Method and device for the simulation of network behavior for real-time dimensioning

The invention relates to a method and a simulation device for the study and to network planning, allowing the need sizing request network. It applies to all types of networks: mobile packet transmissions continuous, optical networks, eg multiplex type division length wave, with or without connection on electronic networks, etc.

The simulation can take into account under relatively timescales important (eg daily cycles or beyond) and a range of events dynamic (traffic, protocol, etc.) associated with control plans and data.

Modeling and simulation of telecommunications systems, to quote a where such networks are exploited, are attracting increasing interest due to their complexity (large capacity, traffic variability, cost of simulation in a real network). The aim is to provide the performance, compare solutions before they are actually implemented and, more generally, reduce costs and improve the optimization and design of networks.

It is also observed that with the convergence of telecommunication networks and data distribution, it is necessary to analyze the high degree of Model complexity to evaluate the different options and determine and the best guidance.

For this kind of task, two approaches are used in the state of the art, say respectively static and per package.

The static approach is primarily intended to networks that operate circuit mode, that is to say with the continuous data, and uses matrices static simulation. In this case, any flow between two points are considered as invariable entities in the network. These tools do not allow capture all network dynamics, protocols, and the variance traffic, which is predominant in our days.

Specifically, static tools can not understand the realistic constraints for example Internet Protocol (IP) or other network packets over optical or electronic resources, including stress due to:<ul><li>variation in traffic in time and in space,</li><li>dynamic mechanisms: resource creation and reconfiguration demand network, traffic engineering (load balancing, control congestions, traffic partitioning), protection or restoration of the network ...) and</li><li>packet transport networks (peripheral units of behavior).</li></ul>

In general, static tools are based on long optimization methods calculating, without real time constraints. These gaps are felt in results, which lack consistency and reliability compared to reality.

The second approach is called "package approach". Examples of such approaches are given in the documents " <i>Optical packet switching with mutiple path routing "</i> Gerardo Castanon, Lubo Tancevski and Lakshman Tamil, <i>"Modeling and Simulating network communication: a hands-on approach using OPNET "</i> I. Katzela.

With the approach package can theoretically informed analyzes are much finer and, in absolute terms, would capture all parameters a network. However, it poses a problem of time of calculation due to the number crippling quickly the number of calculations necessary. by modeling package is including not possible for simulations involving networks said "Terabit" (Which manage more than a bit Tera (10¹² bits) per second) and long-term variations in the traffic (expressed in hours or days) as the absolute number of simulation calculations would be beyond the current possibilities. As an indication, a level simulation package for a network model of Tera 100 bit / s with developments on a scale 24 hours would require the calculation of more than 10¹⁶ events.

The dimensioning of the node components produced from an analysis by packet is generally influenced by considerations taken only on a small radius. However, the quantities of resources (and therefore the sizing and costs) depend mainly large variations.

The packet analysis can be applied to models operating in dependency Mutual with networks based on circuits (behavioral considerations ring), and is limited to networks for optical packets.

Moreover, too fine-grained package level does not allow the study of network in a realistic time scale. (For the record, the granularity is a Quantification of switched basic information in a network, and reports to the type of network: it can for example correspond to the wavelet length for a network multiplexed in wavelength to a fiber in the case of a network of fibers, a packet in a packet switched network, etc. The granularity can be spectral, spatial, temporal or otherwise.)

Thus, in the document " <i>Modeling and Simulating communication networks: year hands-on approach using OPNET "</i> I. Katzela, in Figure 4-6, it is seen clearly that the simulation was performed on a 30-second interval.

Therefore, conventional network operators analyze a network on bases static and then perform a combinatorial optimization, for example with tools linear conversion or the like. When performing packet analysis, they is at best possible to examine some nodes, but never in his network together, and not with any protocol could be used. At best, some just manage to simulate the control plan, that is to say the signaling between the network nodes.

so it is impossible to take into account any new protocols or mechanisms for data transport, for example future optical packets on an optical network.

