DE10104926A1 - Parallel simulation of mobile radio network using multiprocessing system, involves programming network elements such that each network element repeats interaction cycle consecutively during simulation - Google Patents

Abstract

The individual network elements of mobile radio network are programmed such that each network element repeats the interaction cycle consecutively during simulation. The transmission channel is programmed such that it switches between transmission and reception states based on information transmission/reception request received from network elements. The switching between states is based on a comparison of known network elements and number of preceding transmitted/received data from the current into the next state.

Description

The invention relates to a method for parallel simulation of mobile radio networks according to the Preamble of Claim 1. The method allows parallel simulations on commercial ones Computers and is able to perform parallel simulations using discrete events in single Apply partitions.

The rapidly increasing requirements for a higher and more diverse traffic load in Mobile networks require network operators and system designers to optimize existing networks and to develop completely new solutions for mobile messaging systems.

Due to the complexity and the diverse inherent dynamics in a mobile network However, the design and optimization of these networks is a problem that is not analytical can be solved.

This applies in particular to future mobile radio networks. Characteristics of future Mobile networks can change its topology, the allocation of its resources to individuals outstanding services and thereby the overlapping situation of radio signals in the Change the transmission channel in a very dynamic way.

The quality and capacity of mobile radio networks are estimated based on one Analysis of these mutual superimpositions of radio signals in the transmission channel. The Radio signals are sent and received by network elements. Use these radio signals same physical resource, characterized by frequency and time, so they can overlay.

A simulative analysis of these overlays of radio signals is the goal more dynamic Simulations of the radio network level of mobile radio networks. Based on this, the Quality and capacity of a cellular network can be estimated.

Dynamic simulations of the radio network level of mobile radio networks are therefore able to do all capacity and quality relevant aspects of a mobile network in its complexity and that to investigate the dynamic course of time.  

A simulation is an executable computer program that the system to be examined modeled in an abstractness sufficient for the purpose of the investigation.

The generally recognized state of the art for mastering large computer programs is the object-oriented principle. Object-oriented programs consist of objects, which communicate with each other.

For the object-oriented simulation of the radio network level of mobile radio networks, J. Deißner, G. P. Fettweis, J. Fischer, D. Hunold, J. Voigt, R. Lehnert, M. Schweigel and J. Wagner: "Object-Oriented Modeling of a Generic Mobile Radio System for Dynamic Simulation," in Summer Computer Simulation Conference / Symposium on Performance Evaluation of Computer and Telecommunication Systems (SCSC / SPECTS '99), (Chicago, IL, USA), pp. 240-247, 11.-15. July 1999 the object-oriented model of the radio network level of a general mobile radio network presented.

The following were analyzed as important components of a mobile network:

• 1. The network element.
• 2. The transmission channel.
• 3. The network element activation.

These important components of a mobile radio network are each referred to as an object in modeled a computer program. These objects communicate using Network element external events. Communicate objects within a network element by means of events within the network element.

The network element component generally makes several independent of each other Give objects. The independence of the various network elements modeling objects is justified by the fact that the network element internal interactions within a Network element completely independent of the interactions within another network element Are network element.

Controlling the flow of dynamic simulations is not a task of the Objects, which are based on the in J. Deißner, G. P. Fettweis, J. Fischer, D. Hunold, J. Voigt, R. Lehnert, M. Schweigel and J. Wagner: "Object-Oriented Modeling of a Generic Mobile Radio System for Dynamic Simulation, "in Summer Computer Simulation Conference / Symposium on Performance Evaluation of Computer and Telecommunication  Systems (SCSC / SPECTS '99), (Chicago, IL, USA), pp. 240-247, 11.-15. July 1999 Object-oriented analysis of the system to be examined take part in the simulation. Therefore Every simulation must consist of (at least) two main parts: the simulation control, which implements one or more simulation methods and a part which the Functionality of the system to be examined implemented.

