JP2016151902A - Device selection method and signal processing control apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To select an optimal execution device for executing predetermined signal processing, from among a plurality of signal processing devices.SOLUTION: A communication relay section 10 selects an execution device for executing predetermined signal processing, from among a plurality of servers. A vector calculation section 1042 represents the amount of available processing resources in each of the servers, as a plurality of device vectors using the processing resources as coordinate axes, and represents the amount of processing resources to be required for the signal processing, as a processing load vector using a processing resource as a coordinate axis. A device selection section 1044 selects a signal processing device corresponding to a device vector that forms the smallest angle with the processing load vector, as an execution device.SELECTED DRAWING: Figure 4

Description

  The present invention relates to a device selection method and a signal processing control device for selecting an execution device that executes predetermined signal processing from a plurality of signal processing devices.

In recent years, virtualization of communication networks by virtualization of network devices has been studied (for example, see Non-Patent Document 1). In other words, a technique has been studied in which a communication network configured by arranging a dedicated device for each network function is configured using a cloud system using a general-purpose server group or a transfer network using a general-purpose transfer device.
In a virtualized communication network, physical resources are shared between virtualized instances, and it is considered that the use resource distribution for each instance is optimized.

Hidehiro Arimitsu and 6 others, “Current Status of Technology Development for Network Virtualization”, NTT Technical Journal, Nippon Telegraph and Telephone Corporation, May 2014, Vol26, No5, pages 6-9

In the conventional network, a dedicated device is used for each network function, and the transmission method and the size of the communication frame are determined, including SDH (Synchronous Digital Hierarchy) and ATM (Asynchronous Transfer Mode). Yes.
Therefore, the amount of signals that can be processed on the network can be easily derived, and the facility design can be easily performed by increasing the processing resources by the necessary amount.
On the other hand, in the virtualized communication network as described above, it is expected that the choices of the resource type and amount such as memory and CPU will increase and the balance will be diversified. In addition, it is necessary to communicate various communication types, including various call types, and it is expected that selection of processing resources, appropriate distribution of signal processing devices, and facility design will be difficult compared to conventional methods. .

  The present invention has been made in view of such circumstances, and an object thereof is to select an execution device that is optimal for executing predetermined signal processing from a plurality of signal processing devices.

  In order to achieve the above-described object, the invention according to claim 1 is a device selection method for selecting an execution device that executes predetermined signal processing from a plurality of signal processing devices, and each of the plurality of signal processing devices. Represents the available amount of processing resources as a plurality of device vectors with the processing resources as coordinate axes, and represents the usage amounts of the processing resources required for the signal processing as processing load vectors with the processing resources as coordinate axes. A device selection method comprising: a vector calculation step; and a device selection step of selecting an execution device that executes the signal processing based on the plurality of device vectors and the processing load vector.

  The invention according to claim 5 is a signal processing control device that selects an execution device that executes predetermined signal processing from a plurality of signal processing devices, and is a usable amount of processing resources in each of the plurality of signal processing devices. As a plurality of device vectors with the processing resources as coordinate axes, and vector calculation means for expressing the amount of the processing resources required for the signal processing as processing load vectors with the processing resources as coordinate axes, A signal processing control device comprising: device selection means for selecting an execution device that executes the signal processing based on the device vector and the processing load vector.

  In this way, an appropriate signal processing device is selected as an execution device in consideration of the balance between the processing resources of each signal processing device and the processing resources required for the target signal processing. can do. In addition, since the processing capability of the signal processing device is expressed as a vector with each processing resource as a coordinate axis, arithmetic processing can be easily performed even when the number of processing resources to be evaluated is large.

  According to a second aspect of the present invention, in the device selection step, the signal processing device corresponding to the device vector having the smallest angle with the processing load vector is selected as the execution device among the plurality of device vectors. The device selection method is characterized by this.

  According to a sixth aspect of the present invention, the device selection means selects, as the execution device, the signal processing device corresponding to the device vector having the smallest angle with the processing load vector among the plurality of device vectors. The signal processing control device is characterized by this.

  By doing so, it is possible to select a signal processing device having the processing capability closest to the balance of processing resources necessary for signal processing as a target, and to efficiently perform signal processing.

  According to a third aspect of the present invention, when a plurality of the signal processes are executed using a single signal processing apparatus, the vector calculation step calculates the processing load vector for each of the signal processes. In addition, a combined vector of the plurality of calculated processing load vectors is calculated, and in the device selection step, the signal processing device corresponding to the device vector having the smallest angle with the combined vector among the plurality of device vectors is calculated. The apparatus is selected as the execution apparatus.