In both approaches, it is necessary not only to describe the topology the network before the simulation, but also to determine the resources in each of the elements or link capacity and equipment nodes.

Simulation according to conventional techniques can then only account for states of the network during the simulation. Only after the complete simulation and analysis rendered accounts that can be whether the design and network planning are adapted to the conditions of the simulation scenario.

In view of the foregoing, the invention provides, in a first aspect, a method for simulating the behavior of at least one network comprising a set of network elements, characterized in that it consists:<ul><li>to develop and introduce into the network at least one set for flow to simulate a constraint on at least one of the network elements.</li><li>detecting the network behavior in response to at least one constraint imposed by this or these flows;</li></ul>

Preferably, the flow develops on the basis of a model of the variation in time of the traffic intensity in the vis-a-vis network of the or each element which flow is intended in the context of the simulation.

One can then develop the flow in the form of a set of stream, of which each member corresponds to a traffic on a portion of elementary path linking a pair respective nodes specified network.

In the preferred embodiment, the method comprises a step of achieving flow matrix, in which each element expresses a variation in the time of the flux intensity along a portion of respective path of the network, the stream being introduced into the network in accordance with this matrix.

Advantageously, printed on the flow stochastic variation.

The flow can express a traffic intensity variation on a scale of macroscopic time relative to the duration of its transit in the network.

This variation may relate to flow changes on a time scale macroscopic simulating several hours of actual use of the network, including daily operating cycle of the simulated network.

Preferably, creates a flow for intensity modulation scale Macroscopic at which imposes local stochastic flux variations on a microscopic time scale.

The stochastic variation of the flow can be established in accordance with a exponential distribution, preferably a Poisson distribution.

Preferably, characterizes the flow by at least one of the following parameters:<ul><li>an average flow,</li><li>the variance of the flow,</li><li>the Hurst parameter, and</li><li>a quality parameter, including the class of service required by the flow.</li></ul>

According to a preferred embodiment, the method further comprises the steps of:<ul><li>identifying a potential weakness of at least one element facing the stress, and</li><li>Modified optionally at least one element reflecting said weakness to allow it to accommodate the constraint having revealed including by enhancing the design of a performance characteristic of item.</li></ul>

It is thus possible to design a request from a network, or set of networks.

These steps of detecting, identifying and modification may be executed concurrently with the introduction of flows in the network.

Typically, the network element is a node and / or link.

The method may be repeated at least once the introduction of flows in the network to simulate each iteration a statistical variation of the flow, this variation Statistical being obtained in particular on the basis of the stochastic nature of the flow.

According to a second aspect, the invention relates to the method as set forth, put implemented to establish the sizing performance of a network initially virgin for which specified a topology of nodes and links, wherein which is introduced in the network vis-à-vis stream from which it must be dimensioned, and that the detection steps are performed, identification and modification of the aforementioned until the design conforms to the stream.

According to a third aspect, the invention relates to the method as set forth, put implemented to establish a new sizing performance of a network existing, characterized in that one introduces into the network vis-à-vis stream from which it must be dimensioned, and in that the detecting steps are carried out and, if appropriate, the aforementioned identification and modification until a updated design conforms to the stream.

According to a fourth aspect, the invention relates to the method as set forth, put implemented to establish a performance design of a network against a simulated failure, characterized in that a simulation of the network is carried out by modified failure are introduced into the network vis-à-vis which flow the modified network must be dimensioned, and in that said detection steps are carried out and, if appropriate, the aforementioned identification and modification until a design conforms to feed on this modified network.

According to a fifth aspect, the invention relates to the use of the method forth to simulate a data transmission network packet.

Preferably, the flow is developed with an intermediate granularity between the granularity of the package and the intrinsic network switching granularities.

According to sixth aspect, the invention relates to the use of the method forth to simulate a data transmission system circuit mode.

