DE102006045713A1 - Optimized traffic redirection in case of errors for IP networks with additional error correction mechanism on a lower layer - Google Patents

Abstract

The The invention relates to traffic redistribution for fail-safe routes in an IP network formed with links and nodes, the network a first error correction mechanism on the IP layer and a second error correction mechanism underlying the IP layer includes. There are trouble scenarios considered. In this case, in normal operation bandwidth for a fault correction by means of the second Error correction mechanism provided. According to the invention for traffic redistribution to disposal increased bandwidth, by at a fault correction by means of the first error correction mechanism by redistribution of traffic for a fault correction provided by the second error correction mechanism Bandwidth is used to route traffic. According to one Continuing this concept is used in the optimization of link weights applied.

Description

The The invention relates to a method for traffic redistribution for a fail-safe Routes in an IP network formed with links and nodes, where the Net a first error correction mechanism on the IP layer and a second error correction mechanism underlying the IP layer includes.

The currently the most widely used protocol for the transmission of traffic is the so-called Internet Protocol (IP protocol). The IP protocol forms the basis for the global internet, which is a worldwide exchange of data. Originally IP networks were only for usual Data traffic designed, i. those for other traffic types Requirements were usually not met. Current developments go there, IP networks too for to make other types of traffic suitable. It is first Line thought of so-called real-time traffic, i. voice data or video data transmitted with a small time delay acceptable to the service Need to become. In this context, one speaks of quality of service characteristics or of quality of service, which must be guaranteed. A promising one Approach to compliance with quality of service metrics when transmitting from traffic over IP networks is the so-called MPLS (Multi Protocol Label Switching) Protocol. In this protocol, labels are used by the Network leading Defined paths. By defining paths can be Control the bandwidth load to avoid bottlenecks thus guaranteeing compliance with quality of service characteristics can. Typically, then, a reservation will be made in MPLS networks (for example by means of the RSVP protocol), whereby the Total traffic limited and thus the prerequisite for the carriage of the Guaranteed traffic in compliance with quality of service characteristics becomes.

A Property of the conventional TDMS (Time Division Multiple Access) networks is a very high reliability. In the transition from the conventional TDMS technology on IP networks should also have this high reliability be assured again. If possible should the failure of a network element of an IP network not to do so to lead, that quality of service features violated, for example, because packets are discarded or delayed. Therefore, must In IP networks, mechanisms are provided that are available in error cases lead to a rapid redistribution of traffic. Such mechanisms were for For example, MPLS in RFC 4090 (RFC: Request for Comments) proposed. If one network fails, the traffic will be reduced to one Spare path (possibly only a section of the disturbed Paths secured) steered. For this Redirection of traffic must be provided in normal operation bandwidth be available in case of error is.

Next such failure mechanisms at the IP level as e.g. in RFC 4090, there are sometimes additional mechanisms on in the protocol stack further underlying levels or layers. An example is the transport of IP traffic over Sonet / SDH rings (SDH: synchronous digital hierarchy). In the Sonet / SDH ring topology are usually two or more optical fibers (usually in opposite directions) led in a circle. In case of failure of a section a fiber becomes traffic over the other fiber is directed to the destination. In this way, the failure a section corrected at the Sonet / SDH level. This failure of a section corresponds usually the failure of a link at the IP level.

It there is a need for one most efficient Coordination of error correction mechanisms at the IP level with error correction mechanisms lying on lower layers. In front In particular, efficient use of bandwidth in the event of a fault is desirable.

It The object of this application is to specify how the error occurs Bandwidth can be used efficiently if an error mechanism on IP level through a lower layer failure mechanism added becomes.

Of the Invention is based on the observation that for the failure mechanism on a lower level bandwidth can be reserved in normal operation depending on the failure and considered fault scenarios or fault scenarios for a diversion of traffic available stands.

Therefore, according to the invention Bandwidth that is normally used for error correction or fault correction using the error correction mechanism at a lower level (e.g., Sonet / SDH) held in response by the error correction mechanism is used at the IP level to redirect traffic. For example The failure mechanism compensates for link failures at the lower level. additionally is through an error correction mechanism at the IP level the Network against failures other type (for example, node failures) protected. there becomes common a number of likely failure scenarios considered, e.g. the failure of individual links or nodes. At a Reaction at the IP level will then provide bandwidth for an error response provided on the lower level, shared.

The invention has the advantage that in the Feh Langer, which means a loss of available bandwidth, a compensation takes place in that free bandwidth, which was provided for another error case and therefore has remained unused, is used. Bandwidth bottlenecks can be better avoided or the existing bandwidth is better used. As a consequence, the quality of service can be better guaranteed.

According to one Development of the subject invention is the recourse on from for another error case of provided bandwidth for redirection traffic in the optimization and determination of link weights considered. The optimization of link weights is carried out with the proviso that the maximum traffic load of the links of the network in undisturbed operation and for one Number of fault scenarios is limited (e.g., by minimizing them or lowered below a threshold becomes). This optimization or determination is then additionally with the proviso carried out, that for at least one failure scenario, due to reaction or error correction responded by the error correction mechanism at the IP level is for an error correction by means of the further error correction mechanism to disposal bandwidth is used to route the traffic. These linkweights can then assigned to the individual links or for the calculation of paths (e.g., MPLS paths) which are then on the network for the Routes of traffic are set.

