible. - The one-to-one method has a limited scalability in large networks. Comparison of Global and Local Protection Objective: to evaluate the number of required backup tunnels with global path protection, Fast Reroute facility backup, and one-to-one, based on the following assumptions: Network description: - D: network diameter (average number of hops between a head-end LSR and a tail-end LSR). - C: degree of connectivity (average number of neighbors). - N: total number of nodes (LSRs). Others parameters: - L: total number of links to be protected with Fast Reroute L