According to a seventh aspect, the invention relates to a simulation device behavior of at least one network having a plurality of network elements, characterized in that it comprises:<ul><li>Ways to develop and introduce into the network at least one stream configured for simulating a constraint on at least one of the elements network.</li><li>Means for detecting the behavior of the network in response to the least one constraint imposed by this or these flows;</li></ul>

Aspects of the invention discussed above in the context of the method according to a the first to sixth aspects may apply mutatis mutandis to this device and will not be repeated for brevity.

The invention and the benefits that emerge will emerge more clearly in the reading the detailed description of preferred embodiments, given only way of non-limiting example with reference to the accompanying drawings, of which:<ul><li>Figure 1 is a block diagram of the different functional blocks used in a flow simulation tool with sizing in the application in accordance with the invention;</li><li>2A is a diagram showing the nature of the flows and their stochastic emission to a blank network during operation of the tool of Figure 1;</li><li>2B is a curve showing the evolution over time of the distribution stochastic stream of intensity between two nodes of a network to be simulated by the tool of the Figure 1, with a macroscopic time scale corresponding to a cycle time simulated;</li><li>2C is a graph showing the evolution of the intensity of flux Stochastic 2B, but on a microscopic scale, on the order of time transit of a stream in the network, with variations that fluctuate randomly;</li><li>3 shows the network of Figure 2A after a sizing to the request made by the tool of Figure 1; and</li><li>4 shows a network similar to that of Figure 2A, on which occurs a fault simulation by using the tool of Figure 1.</li></ul>

The simulation tool and sizing 2 shown in Figure 1 comprises a set of hardware modules and / or software that functionally depend on central computing unit and management 4, which brings together the intelligence of the whole. access to the tool 2 by a user is performed by a user interface 6 to which are connected a control panel 8 and a keyboard associated with a mouse 10.

The central processing unit 4 controls, inter alia, three units which interact with one or several networks R1, R2 analyzed, namely:<ul><li>a stream output unit 12, which transmits via stream F data traffic simulation. This unit 12 is supplied by a database 14 containing simulation flow matrices, described below;</li><li>a network analysis unit 16 which collects data on the DF operation / network (s) simulated (s); and</li><li>a network modifying unit (s) 18, which transmits data DD network provisioning, especially to selectively increase the level of performance of network elements according to the DF operating data. These data are used, especially to achieve sizing to the request of a network during or after a simulation.</li></ul>

The flow F contained in the database 14 are produced by the CPU 4 according to criteria and setting made by a user via the keyboard 10 and the screen 8 of the interface 6, or possibly by a source such as a beam registration or an online connection (not shown). the base is noted that data 14 contains supplemental information in addition to flow F.

A first mode of operation of the tool 2 to perform a simulated or a set of networks with a possible dimensioning for the application, will be described by reference to the diagram of Figure 2. In the following, is employed for the sake of simplicity the term "network" generically, be it a single network or several, interconnected or not taken into account by the tool 2. In the example shown, the R network is the optical type, operating in multiplex mode by division wavelength (known by the English term of WDM for "wavelength division multiplexing "), which is a technology to use several wavelengths different in the same optical fiber. However, it is understood that the tool 2 can used for any other type of network.

Rather than simulating the propagation and manage each packet in the network, concept uses a new entity that is the flow. The simulation is then to put implement the traffic distribution modeling processes on the network average flow F. The flow F is an intermediate entity between the pack and the granularity intrinsic switching. A wide array R, virtually the entire spectrum of Traffic variation can be apprehended by the dynamic flow of creation.

A flow is defined by one or more characteristics, such as: distributions over time, or start dates and end dates, and distribution Space, or the distribution of flows in the network. One can also take into account the routing, reporting a traffic matrix analysis.

The phenomena that occur within the flow can be characterized several different ways. For example, a simple and effective approach this characterization is to assign the flow F:<ul><li>an average flow,</li><li>a variance, and</li><li>a qualitative parameter, in particular the class of service required by the stream question.</li></ul>

The service class may be of the type called "premium" to carry the voice, or type "best effort" (in English "best effort") to convey data.

The characterization by the mean and variance of flows best fits sporadic flow (also designated "bursty" in English terminology) specific data traffic.