Known object-oriented simulators for mobile radio networks generally work with one Simulation control using discrete events. Communication between objects is treated here by means of events. Events actually take place between the objects information to be exchanged a time stamp, based on the value of which they are in an order can be ordered.

Implementations of this simulation method either work sequentially and are slow or they work in parallel and are then very complex and often require Supercomputer as a computing platform.

In the case of simulations using discrete events according to FIG. 1, it is also disadvantageous that they require a central event memory, which brings all the events occurring in the simulation into a global order relationship.

As a further feature, programs working in parallel must be partitioned to all Distribute calculations during the simulation to the available number of processors can. The goal of partitioning is a balanced computing load on all other Processors participating in calculations (the calculations in a single partition are run on their own processor) and a low one Inter-partition communication traffic.

Known partitioning strategies for simulation programs of mobile radio networks are according to M. Liljenstam and R. Ayani, "Partitioning PCS for Parallel Simulation", in MASCOTS ('97), 1997 partitioning by geographic area or partitioning by group of physical resources in the transmission channel.

A disadvantage of both partitioning strategies is that the model of the transmission channel must be shared in each case. By dividing what is actually only available once  Objects created in the transmission channel may have to communicate with each other. This creates additional inter-partition communication traffic.

With the geographical partitioning according to Fig. 2, all network elements of a certain section of the area to be simulated are combined in one partition.

Disadvantages of this partitioning strategy are:

• 1. If a movement of the individual network elements is modeled, then a high one can be used here Inter-partition communication traffic (a transfer of the moving Network element modeling object in another partition) can be expected if Network elements from one section of the area to be simulated to another Move the cutout.
• 2. This partitioning strategy can no longer or only at the expense of a very high inter-partition communication traffic (communication between objects, which parts of the transmission channel model) are used if that Study area is smaller than the propagation radius of the electromagnetic waves.

With the partitioning strategy of resource separation according to Fig. 3, all network elements are combined in one partition, to which the same physical resource (or a physical resource belonging to a defined group of resources) has been assigned, such as a time slot or a frequency.

Disadvantages of this partitioning strategy are:

• 1. A possible in the course of the operation of the mobile radio network due to several causes Reallocation of resources causes inter-partition communication traffic by transferring the object to another partition, which is the new one Modeled resource element getting network modeled.
• 2. Partitioning according to physical resources can no longer be used, when all network elements send and receive in the same physical resource. This is with cellular standards, which code division multiple access as Use multiple access scheme (e.g. UMTS), the case.

The object of the invention is to provide a method of the type mentioned at the outset, with which the control of the course of the communication of objects as models of  Network elements and the transmission channel of a cellular mobile messaging system can be simulated more efficiently and partitioned into parallel parts.

According to the invention, the object is achieved by a method in conjunction with those in the preamble of claim 1 features solved in that the individual network elements of Mobile network in an interaction cycle determined by sending and receiving be programmed and communicate with the transmission channel, with several Send and receive network elements simultaneously and independently of each other Network element repeats the interaction cycle in succession during the simulation Transmission channel between the states send and receive switchable programmed and switching between the two states based on a comparison of a known number of network elements and the number of previously sent or received information from a current state to the next state. It is not a central event store with a global order relationship of all in the Simulation occurring events more needed.

Advantageous variants of the method are the subject of dependent claims.

All network elements of a cellular network have access to a common one Transmission channel. These accesses follow a certain multiple access scheme, in which generally access more than one network element at a time Transmission channel needed. This access scheme limits channel access for the Transmission of a certain network element to one or more by frequency and time labeled physical resources.

It was found that a successive repetition of the access of the Network elements on the transmission channel. This behavior at the time level is common to everyone digital cellular networks regardless of their multiple access scheme. The reason but the architectures are the protocol stacks of digital transmissions in all Mobile networks from the second generation. The fact of digital transmission is one of the main distinguishing features of the second generation from their predecessors. Data are transmitted here in bursts and further in time slots. These time slots are in Frames fitted, which are then repeated in succession.  