  In a seventh aspect of the present invention, when a plurality of the signal processes are executed using a single signal processing device, the vector calculation means calculates the processing load vector for each of the signal processes. In addition, a combined vector of the plurality of calculated processing load vectors is calculated, and the device selection unit selects the signal processing device corresponding to the device vector having the smallest angle with the combined vector among the plurality of device vectors. The signal processing control device is selected as the execution device.

  By doing in this way, when performing several signal processing using a single signal processing apparatus, the most suitable signal processing apparatus can be selected easily. In addition, since the signal processing devices can be assigned to a plurality of signal processings simultaneously, the efficiency of the processing system can be improved.

  According to a fourth aspect of the present invention, when a single signal processing is executed using a predetermined signal processing device group among a plurality of the signal processing devices, each vector processing step includes the signal processing device. And calculating a device vector corresponding to the signal processing device included in the signal processing device group, and calculating the processing load vector of the combined vector in the device selection step. The apparatus selection method is characterized in that the signal processing apparatus group corresponding to the combined vector having the smallest angle is selected as the execution apparatus.

  According to an eighth aspect of the present invention, when the single signal processing is executed using a predetermined signal processing device group among the plurality of signal processing devices, the vector calculation means is provided for each of the signal processing devices. And calculating a device vector corresponding to the signal processing device included in the signal processing device group, and the device selecting means includes the processing load vector of the combined vector. The signal processing device group corresponding to the synthesized vector having the smallest angle is selected as the execution device.

  By doing in this way, when performing a single signal processing using a plurality of signal processing devices (signal processing device group), the most suitable signal processing device group can be easily selected. In addition, since a plurality of signal processing devices can be assigned to a single signal processing, signal processing with a large processing load can be quickly processed.

  According to the present invention, it is possible to select an optimal execution device for executing predetermined signal processing from a plurality of signal processing devices by combining processing resources of the signal processing device and processing resources necessary for signal processing. it can.

It is explanatory drawing which shows typically the structure of the network to which the communication relay part concerning embodiment is applied. It is explanatory drawing which shows typically the communication path | route between the terminals connected to a different communication network. It is explanatory drawing which shows an example of the format of a SIP signal. It is a block diagram which shows the functional structure of a communication relay part. It is explanatory drawing which shows an example of the combination of a codec, and an execution apparatus. It is explanatory drawing which shows an example of the information for transmission to an execution apparatus. It is explanatory drawing which shows an example of the vector calculated by the vector calculation part. It is a flowchart which shows the process sequence in a communication relay part. It is a flowchart which shows the process sequence in a communication relay part.

Exemplary embodiments of a device selection method and a signal processing control device according to the present invention will be explained below in detail with reference to the accompanying drawings.
In the present embodiment, a case will be described in which a device selection method according to the present invention is applied to a communication relay unit 10 (see FIGS. 2 and 4) that connects different communication networks and functions as a signal processing control device. .

(Embodiment)
FIG. 1 is an explanatory diagram schematically illustrating a configuration of a network 20 to which the communication relay unit 10 according to the embodiment is applied.
The network 20 is a virtualized network system, and includes a cloud system C configured by a general-purpose server 202 and a transfer network N configured by a general-purpose transfer device 204.

In the cloud system C, network functions that have been realized using dedicated hardware are converted into software and operated on the general-purpose server 202 to realize control functions and service functions in the network 20. That is, the control function and the transfer function that are conventionally integrated in the network device are separated, and the control function is converted into software and collected in the cloud system C configured by the general-purpose server 202.
The transfer network N is formed by general-purpose transfer devices 204 connected to each other. Since the control functions are concentrated in the cloud system C as described above, the general purpose transfer device 204 may be a device having a function of transferring data to a designated transfer destination, such as a general purpose switch.

On the cloud system C, a network controller 22 is provided as an application for controlling the network 20.
The network controller 22 manages and controls both computing resources arranged in the cloud system C as a plurality of general-purpose servers 202 and network resources (general-purpose transfer devices 204) constituting the transfer network N.
The network controller 22 includes an orchestrator that manages and controls the resource allocation status of the entire network 20 and a transfer-side controller that controls the general-purpose transfer device 204.