The flow simulation can also be applied to types of said data self-similar ( "self-similar" in ter minology Anglo-Saxon). In this case, the flow characterization parameters are preferably the mean and / or the parameter called "Hurst", which measures the degree of correlation between arrivals packets.

In a circuit network type, the characterization of flow is much simple, since one can directly simulate physical parameters condition transmission, such as wavelength. We then assimilates flow wavelengths. The rate will be fixed, and only plays on the distribution spatial and temporal distribution of these flows, not on the internal characterization flow.

According to the time scale that is intended to simulate and duration of use F flow, they can be representative:<ul><li>or applying to the lowest level: it can be a traffic generated by a particular application (for a micro flow)</li><li>or assumptions that make it an aggregation of flow and micro flow. These last may come to a traffic system from a local network LAN-like (of English "local area network").</li></ul>

One can consider applying for different flow rates, different targets in terms of volume, with or without aggregation. Then, the difference will characterize and pack two types of information: how performance is obtained at the nodes and how they are then used. The latter will subject to data that will actually be used to establish the sizing the network following the simulation.

From scientific point of view, it is more difficult to accurately characterize a a micro flow aggregation in isolation, since the aggregation usually implies statistical multiplexing, capacity buffers, and other service disciplines nodes. This explains why the behavior a deterministic stream aggregation in a network is not envisaged in the state of the technique.

The simulation process used for flow F stochastic matrices 20 each element 22 represents the average traffic between two specific nodes N. The average traffic stream is specified in terms of intensity distribution on a scale of time, which can be long term, corresponding for example a daily cycle of 24 hours. Thus, the matrix 20 comprises, for each element 22 information which can be represented on a curve 24 intensity distribution from which is carried a random draw with a local distribution in time.

Figure 2B shows an example curve 24 intensity distribution for 22ij the matrix element 20 which, according to column formatting and rows thereof, regarding the flow Fij between the N designated nodes Ni and Nj (Figure 2A). The matrix 20 thus includes a number E² such elements to simulate a network comprising a number of nodes E.

The intensity of the flow, defined in units, is shown on the ordinate against time scale x-axis. Curve 24 shows more particularly the modulation, or the envelope variations of the flow intensity, its shape being smoothed over the period cycle (within 24 hours), which corresponds to the macroscopic scale.

However, the instantaneous value of the intensity of a stream is determined by a stochastic modeling. Thus, for a short period of the cycle, the intensity will vary randomly or pseudo-random manner within the constraints set by the modulation curve 24.

2C shows by way of illustration on similar axes to those of the 2B the local variations of the flow intensity in a VL interval of the curve 24, but on a microscopic scale (in this case 30 seconds) to cover intensity fluctuations over a period of the order of the duration of the stream in the network. We notes that this microscopic scale, changes may have Significant excursions.

Thus, it has different time frames contained by traffic modeling matrix depending on whether one considers the evolution ladders macroscopic or microscopic time.

More particularly, for each pair of nodes (for example Ni and Nj), the stochastic matrix 20 defines:<ul><li>the modulation of the traffic distribution in the day (for a cycle daily), which corresponds to an initial distribution on the macroscopic scale (Figure 2B); and</li><li>stochastic fluctuations quantified for short successive times, which corresponds to the microscopic scale.</li></ul>

The stochastic fluctuations can be produced by random nickname random according to an exponential distribution of the duration of the stream, in particular according a Poisson distribution (according to a Poisson distribution) or the like. For achieve these stochastic variations, the flux emitting unit 12 includes means random or pseudo random draw that perform successive draws in a sufficiently short intervals in time to simulate realistic variations. Each print results in an instantaneous and random variation, following the Poisson distribution, the intensity value of the flow indicated generally by the curve 24 in a macroscopic scale.

the so, the flow F of intensity curves do not provide elements of flow deterministic, but probabilities, according to Poisson curves, for example, thereby giving the flux F their stochastic nature.

Since working with stochastic F flow arrivals, Samples for determining the network strength are also of order statistics.

