EP1518353A2

EP1518353A2 - Adaptive control of a network element

Info

Publication number: EP1518353A2
Application number: EP03761480A
Authority: EP
Inventors: Joachim Charzinski; Karl Schrodi; Christian Winkler
Original assignee: Siemens AG
Current assignee: Siemens AG
Priority date: 2002-06-26
Filing date: 2003-06-20
Publication date: 2005-03-30
Also published as: US20060023741A1; WO2004004215A2; AU2003249858A1; WO2004004215A3; CN1666461A

Abstract

The invention relates to the autonomous, adaptive control of a network element in a packet-oriented and connectionless communication network, and to the coupling of a plurality of network elements.

Description

description

Adaptive control of a network element

The subject matter of the application relates to a method for adaptively controlling a network element in a communication network and a method for coupling a plurality of network elements.

There is the concept of configuring network nodes with rules specified by an administrator and stored in a database. This, policy-based networking ^{Λ of} the IETF (Internet Engineering Task Force) serves on the one hand to load quasi-static configuration information into the network nodes. On the other hand, configurations that have to be set depending on specific connection requests can also be given to the nodes at the time of request. For this purpose, a component superordinate to the network is introduced, the Policy Decision Point ^λ PDP, which can read the specified rules from the database and search for the rule suitable for the given situation. As a result, he loads a corresponding one

Configuration information in the network element PEP (for:, Policy Enforcment Point) -

If the PDP is only used for static configuration, it is no longer involved in normal network operation. From then on, the nodes work independently of the network control, but cannot react autonomously. If, on the other hand, incoming connection requests have to be processed (eg RSVP Resource Reservation Protocol), the PDP is the central component of network operation. The rules in the PDP database are created by an administrator, ^• if necessary, automatically checked for consistency and prioritized by the PDP in the event of conflicts according to predefined schemes.

The invention is based on the problem of a method for controlling a network element of a communication network to indicate that with changing operating conditions, such as. B. load change, line break, node failure, responded with a fast autonomous forwarding of data packets.

The problem is solved by an object with the features of claim 1.

According to the invention, a network element in an autonomous communication network is controlled via rules of conduct. For this purpose, the network element is assigned a control instance, Network Control Server (NCS) ^λ , which generates these rules and configures the network element with it. With this approach, the network works without the permanent intervention of the NCS. The NCS only supplies the network elements accordingly if new, adapted rules are required due to sustainable, long-term changes to the network situation.

Based on the knowledge that the operation of a network causes the higher the costs, the more administrative effort has to be made, the approach according to the invention of autonomously working network elements is advantageous from the outset. The method described here of generating the rules automatically minimizes operator costs while increasing availability. By coupling NCSs of several network elements, (sub) networks can also be operated across the board and represent the quality properties required by the user end-to-end with minimal administration effort.

The invention has the following properties:

The network works in a packet-oriented and connectionless manner. The network has network elements that operate autonomously using rules. - With the help of these rules, the network elements can forward packets according to specified criteria (eg quality of service), in particular also via several sensible possible routes (eg for uniform Load distribution) autonomously in normal operation. In addition, they react very quickly autonomously to network failures (e.g. line break, node failure).

In a special embodiment of the invention, the rules of behavior are formed and maintained in a control instance (NCS) individually assigned to a network element. In this way, in a communication network comprising a large number of network elements, one or more or all network elements each generate rules of conduct for themselves, from which it autonomously / automatically selects according to the operating conditions.

An embodiment in which a network element has an individually assigned control entity (NCS) fits into the concept of a non-hierarchical network architecture in which the respective network element has full functionality. A network element thus has the functionality of the control entity (NCS), comparable to the link-state information held in the network element regarding the availability of the connected connecting lines.

Advantageous further developments of the subject of the application are specified in the subclaims.

The invention is described below as an exemplary embodiment to the extent necessary for understanding on the basis of

Figures explained in more detail. Show:

1 shows a schematic representation of the network elements according to the invention in the communication network,

Fig. 2 is a schematic representation of the network and control hierarchy

Fig. 3 input and output variables of the adaptive network control.

In the figures, the same designations denote the same elements. The invention shown here describes the adaptive control. The coupling is also addressed by several networks.

The network elements acting autonomously according to the invention in a network (hereinafter referred to as autonomous network) work under the guidance of an adaptive control but without their permanent intervention.