The interactions between the objects transmission channel and network elements can can be advantageously modeled as follows. Network elements have the assignment and approval of assigned physical resources to the transmission channel. This Interactions correspond to sending a network element in the real system. The The transmission channel then calculates the propagation of the radio waves. Then he answers the network element by returning all assignments of physical resources as well as the Signal strength received at the current position of the requesting network element can be and which, e.g. B. in the case of a co-channel interference analysis, the same use physical resources.

Since the interactions for channel access of the network elements (the network element-external interactions) are repeated one after the other, there is a cycle in the interactions between the transmission channel and the network elements with the points receive and send as shown in Fig. 4.

In a variant of the method, the transmission channel is strictly passive in the sense that only Write network elements to him or receive something from him. The transmission channel handles these interactions immediately. However, it is never made up of network elements from another Reason interact than respond to interactions from them.

The interactions between the objects transmission channel and network elements are in the case of a passive object transmission channel modeled as follows. Network elements have that Allocation and release of assigned physical resources to the Communicate transmission channel. These interactions correspond to sending one Network element in the real system. In contrast, if a network element information from Transmission channel, it must, due to the passivity of the Transmission channel, first send a receive request to the transmission channel. The The transmission channel then calculates the propagation of the radio waves. Then he answers on the receipt request of the network element by returning all assignments physical resources as well as the signal strength, which at the current position of the requesting network element can be received and which, for. B. in the case of a Co-channel interference analysis using the same physical resources.  

Since the interactions for the channel access of the network elements (the network element-external interactions) repeat successively, there is a cycle in the interactions between the transmission channel and the network elements with the items receive requests, receive responses and transmit as shown in Fig. 5.

These points of the interaction cycle mark points in time in the simulation time. Through the Independence of the network elements is the order of communication between the different objects network element indefinitely at a time of the interaction cycle. In plain text this means that which is the first of many network elements to send or receive indefinite. Only the order of the points in the interaction cycle (receive - Send or receive request - receive response - send). Because of the vagueness the order of communication of different network element objects at a time of the interaction cycle does not have to be a global order of all communications in the Simulation are observed. In contrast, compliance is global Order relationship as in simulations using discrete events according to the state of the art Technology an over-specification of the model.

The interaction cycle repeats with every frame or super frame of the investigating mobile radio system. The simulation time can then be measured by counting only the number of interaction cycles and the length of the chosen one Frame duration is multiplied.

Due to the interaction cycle found, the order of communications between the network element and transmission channel objects are now known in advance (Sequence of the times of the interaction cycle or sequence of the individual Interaction cycles) or partially indefinite (different network elements to one Time). This eliminates the need for a global order relationship through an event memory be respected.

In contrast, according to the invention, exactly one object in the executable model of the new domain for sequence control has a special task. In the case of an executable model of the radio network level of a mobile radio network, the object which models the transmission channel can take over the sequence control for the following reasons. It is then the only object that:

• - works with global time,
• - has only one instance in the model and
• - participates in every interaction cycle.

All network elements are proposed as the partitioning strategy according to the invention of the same type regardless of their current position and regardless of the assigned one unite physical resource in one partition (partitioning according to the type of Network element). As an immediate advantage, inter-partition communication traffic never occurs here when moving a network element or reassigning a physical resource on. Another advantage of this partitioning strategy is that it also works with geographic Simulation areas, which are smaller than the propagation area of the radio waves and at Cellular networks with only one physical resource partition the whole Enable simulation models in the first place.

Thus, several network element objects are combined in one partition. The communications of this Objects within the partition as well as the network element internal communications can be searched for the state of the art can be controlled by means of discrete events.

In a further variant of the method, the activation of network elements and Services triggered during the simulation run by random functions, which also the Determine the point in time and the lifespan of a network element or service. The Network element activation generates intermediate arrival and service times, which determine the time and determine the lifespan of a network element in the executable model. In plain text it gives the information from when and for how long a network element of a certain type Is part of the model topology. The interactions between the network element activation and the network elements have an order relationship with the interactions between the Network elements and the transmission channel. So they have to use the same Order relationship are treated. However, unlike consecutive channel access, random behavior.