Each element 22 of the matrix 20 contains similar information govern stochastic flow arrivals of their respective pair of nodes.

It is noted that several sets of such information may be associated with each node, each associated with a type of flow to model between a pair of nodes. Thus, in the example of Figure 2A, the 22ij matrix element 20 is shown as having three curves 24, each similar to that of Figures 2B and 2C, each associated with a particular type of service.

One can thus simulate day / night effects on a large continental network, while respecting a set of realistic loads on the flow of arrivals, requests access, call processing "voice" or packets, etc.

This creates a basis for calculation which will be executed following assumptions Traffic intensity on a given day and following the on-court distributions term.

Other qualitative and quantitative information used to characterize F flows are schematically represented by boxes 26 in all pointing the database unit and the emission stream 14, 12 of Figure 2A. These Additional information may include:<ul><li>the nature of the types of traffic,</li><li>CoS concerned,</li><li>typical rates, the typical duration and other parameters laying down the rules random drawing to form stochastic flows, etc.</li></ul>

For each element 22 of the matrix 20, thus extract the stochastic flow corresponding to integrate it into the node N concerned the network R.

In the example, the network includes two types of nodes: <ul><li>network core nodes ( "core nodes" in English terminology) which do not directly communicate with routers, but only with other nodes; and</li><li>the device access node ( "edge nodes" in English terminology) which are access roads. These nodes, marked by white pellets in 28 2A are connected to routers, namely the type of LSR (the English term "Label switching router").</li></ul>

This is an IP router which can also work in MPLS mode (the English terminology "multi-protocol label switching"), which is the current way to use the Internet with connection oriented approaches. The LSR generate LSP (the English term "switch pass label") connection-oriented, like the Asynchronous Transfer Mode ATM says (the English terminology "asynchronous transfer mode ") or technique" frame relay "between two points. The concept of placing flow implemented allows to faithfully comply with network design, since this last used virtual connections between two points which can also be characterized. Thus, the approach of the invention naturally lends itself to the reality of today's networks.

In the case of Figure 2A, the set of routers 28 serve to receive the different streams and distribute them in a network routing algorithm.

Furthermore, it is possible to prior assumptions input Locations flow in the network when one does not use routers. operation is then performed by multiplexing and determination of the necessary bandwidth.

The evaluation of the bandwidth for the aggregation, in particular between a network electronic and optical network requires to carry out a mapping of flow in wavelengths. This implies an adaptation of the flow interface interfaces between the router and the node to which it communicates.

From these stochastic flow F produced by the stream output unit 12, analyzes the behavior of R network with DF data collected by the unit Analysis 16.

On arrival stochastic flows in the R network capacity are observed nodes (and possibly links L) to treat. The treatment capacity in question includes not only the "raw" capacity but, where appropriate, also the functions of like "conversion" or similar for optical networks.

If the flow F can not pass into the network, it enhances the originally node blocking through these DF sizing data modification unit 18.

This is done in a sizing to the request, the request being caused by the flow simulation.

All these stochastic arrivals of flows in the network will, on a representative sample, give rise to a need for dimensioning of the network passing traffic assumptions originally.

Due to the statistical nature of the samples flow provided in the network, can execute the process iteratively with samples which represent several cycles of the modeled time scale, for example 100 times a day. At each iteration of the process, the array 20 emit streams whose intensity distribution on the macroscopic scale will be the same (Figure 2B), but with different local stochastic variations on a microscopic scale (Figure 2C).

The iterations are continued until afford to get a degree sufficient confidence in the simulation. For each repeated simulation cycle, there will be a large number (for example of the order of one million) of prints. The level of confidence is based on the sample size and the number of cycles (the duration of the simulation).

Furthermore, the simulation time will be variable depending on the fineness of the granularity that is simulated (which may vary for example a micro flow at 10 kbit / s to aggregated traffic of several Mbit / s), because the granularity determines the amount of flux to simulate.

In a practical embodiment, it is possible, with the means of Current treatment to simulate tens of millions of sources of micro flow which are then aggregated and transported in the network.

