EP1428362A2

EP1428362A2 - Connection-free and packet-oriented network

Info

Publication number: EP1428362A2
Application number: EP02776703A
Authority: EP
Inventors: Karl Schrodi
Original assignee: Siemens AG; Nokia Siemens Networks GmbH and Co KG
Current assignee: Nokia Solutions and Networks GmbH and Co KG
Priority date: 2001-09-20
Filing date: 2002-09-20
Publication date: 2004-06-16
Also published as: AU2002339314A1; WO2003026230A3; WO2003026230A2

Abstract

The invention concerns a connection-free and packet-oriented network transporting data packets in accordance with an Internet protocol (TCP/IP), which includes particular boundary node and network configurations, thereby ensuring the type of quality of service as per the agreement between the user and the network operator.

Description

description

Connectionless, packet-oriented communication network

The subject matter of the application relates to a node, an access node for a connectionlessly operated, packet-oriented communication network and a connectionlessly operated, packet-oriented communication network.

The convergence of telecommunications (also called voice networks) and classic data world (also called data networks) to IP (Internet Protocol) based networks and services is a challenging task with regard to IP technology, as this originally primarily focused on 'best effort' transmission is designed and at most provides for the rather vague 'Service Level Agreements' (SLA), while in telecommunications very stringent requirements with regard to compliance with a promised connection quality QoS (Quality of Service), reliability, availability and security of network and services are in the foreground. The 'Internet world' is responding to this task with a multitude of increasingly complex and complex solutions, but has not yet found a complete solution that is also manageable and viable in economic terms.

The QoS requirements of a service or application to a network can be defined using various criteria, some of which are given as examples:

the throughput characteristic of the digitally coded information, ie the required bandwidth or bandwidth characteristic (fixed bandwidth, variable bandwidth [eg with mean value, peak value, 'burstiness factor' or the other characterizing parameters]) and the sensitivity to information losses, - The delay characteristic, ie the effects of an absolute delay (running time from the source to the sink of the information) and the sensitivity to fluctuations in the running time or delay (of course, delay fluctuations can be converted into an absolute delay by 'buffering' - this is usually very time-consuming ),

- the required or unnecessary temporal consistency or 'time invariance' of the information transmitted,

• i.e. whether the information units must be delivered in the exact same order in which they were delivered or not (the ability - or inability - of higher service and application layers may have to be taken into account).

The consequences that result from different QoS requirements are illustrated using two examples:

I. Although unidirectional audio / video applications (eg ^' streaming video') require a real-time presentation to the recipient, in most cases it will be irrelevant whether the absolute delay is 1/100, 1 or 5 seconds , provided there is continuity after the start of the play. This delay tolerance could be used, for example, to compensate for loss of information with the help of repetitions and thereby improve the quality. Alternatively, transmission could also be carried out with redundancy (higher bandwidth) in order to compensate for possible data losses.

II. Interactive, ie bidirectional real-time communication (voice, video, ...) between people must take into account the responsiveness and the typical communication and dialogue behavior of people. Here must the absolute delay (and of course the delay fluctuations) are limited to a few hundred milliseconds (eg 200 ms). On the other hand, somewhat higher loss rates can be tolerated, since the ability of the human brain to 'smooth unevenness' is very pronounced in speech and visual perception, and in dialog the attention to small defects is somewhat reduced. However, real-time dialogues between machines are more complex. Then, under certain circumstances, both the completeness of the information and slight delays close to the physical limit due to the spatial distance (running time approx. 5 ms per 1000 km distance) must be targeted.

If the QoS requirements are defined, a network, if it still has reserves in one of these areas, can also use these to compensate for deficits in another area. This compensation is explained using two examples:

I. If an application tolerates relatively high information losses, the delay fluctuations can be reduced by rejecting information units that have received a high delay. Conversely, larger delay fluctuations can of course also be used to achieve lower losses, but this leads to large buffer memories.

II. If the maximum of the delay fluctuations lies below the minimum time interval between the information units delivered (so-called 'fast network'), problems with the time consistency of the transmitted information can be excluded. If measures to restore this temporal consistency are provided, relatively large fluctuations in delay can be tolerated as long as the framework of the absolutely permissible other delay is not exceeded.