The elements of the autonomous network (see Fig. 1) are: on the one hand, the network nodes that autonomously transmit / forward the traffic (router), which are differentiated into edge nodes (edge router) and core nodes (core router), - on the other hand, the resource control entities ( RCA), which are arranged on the edge of the network.

The RCAs are assigned to the edge nodes. Your task is to accept resource requests (e.g. connection setup / cleardown) at an assigned input or output edge node (e.g. from a separate service control (not shown here) (see Fig. 1, ©)) and to allow them to be admissible and feasible check to accept or reject them. As a result, the RCA provides the corresponding edge node with parameters (see FIG. 1, ®) which enable it to set the use and usage monitoring of the resources and to configure the rules for handling the data packets associated with the corresponding traffic flow (eg marking, Policing, scheduling).

Like the routers, the RCAs work autonomously based on rules of conduct. These rules of conduct describe their control task and contain the parameters to be passed on to the Edge Router during operation explicitly or implicitly (e.g. as a calculation rule).

There are several options for implementing an RCA: • as a separate server

• integrated into an edge router

An RCA can be responsible for: »One edge router each

• for several edge routers

The elements (router, RCA) of the autonomous network work according to rules of conduct. These can be given by the NCS to the network elements or otherwise, e.g. via the network management. The NCS can therefore be responsible for:

• CoreRouter (see Fig 1, ®)

• Edge router (see Fig. 1, ©)

• RCAs (see Fig. 1, ©) • any combination

Network and control hierarchy

There are four levels for the network and control hierarchy, each with different priorities / objectives for control and a different temporal behavior. From bottom to top, these are (see Fig. 2):

• The transmission infrastructure

• The autonomous IP network • The adaptive control of the network elements

• The network management

The transmission infrastructure is primarily responsible for the transmission of the data and may include mechanisms for a very fast equivalent circuit in the event of a fault (e.g. line break etc.), e.g. for SDH or similar approaches in the area of optical networks. This is a control task that is performed independently by the transmission infrastructure in the millisecond range.

The autonomous IP network described above autonomously processes resource requests, a control task in Cooperation with a service control, distributes the traffic in the network and reacts quickly and independently to errors. Only those errors are processed that could not already be eliminated on the transmission level.

In contrast to the lower two levels, the adaptive network control (regulation) according to the invention has no real-time requirements. It monitors the network and generates new rules if there are significant deviations from the target operation. The time horizon is hours or more.

The network management serves to set the basic configuration in the direction of the network. As a rule, it is only active at very large intervals, e.g. when expanding the network.

There are several options for implementing an NCS:

• as a separate server

• assigned to a respective network element, e.g. B. integrated

An NCS can be responsible for:

• one network element each

• for several network elements

Rules and rule generation

A. Quasi-static rules:

In the simplest case, these rules are quasi-static, so they only depend on the network topology and the static properties of the network.

In contrast to, policy based networking ^Λ but they are not fixed by an administrator but automatically generated by NCS.

The NCS can provide the basic information, e.g. from

Network management and / or obtained from the network element (s) itself. This may include: network topology, Line bandwidths, properties of the network element (s), (preferred) routes, traffic matrices, traffic classes, etc.

When changing this basic information, e.g. Changes to the network topology, the rules are recalculated accordingly and loaded into the respective network element.

The rules are set in such a way that the network can ensure the properties described above in autonomous operation. The NCS is basically not part of regular operations.

B. Dynamic rules:

In this more complex case, the rules are also adaptively changed, adapted or generated depending on the network state. It should be noted that the rules are adjusted on a larger time scale (e.g. 15 minutes or 2

Days) and the network continues to respond to dynamic changes

(including errors) reacted quickly autonomously.

Two variants are conceivable for the rule generation by the NCS:

• The NCS selects rules from a set of predetermined rules according to the network situation. • The NCS also adjusts the predetermined rules according to the network situation.

• The NCS generates rules according to the network situation.

Information from the network is e.g. Statistics about traffic and queues, error messages from the network, current route guidance, etc. impending permanent unbalanced loads (e.g. due to extensive failures or permanent changes in user behavior and traffic matrices) are corrected.