The simulation of the F flow basis may include:<ul><li>model the traffic transmitted by an application (voice, video, file transfer, HTTP, ...);</li><li>model a micro-flow aggregation (output of a local LAN-like network (The English term "local area network), etc.);</li><li>be specified by a set of behavioral modeling parameters traffic (average rate of passage, sporadic, mathematical models, MMP, self-similar, CoS, VPN, ...); etc.</li></ul>

However, the behavior at the packet level remains implicit and is not simulated, which means that the number of events to be simulated is lower of several orders of magnitude with respect to a packet level modeling as used Conventional approaches.

The flow properties can be managed according to many different distributions (inward flow, flow duration, destination stream, update dynamic parameters with TCP (the English term "Transport Control Protocol) etc.

The simulation technique may be seen as incorporating the following steps:<sl><li>i) introducing a stream F stochastic simulation in the network;</li><li>ii) detecting network performance face this stochastic flows F; and</li><li>iii) enhancement potential weak or inadequate parts of the network to treat imposed flow.</li></sl>

These steps can be continued interactively, with step iii) raising automatically initiated as a function of step ii) of detecting, step i) introduction of streams that can occur independently and concurrently as a specific program,

The simulation can be made on an R network said virgin, that is to say with a strict minimum pre-established characteristics or the initial topology, characterized by a set of nodes N and links L between them. The capacities of nodes N and links are not specified for virgin network: it is only a model and network nodes with a set of limits, but without capacity.

The idea is to bring a stochastic flows F constraints in the network R to identify their needs and modify the boundaries.

In response to flow constraints, the respective functions and capabilities network will be updated on demand (through targeted enhancement of performance through DD sizing data). These enhancements may be considered potentially many parameters, such as the performance of nodes at package, service quality, priorities, etc. Thereby, data DD sizing are determined not only by data DF collected operation, but also according to external parameters, such as an integrated specification implicitly scalable load in the stream.

3 schematically illustrates the originally virgin R network 2A after sizing on demand by the above process. We notes that some of the nodes N and links L have undergone a level of enhancement performance, especially in terms of capacity, indicated by the arrows designated RN and RL respectively.

DD sizing data are compiled, according to a specific protocol, to indicate both: i) the location of specific network sizing (designation node (s) or link (s) individual (s)), ii) the characteristic concerned by the sizing (capacity, speed, number ports, etc.), and iii) the quantification of that feature (eg a percentage increase, a new value capacity, etc.). Sizing Data can also specify an addition or displacement of a node or a link using a predetermined signaling protocol.

Note that the invention is remarkable for its ability to manage such a network that this would be with protocols and traffic engineering or management techniques realistic, even though it is in the design phase.

Any dynamic event related to the management, traffic control, fault (which have an impact on the network architecture) etc. can be simulated by this approach.

One of the typical actions of the dynamic sizing to the ability selectively enhance the ability of nodes N. In this approach, it is possible taking into account different granularity enhancement, including the specifications current network load the manufacturer.

The approach according to the invention, based on flow analysis, constitutes a solution that can be qualified as an intermediary, to remain faithful to dynamic routing protocol and take into account elements dynamic speaker on the network, such as breakdowns, algorithms traffic engineering, flow control, etc.

Today, we seek to create networks of said type of "multi-granularity", where we integrate the various layers and the various stages in the network, with aggregation to create traffic that are switched using different techniques.

The process used according to the invention allows the application from the Construction of the network protocols that are currently often adopted a posteriori. For example, if is implemented in a network balancing technique fillers (known by the term "load balancing"), we have a protocol that function to split traffic to distribute across multiple paths. This technology can then be taken into account from the network dimensioning.

This is possible because the simulating entities that are relatively thin and on which real protocols can be applied.