In addition to the QoS, the general availability of the services is also an essential parameter, which is highly dependent on the network and its properties. Is in the event of an error, e.g. in the event of failure of individual network components or connecting lines, an alternative route is available and how quickly can this be used? Are there any noticeable interruptions for the user and how long are they? Does the network operator or even the user have to intervene in any way to restore the service? The reliability of the network itself and the way in which it can help to bridge errors and possibly restore applications is of great importance.

A uniform network must therefore be considered in a qualified manner with the boundary conditions of the type presented here - and of course it should also be as efficient as possible, i.e. with possible can be solved with little effort and economically advantageous.

So far the following network technologies are known:

1. Line switching technology

The simplest approach. is the tried-and-tested technology of line switching, in which a dedicated path (a 'connection') with a dedicated and absolutely reserved bandwidth is switched for each communication relationship (in the bidirectional case or, in the case of multiple relationships, possibly two or more). Such paths are either designed explicitly as individual physical lines (for example copper wires) or as channels in so-called transmission or switching systems that allow multiple use of physical lines. It is also possible to mix with different sections. The possible data throughput of such a path is determined by the bandwidth allocated or allocated to it, the delay time for the transport consists of the 'Propagation Delay', ie the distance-dependent runtime on the line, and the 'Switching Delays', ie the inherent processing times that are involved in 'communicating' the digitally coded information ('data') in the network nodes (switches) arise together. 'Transfer' here means converting the information ('data') from a specific incoming line / channel to an outgoing line / channel defined when the connection is established. Both deployment components can generally be assumed to be constant (that is, if the systems are operating properly) for the duration of a communication relationship (with the path switched through or an existing 'connection'). In the fault-free case, the same quasi 'optimal' QoS is specified and achievable ^' for all applications (no loss of information, constant, generally relatively small delay, no mix-ups). For this, however, the 'line' (the path) must be permanently switched (and reserved) for the duration of the communication relationship ('connection'), even if the application uses it very little (eg only sporadically). The reliability / availability can be improved in that the alternate path is switched provided in case of failure rapidly as possible to a pre-, (twice the capacity needed) or a ^Ers' is atzweg connected immediately (delay and effort, especially if due to a failure many connections are affected at the same time).

2. Parcel switching technology

The packet switching technology ^' aims at a better and more flexible use of the resources (bandwidth) through quasi simultaneous use (sharing) of (virtual) lines and channels or switching and transmission media through multiple communication relationships. The two most important representatives are ATM technology (with 'cells' of fixed length and connection-oriented') and IP technology (with packets of variable length and 'connectionless'). ATMU is also used by ITU-T under this term and with the aim of a 'broadband ISDN' (B-ISDN). ATM has very sophisticated mechanisms to provide a wide range of service classes with defined and (on average) guaranteed QoS even with very scarce resources (available bandwidths). The resulting systems and networks, however, become quite complicated and expensive. Sizing and operation require well-trained specialist personnel. ATM works connection-oriented, with a network of 'virtual' paths and channels that are hierarchically assigned to each other. Bandwidths can be reserved individually for a variety of service classes and can also be 'guaranteed' in accordance with the underlying traffic statistics. For this purpose, sophisticated queuing and scheduling mechanisms are used, which are set in each node per path and channel (connection) by means of appropriate parameters. By means of highly complicated dimensioning and connection acceptance regulations, information losses and the variable proportions of the switching delays (which are essentially determined by queuing) can be limited in accordance with statistical rules. An exchange of information units is not to be expected due to the connection orientation. As a result of the connection orientation, however, all inherent complexities have to be run through again when handling error cases. The basic ideas are often very similar to those of circuit switching technology.

IP technology is a rather pragmatic approach that has established itself in the data world due to its simple basic mechanisms. It has made massive progress in recent years so that systems and networks based thereon can be compared in terms of their performance (data throughput, control efficiency) with systems based on ATM technology. The success of IP technology is significantly due to the fact that a large part of the services and applications are already in the terminal device on packet-oriented Internet Set up protocols (IP). It is currently forecast that the growth in IP-based services will also be many times greater in the future than in other technologies, which is why it seems likely that all services will be largely migrated to transport over IP-based networks.