It must be ensured that the interactions during the interaction cycle in the in the correct order. Here are the random ones in particular  Configuration events, which are output by the network element activation, in a Order relationship with the interactions between the transmission channel and the Classify network elements.

It can be determined that an interaction cycle has ended when all Network elements point the allocation of resources for the next interaction cycle Have announced the transmission channel. Only then can the transmission channel start the calculations for the receive responses or new receive requests accept. Receipt responses would be calculated, if not all Interactions in the point of sending are processed, then there is a causality in the Resource allocation is not guaranteed and the receive responses would not be the exact one Superposition situation of radio signals in the transmission channel at the time of the current one Reflect the interaction cycle. It must therefore be guaranteed in the execution that the first Receive response is only output from the transmission channel when all Interactions in the point of sending in the transmission channel were processed and all Receive requests have been sent from all network elements. Otherwise independent running simulations in the objects transmission channel and network elements must meet exactly at this point. On the other hand, the network elements can interact to transmit and receive request at any point in the cycle, after receiving the receive response. The sequence control of the simulation itself ensure that these interactions are processed in the correct order.

The invention is explained in more detail below on the basis of exemplary embodiments. In the associated drawings show:

Fig. 1 is a representation to illustrate discrete events,

Fig. 2 is an illustration to illustrate geographic partitioning,

Fig. 3 is an illustration to illustrate the partitioning by resources,

Fig. 4 is an illustration for illustrating the method according to the invention,

Fig. 5 shows the procedure for passivity of the transmission channel,

Fig. 6 illustrates a partitioning according to the invention,

Fig. 7 shows a representation with passivity of the transmission channel with activation events,

Fig. 8 shows the saving of computing time when using the simulation method according to the invention on a multiprocessor computer.

Figs. 1 to 3 serve to illustrate the state of the art and have already been explained in more detail there.

Eight time slots are used as an exemplary embodiment of the interaction cycle in the GSM system with a duration of 0.577 ms in a frame with a duration of 4.615 ms. It now depends on the desired time resolution of the simulation of the radio network level, which duration as Basic element of the interaction cycle is used. When implemented in one The simulator for the GSM radio network level is the frame duration of 4,615 ms as the basic element used. This use has the advantage that every active network element on each Interaction cycle takes part because each active network element one of the eight time slots must be assigned.

As an exemplary embodiment for implementing the invention, one should run itself controlling variant of the simulation of the radio network level of a mobile radio network is described become.

A good way to describe a model of a system, the parts of which act in parallel, but must interact with each other is to use the notations of the process calculus, e.g. B. the Communicating Sequential Processes (CSP), the Calculus of Communicating Systems (CCS). Process calculi are formal algebras for the description of interacting only Sharing a system. These parts are commonly called processes. All that Happens within a process, is not seen from outside and is not considered.

CSP uses rendezvous-style interactions, or in other words, looks at everyone Interactions as synchronous. This means that an interaction between processes can only take place when all participating objects (sender and receiver) are ready for it. If one of the Objects is not ready for interaction, then all other objects must wait (the objects are blocked).

In contrast, the sender of a message does not have to be in an asynchronous interaction wait for the recipient to be ready to receive the message because the message is from of interaction until the receiver is ready. But that means whenever  Receiver wants to receive something, but the sender isn't ready for interaction then the receiver must wait (the receiving object is blocked).

Synchronous and asynchronous rendezvous interactions can now be used to Communications between the transmission channel, the network elements and the objects Describe network element activation.

The first step in describing a system using the process calculus notation is to define an interaction alphabet. Since all internal calculations of an object are completely hidden inside, an object is fully described by its interaction alphabet. The interactions in the radio network level of a mobile radio network can be divided into two groups:

• 1. The points of the interaction cycle.
• 2. The randomly occurring interactions of the network element activation.