The orchestrator performs control such as dynamically changing the resource allocation in response to a change in demand for each service provided on the network 20 and a request for resource allocation from the service provider.
The transfer controller receives an instruction from the orchestrator and controls the transfer network N configured with the general transfer device 204. The interface and protocol specifications between the transfer-side controller and the transfer network N are called Southbound API, and OpenFlow or the like can be applied.

A plurality of servers 30 (30A, 30B) are connected to the network 20. The server 30 may be a virtual server or an individual physical server. Further, the server 30 may constitute a part of the cloud system C.
Each of the servers 30A and 30B has processing resources such as a CPU and a memory, and has different signal processing capabilities.
In the present embodiment, the communication relay unit 10 (signal processing control device) selects a signal processing device that implements the codec conversion processing unit 12, that is, an execution device that executes codec conversion, from the plurality of servers 30. Have

In the present embodiment, communication between terminals 40A and 40B connected to network 20 will be described as an example.
In the present embodiment, terminals 40A and 40B are SIP (Session Initiation Protocol) terminals, and after starting a session using SIP signals, media signals such as voice data, video data, and text data are transmitted to each other. Real-time communication. Hereinafter, the terminal 40A is referred to as a caller (user agent / client), and the terminal 40B is referred to as a callee (user agent / server).

FIG. 2 is an explanatory diagram schematically showing a communication path between the terminals 40A and 40B.
In FIG. 2, terminal 40A and terminal B are connected to different communication networks (communication network A and communication network B).
The communication relay unit 10 (signal processing control device) relays communication between terminals 40A and 40B respectively connected to different communication networks (communication network A and communication network B) at the boundary of the communication network.
Different communication networks include, for example, IP telephone service networks operated by different carriers, internal networks operated by small and medium-sized enterprises, and interconnection of unified communication services.

The communication relay unit 10 is a functional unit corresponding to a conventional session border controller. The conventional session border controller mainly has the following functions.
Among the following functions, the codec conversion function is executed by the codec conversion processing unit 12 described later.
1. Codec conversion function A function to perform codec conversion in order to match the codec of U plane (user plane, function group for exchanging media signals) data.
2. SIP difference absorption function A function for absorbing differences in order to allow communication networks to communicate with each other within the SIP protocol range defined by RFC3261 and the like. That is, when a difference occurs in the header part of the SIP signal, which will be described later, between the communication networks, this is complemented.
3. Topology hiding function A function for concealing the network configuration in the communication network, such as the IP address and port number, from the outside of the communication network.
4). QoS (Quality of Service) control function A function for adjusting the priority DSCP (Differentiated Services Code Point) of packets that differ between communication networks.
5. Pinhole control function A function that opens and closes a U plane port in order to input and output U plane data under the conditions determined by the C plane (control plane, a function group for exchanging SIP signals described later).

  The communication relay unit 10 may be connected to a communication network other than the communication network A and the communication network B. The communication relay unit 10 sets a scenario for each route for each communication network. A scenario is configured by rules for each of incoming and outgoing directions to a communication network, rules for handling SIP headers, rules defining handling and operation of parameters of each header, and the like. Specifically, for example, the execution device 602, the setting protocol 604, the setting address 606, the setting port number 608, the transmission destination address 610, the transmission destination port number 612, and the like shown in FIG. 6 are included.

In the present embodiment, the communication relay unit 10 and the SIP server 212 are realized as applications on the cloud system C.
The codec conversion processing unit 12 is realized by any one of the servers 30 selected by the communication relay unit 10.

Connection between the terminals 40A and 40B is performed via the SIP server 212 (212A and 212B). In FIG. 2, the terminals 40A and 40B are connected to the SIP server 212 via the edge routers 214 (214A and 214B), respectively. Further, the SIP servers 212 (212A, 212B) are connected to the communication relay unit 10 at the boundaries of the respective communication networks. This path is the C plane.
In FIG. 2, two SIP servers 212 are shown in one communication network, but the number of SIP servers 212 can be set arbitrarily.

The connection between the terminals 40A and 40B will be described in more detail. The terminals 40A and 40B, which are user agents, each transmit a registration message including identification information and IP address of the own terminal to the SIP server 212 at the time of activation. The SIP server 212 records the registration information of the terminals 40A and 40B in the database.
When the terminal 40A makes a voice call or the like to the terminal 40B, the terminal 40A transmits a SIP signal including an outgoing message (INVITE message) specifying the identification information of the terminal 40B to the SIP server 212.
The SIP server 212 searches the information in the database and transfers the outgoing message to the IP address corresponding to the specified identification information. When there are a plurality of SIP servers 212 between the terminal 40A and the terminal 40B, outgoing messages are sequentially transferred between the SIP servers 212.
Further, at the boundary between the communication network A and the communication network B, the communication relay unit 10 complements the difference of the SIP signal as appropriate by the SIP difference absorption function.
The terminal 40B that has received the outgoing message returns a response message to the SIP server 212, and the SIP server 212 transfers the response message to the terminal 40A.