It can be provided in addition to modify dynamically in function of the simulation, the structure of the network, in particular by adding or removing nodes and links. This requires interaction between the said mechanisms "Offline" and centralized, not directly involved in the simulation process (Mechanisms called "off line" in English terminology), and mechanisms enligne distributed operating in real time (called "on-line" according to the same terminology). In this case, can be achieved off-line analysis via a node adapted for this purpose, for example with an analysis function of the network state, and in particular the distribution of traffic. For this node, then we can change dynamic network topology, including establishing at least one additional link between two nodes.

Compared to conventional techniques that simulate the control plan network at the nodes, the invention also predicts the plan of attack with its impact in the design, within the same tool and the part of the same process.

In the embodiment with the sizing application, the tool intervenes in particular in two cases:<ul><li>in the case of a virgin network, as described above, where it performs a complete calculation of sizing on demand (since it allows to supply real-time to provide the means to different nodes), and</li><li>in the case of an already dimensioned network, where it uses only performance of different algorithms and protocols.</li></ul>

It is also possible to perform simulations looped, where the result of a dimensioned network during a previous simulation is again subject to simulation or with the same stochastic flows (loan random factor) or with new settings flow. This loop can be repeated a number arbitrary times until a network according to different possibilities flux.

This allows to observe the network to another larger time scale, for example on years while the flow matrix is ​​based on cycles daily.

In general, it has a traffic matrix that is dynamic, that represents changes over time on a scale of a day. But one operator often wants to know the evolution of the network over several years. When one has a network dimensioned on the basis of a given traffic matrix, one can take the hypothesis such that a given number matrix months later will evolved by a multiplicative factor. We leave then the result before and again applies the same principle of selective enhancement of the performance of nodes, but starting from a given position which is by a blank grid, being that which results of a first simulation.

Such a possibility is open that the outputs can be inputs, Network level: nodes, protocols used, links, and physical parameters of the case applicable.

According to an optional aspect, the tool is also able to simulate failures network dynamically. Next algorithms of protection and restoration, it creates random or exhaustive breakdowns occur in that network Additional capacity requirements in the nodes and links on which traffic will be re-routed. This is taken into account in the simulation phase. In dynamic simulation, the additional resources required will be determined by restoration scheme algorithms or protection and taken in time real.

For example, Figure 4 illustrates a simulated failure on a link between two core network nodes. Routing on the network R imposed in response an overload on shedding links that connect these two nodes. The analysis is then whether these links and nodes involved can ensure this overload with the simulated flow.

You can also use the tool to make 2 following comparative studies several approaches, with protection schemes applied differently in different networks.

Technical comparisons can also be obtained from a packet and a circuit mode, for identical topologies and traffic, allowing to reach Cost comparisons taking into account the unit costs of components used.

Thus the tool can be used for scientific studies (analysis of new nodes, new types of nodes, or available features, etc.), or yet as network planning support tool for an operator.

The invention is applicable to any transport stream network, the term flows being taken its broadest sense: it therefore covers not only data transport computer and electronics, but also the distribution of energy or services (Gas, electricity, telephone), or material entities, transport networks vehicles (rail, road, sea, air), monetary flows in a network macro or microeconomic (stock exchanges, transactions between banks, shops, etc.), flow of parts or tasks in industry, etc.

The principle of the invention applies for both node type circuit, e.g. which switch wavelengths in the case of an optical network), or packet switches.

In the case of a circuit mode, it is relatively simple to whether a flow Arriving on a node can be switched or not: to take the example of a network optical or there is a port adapted to the wavelength of the incoming stream, wherein If this stream will be handled automatically or the node does not have such a port and the stream will refuse.

In the case of a packet switch, electronic or optical, the situation is more complicated. Indeed, when a stream reaches node, it must Knowing the equivalent bandwidth of the incoming flow, the more it flows before before determining whether the incoming flow can pass. This results from the fact that operates statistical multiplexing, the contention, considerations of performance at the packet level, etc. that do not exist in circuit. these aspects are taken into account by a simulation tool node characteristics as such. The information required on the node, to increase its performance can be obtained analytically, by result tables, by specific simulations, and with various degrees of following approximation performance, which techniques are known in themselves.