Degree of information and intelligence of the, NCS Graduated embodiments are possible for the NCS, which differ in the dimensions of information level and intelligence. The more information (sources) available to the NCS, the more optimized rules it can generate. This is closely linked to the necessary and possible intelligence of the NCS, which can range from simple logic, optimization procedures and dimensioning procedures to expert systems or neural networks. With increasing information, the need for intelligence of the NCS increases.

Possible sources of information (also in various combinations) include the network elements themselves (e.g. statistical information, network load, routes), network management (e.g. topology, error events), administrator inputs, static and dynamic basic data (e.g. traffic matrices).

Information flows from / to adaptive network control (NCS): The NCS obtains information from several sources for its task and also supplies data to different customers (see Fig. 3).

Input:

• Network management / operator (among others): • Network operating strategy

• Network configuration

• Additional configuration data (e.g. network segments to be specially protected, etc.)

• Autonomous network (among others):

• Statistics

• Operating ^states • • Routes

• Service provider (among others):

• Information about services and applications and their properties and requirements Output:

• Network management / operator (among others):

• Info for operators, e.g. Need to expand the network etc.

• Statistics

• Events

• Autonomous network (among others) • Rules of conduct

• parameters

Coupling of (sub) networks

If several networks that work according to the autonomous principle described are closely linked so that they can represent the properties of the autonomous network such as load distribution, error response and quality of service, the rules of the subnetworks must be coordinated.

For this purpose, the NCSs responsible for a (sub) network are linked to each other using a suitable protocol and exchange information to compare the rules. Then, as described above, they create adapted rules and thus supply the network elements of their (sub) network.

Options and extensions

• Each network node receives an individual set of parameters / all network nodes receive the same parameters.

• The parameters also include the selection of an algorithm if several algorithms can be used to handle the task. • The NCS can be centrally located in the network / there are one or more backup devices / there are several equal coordination devices that work with With the help of a special coordination protocol, their rule specifications are mapped / different areas of the network or network elements are controlled locally by various NCSs that communicate via a special communication protocol.

• The rules are changed depending on the load on one or more links.

• The rules are changed depending on an observed quality of service. • The rules are changed depending on the queue lengths observed in the network nodes.

• The NCS is used to additionally specify the parameters in devices for connection acceptance control at the network edge. • The NCS communicates with other network control servers in the networks of other network operators.

• The NCS generates current tariff information from the existing status information (and possibly other parameters supplied by network management), which it forwards to the transport control (RCAs).

• Communication between network elements and NCS can originate from the NCS or from the network elements. In the first case, the NCS actively supplies the network elements with new rules and / or parameters as soon as they are available. In the second case, the network elements can call up the current rules / parameters if required. Both forms of communication can be used in a network, the second being particularly suitable for the automatic configuration of new network elements (e.g. when starting up or restarting) and / or for configuring the communication parameters for the first form of communication.

• When creating the rules / parameters, the NCS takes into account the order in which the new rules / parameters are imported into the network elements. Since not all network elements absolutely new rules / parameters at the same time such an intelligent coupling of creation and distribution of the rules / parameters can help to avoid transient states of overload or instability.

Claims

claims

1. A method for controlling a network element in a communication network, accordingly - a network element holds a plurality of rules of conduct and the network element autonomously / automatically selects a behavior rule in accordance with the operating conditions and forwards data packets in accordance with this behavior rule.

2nd Method according to claim 1 d a d u r c h g e k e n n z e i c h n e t d a s s s the operating conditions by any combination of line break, node failure, network utilization,

Connection establishment, network reconfiguration are given.

3. The method according to any one of the preceding claims, characterized in that a behavioral rule includes the selection of one of several ways.

4. The method according to any one of the preceding claims, characterized in that the rules of behavior are formed in a control instance (NCS).

5. The method according to any one of the preceding claims, characterized in that the rules of behavior are formed in a control entity (NCS) individually assigned to a network element.

6. The method according to any one of claims 1 to 4, characterized in that the rules of conduct can be fed to the network element by a control entity (NCS) superordinate to a number of network elements.

7. The method according to any one of the preceding claims, characterized in that the rules of behavior are generated automatically.

8. A method for coupling a plurality of network elements, in particular according to one of the preceding claims, according to which two control entities (NCS) are coupled to one another via a protocol, by means of which they exchange information for comparing rules of conduct.

9. The method according to any one of the preceding claims, characterized in that it is used in a packet-oriented and connectionless communication network.