In contrast to ATM networks, IP networks initially work without a connection and only offer a 'best effort' service, with which even with generous dimensioning of the networks, hardly any predictions and certainly no guarantees for an achievable QoS are possible. In addition, the following approaches are known:

a) Use an ATM network as the core network. Edge devices implement the IP data flows in ATM connections of suitable service classes and the transport takes place in corresponding connections in the ATM network. Scalability, complexity and effort for setup and operation (see above, ATM technology) are problematic. This solution initially helps in the core. With additional The same disadvantages apply when using Access. The following solution is an alternative in Access.

b) Use a signaling protocol and establish connections with reserved bandwidths over the IP network (Integrated Services, RSVP). Basically goes end-to-end (from end device to end device) or on sections. Can be used per communication flow or (in the core) also for aggregated flows. Disadvantages: cumbersome, complex, does not scale (control effort), efficiency problems, i.e. Disadvantages similar to ATM technology.

c) MPLS: This approach is based on ATM technology. Paths (connections) are set up in the network via which the traffic of individual (generally aggregated) flows is routed in a targeted manner. Often suggested for QoS in conjunction with RSVP and DiffServ. Can also be implemented on the basis of ATM transport. Falls back into the complexity of connection-oriented mechanisms with all associated problems (from bandwidth control to monitoring the presence of the connection). In connection with DiffServ, it should alleviate the problems mentioned above (targeted traffic control via paths). Complexity and problems as with ATM technology.

'Differentiated Services' (e.g. IETF RFC 2475). The data packets are classified and marked in an edge device on the basis of their belonging to certain services, applications or communication relationships, etc. In addition, (flow-related) access control and monitoring can or should be carried out (e.g. for the availability of resources and compliance with the bandwidth and QoS characteristics registered). The packets then follow the route through the network specified by their packet header information (eg destination address) and the routing protocols, whereby they are treated (eg prioritized) in each node according to their marking with a corresponding 'per-hop behavior'. The 'DiffServ' approach allows the freedom of 'per-hop behavior' within a 'single routing domain', eg the (sub) network of an operator, but requires complete 'edge' treatment between such domains (subnets ). The DiffServ approach cannot prevent temporary and / or local bottlenecks, as there is usually no consideration or coordination with the routes specified by the routing protocols. Usually packets with the same. From the moment they meet at a node, they follow the same set route. This can cause considerable unbalanced loads and bottlenecks in the networks with correspondingly large (queuing) delays up to packet loss. result. The engineering of the networks and routes is, however, a complex task, with the aspect of reliability and availability (eg U routes in the event of a fault) making it more difficult.

In principle, almost all combinations between these approaches are also conceivable and for the most part have already been discussed. However, all these approaches have in common that they basically work on the basis of paths (or at least predefined routes in the case of pure DiffServ) and (with the exception of DiffServ) bandwidths reserved along the paths and possibly other resources. Associated with this is usually a very large administrative effort for the preparation and (static) establishment of paths and routes in the network or a correspondingly high control effort for the dynamic selection and switching of the routes. In addition, storage facilities for holding path- and connection-specific information must be provided in each network node, which can be lost in the event of an error or have to be reconfigured in other ways. Even with the pure DiffServ teaching, the traffic follows the routes predetermined by the routing protocols, which must therefore be dimensioned and monitored very carefully. As a rule, however, neither all fluctuations in traffic volume nor all reactions of the routing protocols to any events in the network can be predicted exactly.

The object of the invention is to demonstrate a simple, pragmatic and cost-effective approach of how various types of services can be provided reliably, efficiently and in compliance with their specific QoS requirements in IP-based networks.

The object is solved by claims 1, 2 and 3. Advantageous further developments of the subject of the application are specified in the subclaims. The subject of the application is explained in more detail below as an exemplary embodiment to the extent necessary for understanding with reference to figures. 1 shows a packet-oriented, connectionless communication network in accordance with the application, which is formed exclusively with edge nodes, FIG. 2 shows an embodiment of a network in accordance with the application

Edge node and central (core) node, FIG. 3 shows a possible topology of a real network.

In the figures, the same designations denote the same elements.