A Such definition of link weights or paths in the network in turn leads to a better utilization of the network in case of error. This can Overall, more traffic will be transported or given a given Traffic matrix is the maximum link utilization lower.

The inventive method are preferably on a device centrally responsible for network settings carried out.

Of the Subject of the invention is described below with reference to figures in the context an embodiment explained in more detail. It demonstrate

1 a sonet / SDH ring

2 Layers of an IP network

3 a network section to illustrate the invention.

In 1 a Sonet / SDH ring is shown. The architecture shown is commonly referred to as "Four Fiber Bidirectional." There are four fibers arranged in a ring, with two each designed for traffic in one direction, with one fiber used for both working and standby If the fiber fails during operation (working), it is redirected to the other one (standby), which is routed through one of the nodes usually formed by ADD / drop multiplexers ADM or digital cross-connects DCC The topology shown is the most robust of the currently used Sonet / SDH ring topologies and has been used here because it is particularly suitable for demonstrating the principle of the invention any error correction mechanisms can be used at levels below the IP level, especially for Sonet / SDH ring topology with two fibers instead of four.

In 2 Three layers of a network are shown: Traffic is routed on an IP layer. At this level, an error mechanism (not shown) is provided which, for example, sets up MPLS paths and causes switching to another path in the event of a fault. At a lower level, the Sonet / SDH level, a ring topology is used, so there is a second failure mechanism to insure the failure of portions of these rings. This second mechanism, in effect, provides protection against link failure at the IP level. In the following, it will be assumed for simplicity that this ring topology is that of 1 equivalent. In the context of this embodiment, it is assumed that the amount of failure scenarios considered is given by the failures of individual links or nodes. Since individual link failures are already secured by the lower layer, the mechanism at the IP level is limited to protection against the failure of individual nodes. In the event of a node failure, the bandwidth provided for the interception of link failures at the lower level is shared (ie, the standby fibers not normally shared). This is based on 3 explained in more detail. 3 shows an extremely simplified network configuration, from which the principle of the invention is easier to see. Four nodes are shown, K1 to K4, interconnected by four links L1 to L4. Links and knots together form a ring. It is assumed that this is a Sonet / SDH ring, ie that the lower level links are protected against failure (corresponding to 1 ). That is, each of the links shown represents lower-level realization as four parallel optical fibers on which traffic is transmitted in the opposite direction. The four fibers form four rings, so to speak, of which two are used and the others as a redundant capacity for the compensation of error cases is available.

considered now becomes a path P1, which leads from node 1 via node 2 to node 3. One alternative path is given by the path P2 passing over the Nodes P1, P4 and P3 leads. Path P1 is now protected against failure by path P2 at IP level, i. this second path P2 is a replacement path, which is not in normal operation is used, so traffic on failure of path 1 up Path 2 can be switched. Each of the links have a capacity 2C, where C the capacity a single fiber. (This is based on a definition, associated with a direction with links; the capacity 2C thus takes into account only those for transmission available in the direction of the path Capacity.) Actually used In normal operation, only one fiber will ever be working (working). That the half the capacity (Standby fiber) is used for Linkaus falls to disposal held. The path P1 thus has a capacity C available for the transmission of traffic. Conventional would be Failure of node K2, i. if path P1 fails, the path P2 with a capacity C available.

According to the invention is at an error compensated at IP level additionally the back-up capacity or the for link errors provided additional capacity availed. In case of failure of the node K3 not only the under normal conditions used fibers of P2 but also the standby fibers for redirection of traffic used. So is actually a capacity or bandwidth from 2C for the Diversion available.

It It is easy to see that this is a considerably more efficient use represents the bandwidth, since in this case the back-up path P2 even under normal conditions with traffic (i.e. to a capacity C), because in case of failure of the path P1, the double capacity 2C transmitted becomes.

Usually more complex topological conditions are the rule. For an implementation of the inventive principle is in the context of a development of this for the Determination of link weights used. From the link weights in turn can be determined paths.

Link weights are determined by optimization below. The goal is to minimize the maximum link utilization. The optimization is done in such a way that IP-level troubleshooting uses the unused capacity provided for troubleshooting at lower levels. Specifically, that means in the basis of 1 and 3 described case that in a node error, the double capacity on the links used (ie, not affected by the node error links) is available.

Of the most important application is the protection of node failures IP level and the protection of link failures at a lower level. That this under normal circumstances minimizing to a smaller value of maximum link utilization leads, is easy to see. Because conventional in most cases when a node error reaches the maximum link utilization. The is intuitively understandable that in case of a node error, all are going to or out of the node Links can not be used i.e. traffic to one compared to normal and to a single link failure less number of in case of failure usable links must be distributed. Consequently, one has in this Case also under normal circumstances the highest Link utilization. When using the mechanism for lower Level held backup capacity increased (doubled in the above case) the available capacity in the node failure on the Left. The calculated minimum of maximum link utilization is therefore smaller. Due to the calculated link weights, paths and their usable ones can be used capacity be calculated. Overall, this results in a better load capacity of the network, i. More traffic can be transported through the network be, where for the failure scenarios more bandwidth is available.