The tool according to the invention can accept input this information, some its origin. Preferably, this information comes from an analytical tool the characteristics of the network, if available.

The invention is particularly well suited to applications oriented connection. One can consider dynamically mechanisms congestion control, which change in real time the parameters of flow compared to states observed in the network. This will be modeled by the behavior stream level, without descending to the packet level.

In the following summary is presented so some of the advantages and features of the tool 2 of the invention.

Added value for the development of a network strategy:<ul><li>relevant results for comparison and orientation of architectures systems and networks; </li><li>control and management of evaluation protocols for scenarios existing or new;</li><li>effective and original methods of design and planning can apply to models in the short term or the long term.</li></ul>

Solution for typical studies of complex network creation:<ul><li>taking account of the dynamic aspects (traffic, real-time mechanisms, ...);</li><li>taking account of the scales (Terabit network);</li><li>incorporating a large set of constraints (physical, of conversion, protection, ...);</li><li>accepting any Internet business model.</li></ul>

Simulator dynamic flow:<ul><li>represents the distribution of traffic in a network by means of flow;</li><li>a stream is an intermediate entity between a package and the granularity of Intrinsic switching;</li><li>to model the traffic transmitted by an application (voice, video, file transfer, HTTP, ...);</li><li>to model a micro-flow aggregation (LAN Release, ...);</li><li>can be specified by a set of parameters modeling traffic behavior (average passing rate, sporadic, mathematical models, MMPP, self-similar, ...), CoS, VPN, ...</li><li>behavior at the packet level remains implicit and is not simulated, which implies a reduction of several orders of magnitude in the number events simulating;</li><li>the properties of flows can be managed in many distributions (flow arrivals, flow duration, destination flows, making Dynamic Update settings with TCP, ...).</li></ul>

Stresses applied to a network model by applying the Stochastic flow and dynamic event:<ul><li>sizing demand (DAD) of the capacitance (and optionally configuration) of the network and its resources;</li><li>successive streams are processed in the network (routing, distribution, ...);</li><li>the corresponding required resources are enhanced compared to services, protocols or other design constraint; </li><li>any optimization can be added at this stage: an optimization time-consuming (topology optimization, ...), or constraints anticipation of future exploitation technique known as network of "Score and validation" (the English term "point and click" (VPN ...);</li><li>real-time behavior of the "life" of the network (s):</li></ul><ul><li>use based on existing resources;</li><li>Application scenarios for survival and traffic engineering mechanisms (Dynamic supply routing "intelligent" dynamic setting, load balancing, partitioning traffic, congestion management, signaling protocols, ...);</li><li>preliminary assessment of performance at the packet;</li><li>dynamic impacts visible to the edges of the network, with performance edges for evaluating resources equivalent;</li><li>ability to enhance network resources to "exact" needs:</li></ul><ul><li>making network status update with new flows;</li><li>applicable to a wide range of algorithms and protocols for routing and procurement;</li><li>database of the state of nodes;</li><li>enhancement of the size and functionality of the nodes relative constraints in terms of service, enhancement systems (Switching and transmission), performance evaluation required for OPS routers like / OBSTout type of switch or router packet / frame;</li><li>be taken into account any dynamic events related to the "life" of network: traffic engineering, breakdowns, ...;</li></ul>

Potential of the dynamic flow simulation:<ul><li>comparisons (cost, performance, ...) network solutions;</li><li>systems (packet vs. circuit) multigranularité, ...);</li><li>architecture (vs. pair overlays, topologies, ...);</li><li>engineering protocols and traffic;</li><li>potential transparent assessment in the network;</li><li>conversion needs assessment;</li><li>assessment restore performance;</li><li>Comparison of survival strategies (amount of resources to be added);</li><li>determination of the gain with dynamic supply at various grain sizes; </li><li>performance of congestion control mechanisms in network Terabit kind;</li><li>compatibility with future developments (elements of independence model and methods).</li></ul>

It will be appreciated from the foregoing that the invention enables many various embodiments and variants.