The figures show different embodiments according to the invention of connectionlessly operated, packet-oriented communication networks which have a plurality of nodes, between which a number of possible routes via connection sections

(left) are formed. In such a communication network, the edge nodes are BNI with regard to a supplied or a forwarded data stream as access edge nodes

(Boundary Node Ingress) and / or as an output edge node ESD

(Boundary Node Egress) operable.

3 shows an interconnection of a data stream belonging to a connection via different paths from an access edge node R to an output edge node S.

The invention is based on the following considerations:

a) Quality of service (QoS) is a relative term. Even if information is transmitted using circuit-switched technology, data loss cannot be ruled out (e.g. due to interference (-> bit errors) or frame slippage). Such weaknesses are either tolerable (e.g. in digital telephony) or they are intercepted by appropriate security measures on the same (e.g. by redundancy) or higher layers (e.g. by repetitions) (data technology). Ultimately, the decisive factor is the (subjective) quality perception of the recipient of the information. Real-time, interactive communication involving people, for example, always takes place via their (analogue) sense organs, which can certainly deal with incomplete information

(Otherwise (especially mobile) phone calls, film and television in today's form would certainly not be possible). For the interactive control of machines (e.g. remote control of robots), the requirements may be significantly higher, so that a detailed examination of each individual case may be necessary. In no case can the physical limits, e.g. in terms of distance-dependent transit times.

So QoS does not necessarily require an absolute guarantee (which does not exist anyway, not even using paths and reservations), but compliance with the corresponding specific requirements of the respective service from the perspective of the information recipient. In the case of packet-oriented transmission, this primarily concerns the type and scope of possible information losses, fixed and / or variable delays and the temporal consistency (sequence) of the information. ATM technology, for example, relies on switching nodes and transmission routes dimensioned according to the rules of statistics and the principle of connection-oriented transmission with appropriately reserved resources along the paths, with the correct distribution of resources along the paths using sophisticated and highly complex queuing and scheduling mechanisms to be ensured in the network nodes. Modern high-speed (data) networks work in 'Wire Speed' IP-based networks (Internet) initially only see data packets and treat them a priori all the same: the packet that is first received is also forwarded first, because there are not enough transmission resources available , the packets are initially saved (queuing, buffer) and if there is no more storage space, excess incoming packets are discarded ('best effort' principle). The network nodes in these networks, the so-called routers, are traditionally computers which have implemented the complete functionality of analyzing and forwarding the data packets in software programs. Accordingly, until recently these networks were comparatively slow, but with the help of appropriately dimensioned buffer memories and suitable data protection mechanisms in the higher protocol layers (TCP), a sufficiently reliable and usable transmission of time-uncritical information could be achieved (often with a long delay, but) become.

Technological advances made it possible to implement the elementary router functions in hardware (A ~ SICs, FPGAs) and thus opened the way for fast, and thus also real-time, forwarding of data packets on connecting lines. All that remains as a delaying element is the inevitable buffering of conflicts with the simultaneous arrival of several data packets that are routed to the same output. However, these delays become less and less significant with increasing bandwidth (or better: speed) of the connecting lines between the routers (see enclosed working paper HSR-NETZ .DOC, Chapter 2). This applies in particular when different traffic flows are distinguished by appropriate marking and can be treated differently in queuing and scheduling (DiffServ, 'prioritization'). It remains to be solved, such as the aggregation of traffic flows on the routes in the network, which is why even with careful control of the traffic flows at the network entrances deeper in the network, unbalanced loads that may affect the QoS cannot be predicted and therefore cannot be avoided, or the high effort and the resulting long period of time for reconfiguring the routes in the event of a fault, which can considerably impair the availability of the network and services for the users.

A communication network according to the invention comprises the following properties and functionalities (basic idea):

- it works package-oriented and connectionless,

- it offers a variety of input and output ports,

it consists of a plurality of network nodes which are meshed with one another in such a way that (as a rule) a plurality of paths exist between different input and output ports,

It contains mechanisms that, taking into account the respective destination (output port) of the data packets, aim at an even distribution of the traffic load in the network at all times (i.e. at every decision point in the network if possible).