The communications between the named objects can now be made using the Knowledge of the course of the interaction cycle can be implemented as follows.

It is assumed that J (z) network elements are active in the interaction cycle z. The following communications are then possible:

• 1. For the first J (z) - 1 network elements for all points of the interaction cycle (Communications receive request - receive response - send or receive - Send): Asynchronous rendezvous interactions. Justification: The network elements are independently of each other, the order of their interactions with the transmission channel a point in time of the interaction cycle is indefinite.
• 2. For the network element J (z) for all mentioned points of the interaction cycle: synchronous Rendezvous interactions. Reason: According to the invention, the order of the points of the interaction cycle. There is one for each interaction cycle Information necessary if the last network element per point is communicating performed. This information can e.g. B. by internal counters in the process controlling object transmission channel can be delivered.
• 3. The interactions of the network element activation break into the interaction cycle according to Fig. 7. These breakpoints must be planned by the transmission channel object, since this object controls the simulation process according to the invention.

The invention was used as an exemplary embodiment for implementation Simulation control based on PtolemyII 0.3 implemented by UC Berkeley. PtolemyII is a library for heterogeneous concurrent modeling and design, which written entirely in the Java programming language and released to the public has been. It offers a general program infrastructure that can be used to executable models of heterogeneous systems using a variety of To implement simulation procedures. The semantics of a special simulation procedure is implemented in a so-called domain, which is able to model a System using this special simulation method. Furthermore, the PtolemyII program infrastructure allows hierarchically composed models of one System, their execution different domains and thus simulation methods includes.

In order to implement the simulation control according to the invention, a new domain was created created for PtolemyII. This domain allows synchronous and asynchronous as well Breakpoint-based interactions between Objects which are referred to as actuators in the PtolemyII conceptual world.

This simulation control was used to simulate a cellular network with a UMTS-UTRA / FDD air interface to control.

In the following investigations, the new domain is used as a simulation control for the Network element-external communications used during a sequential Simulation control using discrete events for the network element internal Communications and is used for communications within the partitions. This overall hierarchical simulation control is carried out sequentially Simulation control using discrete events on various computer platforms tested.

The performance of both simulation controls compared on a computer with two processors:
A SunOS 5.6 sun4u sparc SUNW, Ultra-2 computer with 300 MHz clock speed and Solaris VM JDK1.2.1.02, native threads, using sunwjit. The results are shown in Fig. 8.

On this multiprocessor computer, the heterogeneous simulation control leads to faster Simulations.

Claims (7)

1. A method for parallel simulation of mobile radio networks, which consist of sendable and receiveable network elements, which transmit into and receive from the transmission channel, network elements and transmission channel being modeled as objects in object-oriented programming in a program, and the transmission and reception the network elements in the program is programmed by communication of the objects, characterized in that

• a) the individual network elements of the mobile radio network are programmed in an interaction cycle determined by sending and receiving and communicate with the transmission channel, with several network elements transmitting and receiving independently at the same time and each individual network element repeating the interaction cycle in succession during the simulation,
• b) the transmission channel between the states send and receive is programmed switchable and
• c) the switching between the two states takes place on the basis of a comparison of a known number of network elements and the number of information previously sent or received from a current state to the next state.

2. The method according to claim 1, characterized in that the switching of Then transmission channel from send to receive or from receive to send takes place when the comparison of the number of network elements and sent or received information results in equality. 3. The method according to claim 1 or 2, characterized in that the simulation time is determined by counting the number of interaction cycles and the sum with the Length of the frame duration of the cellular standard to be simulated is multiplied. 4. The method according to claim 1 to 3, characterized in that the transmission channel is operated passively, the points reception request, Receive response and send are included.   5. The method according to claim 1 to 4, characterized in that in the interaction cycles additional events for the activation of further network elements are recorded. 6. The method according to claim 1 to 5, characterized in that for partitioning Network elements are grouped together and the groups parallel information process. 7. The method according to claim 6, characterized in that the merger of the Network elements are made into groups according to the type of network element.