A Calculation of link weights k (l), 1 ε L (amount of links) is outlined below.

It let ρ (l, s) the utilization or use of the link l in the case of the scenario s. Let L be the set of links. The set of scenarios S sat down together from the set of link failures S (L), the node or Route failures S (R) and normal operation S (0).

The maximum ρmax (K) of link utilization depends on the set K of link weights k (l) ρmax (K) = max (ρ (l, s, K)), lε L, s ε S, where ρ (l, s, K) is the utilization of the left l in the scenario s.

It Assume that, as in the above example, half of the capacity for the Securing mechanism is kept free on the lower level. The in normal operation available standing capacity Let C. Sei v (i, j) be the one to be transported from node i to node j Traffic (traffic matrix). Then let ρ (l, s, K) be dependent the traffic matrix and the quantities u (i, j, l, s, K) (share of the traffic flow v (i, j), the scenario s via link l is conducted).

For s ε S (L) ∪ S (0), ie for failures of individual links and normal operation ρ (l, s, K) = Σ v (i, j) · u (i, j, l, s, K) / C, sum over i, j.

In contrast, twice the capacity can be used for node failures, ie for s ε S (R) ρ (l, s, K) = Σ v (i, j) · u (i, j, l, s, K) / 2C, sum over i, j.

That the over after failure of a node r (scenario s (r)) still available link l directed portion of traffic between nodes I and j, i. u (i, j, l, s (r), K) can be twice as large as conventional. As a rule, this results in a reduction of ρmax (K).

The Criterion for the calculation of the link weights k (l) and K is then the minimization of the maximum ρmax the link utilization.

The Link weights K are now determined by searching for the minimum of ρmax (K). It can e.g. iterative methods are used in which K is incremental in the direction of lower ρmax (K) is varied. In this case, e.g. also sought a local minimum of ρmax (K) be (simulated annealing) or when falling below a threshold value for ρmax (K) broken off and the value for K can be used. Such methods are known and are not described here.

The Determination of link weights leads to link weights, which allow a transport of more traffic in the network. Based on these link weights, paths (e.g. Dijkstra algorithm).

Claims (14)

A method of traffic redistribution for fail-safe routing in an IP network formed with links and nodes, the network comprising a first error correction mechanism on the IP layer and a second sub-IP error correction mechanism that takes into account: - fault scenarios; in normal operation bandwidth is provided for a fault correction by means of the second error correction mechanism, characterized in that - for at least one fault scenario in which a fault correction by traffic redistribution takes place by means of the first error correction mechanism, bandwidth provided for a fault correction by means of the second error correction mechanism for routing used by traffic. Method according to claim 1, characterized in that that took into account Fault scenarios by the failure of at least one node or a link are given. Method according to one of the preceding claims, characterized marked that - at fault scenarios with failure of at least one node, the first error correction mechanism is used, and - at fault scenarios where only one or more links are affected, the second Error correction mechanism is used. Method according to one of the preceding claims, characterized marked that - of the Failure of one or more links not in the context of the first error correction mechanism considered becomes. Method according to one of the preceding claims, characterized marked that - the considered Failure scenarios are given by the scenarios with failure of one Node or a link. Method according to one of the preceding claims, characterized marked that - in the Frame the second error correction mechanism on the SDH / SONET layer Traffic is redirected. Method according to Claim 6, characterized that - the Diversion over is provided by at least one SDH / SONET ring provided spare bandwidth. Method according to one of the preceding claims, characterized marked that - in the network MPLS paths are set, and - in the course of the first failure mechanism Traffic over at least one path is redirected, which is a replacement path for at least represents part of an MPLS path. A method for optimizing link weights for routing defected by a method according to any one of claims 1 to 8 in an IP network formed with links and nodes, the network having a first error correction mechanism on the IP layer and a second, below the IP layer lying error correction mechanism, in which - fault scenarios are taken into account, - in normal operation, bandwidth is provided for a fault correction to be carried out by means of the second error correction mechanism in the context of at least one of the considered fault scenarios, - The linkweights are optimized with the proviso to limit the maximum traffic load of the links of the network for undisturbed operation and the fault scenarios, characterized in that - the optimization is carried out according to the proviso that for at least one fault scenario, in which by means of the first Error correction mechanism takes place by means of traffic redistribution error correction, for a correction of errors provided by the second error correction mechanism bandwidth is used for routing traffic. Method according to claim 9, characterized in that that the maximum traffic load of the links of the network for undisturbed operation and limited the error scenarios is minimized or undercut by the maximum traffic load a threshold the optimization is terminated. Method according to one of claims 9 or 10, characterized that - one Setting the link weights for the links of the network is done by the optimized link weights the Be assigned to the left of the network. Method according to one of claims 9 to 11, characterized that - the optimized link weights for the determination of paths can be used. Device with means for carrying out a Method according to one of the claims 1 to 13. Device according to claim 13, characterized in that that the device is a central device of the network for making of network settings.