Other features can be added in different designs and compositions (but not all combinations are reasonable and possible). Some features, along with some possible alternative solutions, are detailed below: The goal of traffic distribution is to achieve the most even possible distribution of the traffic load in the network. It can take place in different granularity, for example on the basis of aggregated traffic flows, per individual traffic flow or on the basis of individual data packets. However, the finer its granularity, the more efficient the distribution will be. The distribution decision should be made "ad hoc" and automatically in every network node. The decision criterion is information that is delivered with the data packets, for example a source / destination address combination, possibly together with further information that is used, for example, to assign a specific traffic flow. In the case of a distribution based on traffic flows, however, all data packets belonging to the same traffic flow will also have to take the same route through the network, which in the case of insufficient statistical mass or very uneven traffic flows (different bandwidths) unbalanced loads up to overload cannot reliably prevent individual network sections.

An example to illustrate the basic principle for traffic distribution in a regular network (theory) meshed over different levels (network levels) is shown in FIG. 2. FIG. 3 shows one (of many possible) more specific embodiment (s). In the case of such a predetermined topology with regular linking, the route information for the traffic distribution and the resulting 'branching pattern' can be more or less fixedly preset in the network nodes. In a real, grown data network (practice, e.g. see Fig. 3), however, meshing will normally not be regular and rather incomplete. There will also always be changes to the network configuration during operation. For this reason, a flexible update of the possible routes and branching patterns (as required or regularly) must be possible and / or the node must be able to derive the new branching patterns from a changed route information. Corresponding protocols from the Internet are to be used as mechanisms for the distribution of the route information. net environment (routing protocols, e.g. OSPF, BGP) or variants / further developments derived from it. Of course, this information can also be specified via a network control (or whatever) or a network management system.

With the branching patterns, further criteria such as, for example, different bandwidths, different distances to the destination, route costs, etc. can be taken into account in the algorithms for the specific route selection. For example, In the case of packet-wise distribution between an STM-4 and an STM-1 link, only every 5th packet can be placed on the STM-1 link by appropriate weighting. The load can be distributed even more finely if the individual package lengths are also taken into account (but complexity !?). A weighting of the links according to appropriate criteria is also advantageous in order to avoid the formation of loops in a complex meshed network or to e.g. Limit packet delay fluctuations. Different packet delays in different ways can lead to a change in the packet order, which must then be restored at the network output (resequencing) if the application requires it.

The traffic distribution itself creates a QoS that is largely balanced for all services and applications. However, without further measures, this still only has the 'best effort' character and the services and applications will suffer more or less noticeably as the load increases, depending on their properties and requirements. This behavior can be significantly improved if the total traffic load in the network is limited according to the actual network capacity. In addition, the bandwidth of the individual network accesses must be considered both on the input and output side and taken into account both separately and in the overall picture. Based on the statistical traffic characteristics of the various services and Applications and based on the topology of the network and the capacity and performance of the network nodes and connecting lines, assuming certain behavior patterns of the participants or certain resulting traffic characteristics at the network accesses, the network can be dimensioned in such a way that certain limit values of the QoS-determining ones determine these boundary conditions Factors (eg packet loss, delay, delay fluctuation) can only be exceeded with a well-defined, sufficiently low statistical probability (network dimensioning). Conversely, of course, the traffic in a given network can also be limited so that the corresponding boundary conditions are met. To do this, however, all communication relationships and data streams in the network must be appropriately parameterized and individually recorded each time they occur and, depending on the current load situation in the network, approved or rejected (admission control). For most of the traditional and future Internet applications, which are designed for a 'rough' best effort environment, this is neither practical nor economical. Nor is it economically justifiable to oversize the networks to meet the requirements of the more sensitive (higher quality) services.

A differentiated and adapted to the requirements of the respective service QoS can be done by differentiating and dividing into different traffic classes, which can be treated (prioritized) accordingly. The number of traffic classes is at least two (but there may be more), although strict prioritization is preferred for the treatment in the network nodes (ie at the queuing points). Alternative procedures that also guarantee low-priority traffic classes under all circumstances may have to do this under high loads at the expense of higher-priority traffic (eg weighted fair queuing, WFQ). In a communication network of the type described above, the prioritization feature can be used advantageously as follows:

All data streams are divided into corresponding priority classes according to their requirements. The lowest class is only taken into account when dimensioning the network (within the scope of the expected total traffic volume) and is generally only treated according to the best effort principle. For all communication relationships or data streams in higher priority classes, an admissibility check (admission control) is carried out at the network input (in the input direction) and at the network output (in the output direction). To do this, these data streams must be registered and evaluated at these two points with appropriate parameters (e.g. average data and / or packet rate, peak rate, etc.). The decisions at the entrance and at the exit are independent of each other and the data stream is only allowed if both decisions are positive. As a decision criterion, e.g. serve a threshold value, which is determined depending on the capacity of the port, the total network capacity, the desired quality with regard to possible packet delays and losses, etc., the respective priority class and possibly other criteria. It is also conceivable that there are several threshold values for each class on the basis of different evaluation parameters, all of which must be complied with individually or with corresponding interdependencies.

The admissibility check is intended on the one hand to limit the total traffic volume of a certain priority class in the network, and on the other hand to limit the associated traffic volume at each individual input and output port. Due to the even distribution of traffic on the network (ideally on a packet basis) and the corresponding preferential treatment (ideally strict priority, ie complete suppression of low-priority traffic if necessary), this traffic is always sufficient with correctly set thresholds Find resources (free link capacity, buffer memory) in the network in order to be able to meet both the delay and loss limit values of its quality requirements. The network can certainly be fully utilized and operated economically, because all bandwidth not used by high-priority traffic can be used by low-priority traffic at any time.

Adherence to the registered traffic parameters of the individual data streams must of course be monitored, because within the framework of the traffic distribution, even a single data stream, which slips properly, can significantly disrupt all traffic in the entire network. However, the monitoring function (traffic enforcement, policing) can, if necessary, be designed to be relatively insensitive (and inexpensive) because a random, short-term, slight exceedance can be appropriately averaged out by the traffic distribution.

The monitoring function is usefully applied to the individual data streams as they were registered. It would also be conceivable to aggregate per port of some kind, whereby only the total limit is checked (but with the unsightly consequence that if an aggregate is exceeded, it would have to be 'randomly' struck and possibly across all the data streams it contains). In principle, any and of course all relevant mechanisms (e.g. leaky bucket) can be used and the same applies to the reaction options (discarding packets, marking packets, switching off / blocking the data stream, ...). Marking may also consist in converting the packets violating the agreement (or rather the entire associated data stream) to a lower class or 'best effort'.

The principle of traffic distribution (especially if this is done at the packet level) can also be used very advantageously to improve the reliability and availability of the network and services. To do this, it is sufficient for the network nodes to remove the associated link (s) from the branch fan when an error is detected (eg link failed, neighboring node failed) and to continue the distribution only via the remaining links. The decision can be made immediately and autonomously if the fault condition is recognized, and if the fault is recognized as soon as the information is available. If the network is adequately dimensioned, in the worst case, such a reaction will lead to a somewhat more displacement of best effort traffic, but will not result in any loss of quality in high-priority traffic.

Further developments of the invention are as follows:

A very interesting variant of the basic idea arises if the traffic distribution method is only applied to the higher priority traffic class (es). Best Effort traffic then selects its routes according to the principles of the Internet that are common today. The higher priority traffic is distributed evenly in the network and fills it up from below, while the best effort traffic 'floats' on it and is increasingly crowded out as the 'tide' rises. An advantage could include is that the QoS solution can be added to existing networks, while the existing mechanisms remain unchanged. In this case, however, best effort traffic would not be able to benefit from the basic advantages of traffic distribution.

The principle presented is also in a cell-based network, e.g. an ATM network, applicable.

The reliability of the network can be improved by self-monitoring mechanisms in the router, especially as part of the distribution development process can be further improved.

In order to increase reliability over the entire network, a type of quick feedback mechanism can also be used between the routers, which e.g. enables problems to be redistributed 'upstream' early in the event of problems occurring somewhere 'downstream'.

The admission control can be designed in such a way that it automatically offers the user the next lower class in the case of 'overbooking' a high-priority traffic class.

The resequencing function is not implemented in most of today's TCP applications. It is therefore generally, e.g. provided as a standard function at the mains outlet.

This has several other advantages:

- Even traffic distribution enables optimal use of resources with the highest quality and thus the most economical dimensioning.

- A connectionless operated network does not require any control services for establishing and clearing the connection, no search for a route, no reconfiguration of routes, no recovery of paths in the event of an error, ... - is much easier to control and operate more economically (hardly any administrative intervention is necessary, ' self-organizing ').

- Strict priority gives minimal delay and minimal loss.

The invention is explained in more detail below on the basis of exemplary embodiments which are illustrated in the figures. A preferred embodiment of a 'sample solution' includes the following basic functionalities:

- A connection-free, packet-oriented communication network (according to the minimum requirements according to section 3),

- that distinguishes (at least) two traffic classes, one of which is treated as pure best effort traffic, while the other (s) (and ideally among themselves) is (are) strictly prioritized, in which the network nodes individually and autonomously control the traffic with the aim of evenly distributing the traffic load in packets according to certain rules on all (or at least a plurality of) routes (n) towards their destination (network exit),

- in which the network nodes exchange / disseminate the information about available routes with appropriate protocols,

- in which the network nodes adapt their traffic distribution pattern immediately and autonomously in the event of a fault,

- that performs an admission control on the basis of certain traffic parameters for the data streams of the higher traffic class (s) at each input and each output (which, for example, further traffic of this traffic class (n) from a total load of x% (x%, (x + d)%, (x + nd)%) of the port capacity no longer allows (other rules are also conceivable)),

- which only accepts a data stream of the (higher) traffic class if both admission controls (at the input port and at the output port - independently of one another!) have made a positive decision that monitors the registered traffic parameters of the data streams of the higher traffic class (s) at each input and intervenes with suitable measures if necessary,

- That offers a resequencing function at each output for optional use by (all) data streams.

It should be emphasized that the description of the components of the uniform communication network relevant to the invention is not to be understood as restrictive. For a relevant specialist it is clear that the terms used are to be understood functionally and not physically. The components can thus also be implemented partially or completely in software and / or distributed over a number of physical devices.

Claims

claims

1. Access node for a connectionless operated, packet-oriented communication network in which

the traffic flows that can be fed to the node can be distinguished according to a high-ranking priority class and at least one lower-priority class,

- The node access means are assigned in such a way that the data packets of a traffic stream with a high priority class are only permitted to be switched through by the communication network if both the requirements for the data rate of the traffic stream are met by the node as an access edge node and the node receives a signal that allows the forwarding of the traffic flow from the communication network from a remote outgoing node.

2. Node for a connection-free, packet-oriented communication network in which

^' - the node can be operated as an access edge node,

the node can be operated as an output edge node,

- The node access means are assigned such that the data packets of a first traffic stream with a high priority class are only permitted to be switched through by the communication network if the requirements for the data rate of the traffic stream are met by the node as an access edge node and

- Means are assigned to the node as an output edge node in such a way that for the data packets of a second traffic stream with a high priority class to be passed on from the communication network, the signal is transmitted in the direction of a remote access node of the second traffic stream when the node ten has sufficient forwarding capacity.

3. Connection-free, packet-oriented communication network with a plurality of edge nodes and a plurality of connection sections (left) between the edge nodes, in which at least one edge node can be operated as an access edge node BNI (Boundary Node Ingress), at least one edge node as Output boundary node BNE (boundary node egress) can be operated,

- The traffic flows that can be fed to the edge nodes can be distinguished according to a high-ranking priority class and at least one lower-priority class and

- An access edge node BNI access means are assigned such that the data packets of a traffic stream with a high priority class are only permitted to be switched through to an output edge node GNI if the requirements for the data rate of the traffic flow both from the access edge node BNI and can also be fulfilled by the output boundary node GNI.

4. Communication network according to one of the preceding claims, characterized in that at least one central node (core node) is connected via connecting sections between the edge nodes.

5. Communication network according to one of the preceding claims, characterized in that a node has distribution means in such a way that the data packets of a traffic flow are distributed over different connection sections.

6. Communication network according to claim 5, characterized in that the distribution means are designed such that the distribution of data packets to data packets takes place.

7. Communication network according to one of the preceding claims, characterized in that monitoring means are assigned to an access edge node in such a way that the data packets of a traffic stream which exceed the agreed traffic parameters, in particular the data transmission rate, are held.