Through this series of flows, the terminal 40A and the terminal 40B can know the IP address of the other party included in the message.
After such a connection process, the terminals 40A and 40B directly send media signals via the U plane, not via the SIP server 212.

FIG. 3 is an explanatory diagram showing an example of the format of the SIP signal.
The SIP signal is divided into a start line, header information, and a body part sandwiching a blank line. The header information includes an identifier sip: alice @ atranta. com, an identifier sip: bob @ biloxi. com, message type (INVITE), and the like.
In the body portion, session information such as a media streaming IP address such as voice and a compression format is described using a description syntax called SDP (Session Description Protocol).

In particular, the following information is described for the third and lower lines (c line, m line, a line: reference numeral 201) from the lower part of the body part.
Line c: Connection information, the address and port number of the medium that the user wants to receive are described.
m line: Media type The following information is described in more detail. A plurality of m lines can be set.
m = <media type><portnumber><transportprotocol><mediaformat><0 = G. 711 u-Law, 18 = G. 729>
Line a: Media attribute (RTP (Real-time Transport Protocol) attribute in FIG. 3)
In more detail, the following information is described. Multiple rows can be set for row a.
a = rtpmap <payload type><codec><samplingrate>

At the start of communication between the terminals 40A and 40B, a codec and the like to be used are negotiated using the SIP signal as shown in FIG. This is because the conditions such as codec and bandwidth that can be communicated differ depending on the specifications of the terminals 40A and 40B and the specifications of the communication network to which the terminals 40A and 40B are connected.
The terminal 40A on the transmission side designates the codec used in the current communication in the INVITE message. The receiving terminal 40B returns an OK response if the designated codec is usable. If the specified codec cannot be used, an error response is returned.
In the present embodiment, even if there is no codec that can be used in common between the terminals 40A and 40B, the codec conversion function of the communication relay unit 10 can perform conversion and transmission.
The media signal after the connection processing is transmitted through a route not via the SIP server 212, that is, a route via the edge router 214 (214A, 214B), the core router 216 (216A, 216B), and the codec conversion processing unit 12. This path is the U plane.

FIG. 4 is a block diagram illustrating a functional configuration of the communication relay unit 10.
The communication relay unit 10 includes a conversion determination unit 102, a processing device selection unit 104, a transmission destination control unit 106, and a control signal output unit 108.
The conversion determination unit 102 determines whether it is necessary to convert the codec of the media signal transmitted between the terminals 40A and 40B.
The conversion determination unit 102 determines the necessity of codec conversion from the SIP signal transmitted and received at the start of a session between the terminals 40A and 40B. That is, referring to the SIP signal, it is determined that it is not necessary to convert the codec when there is a codec that can be used by each of the terminals 40A and 40B, and when there is no matching codec, the codec is converted. Judge that it is necessary.

In each communication network (communication network A, communication network B), there are a plurality of codecs that can be handled depending on the types of the terminals 40A and 40B and the conditions of the communication network, and the codec to be used is a session by negotiation between the terminals 40A and 40B. Decide for each. That is, a plurality of codec conversion patterns (combination before and after conversion) can be taken between specific communication networks. Therefore, it is necessary to select which conversion pattern is to be applied from among a plurality of conversion patterns.
In addition, a plurality of media that can be allowed between terminals or in a communication network may be specified at the time of negotiation. This is so that a more appropriate codec can be selected as a result of the negotiation, and the codec is selected so that higher quality communication can be performed even when codec conversion is involved.

When it is determined that the codec conversion of the media signal is necessary, the processing device selection unit 104 determines an execution device that executes the codec conversion processing. That is, the server 30 to be the codec conversion processing unit 12 is selected.
The processing device selection unit 104 is further configured by a vector calculation unit 1042 (vector calculation unit) and a device selection unit 1044 (device selection unit).
The vector calculation unit 1042 represents the usable amount of processing resources in each of a plurality of signal processing devices (servers 30A and 30B in the present embodiment) as a plurality of device vectors with the processing resources as coordinate axes, and a predetermined signal. The amount of processing resources required for processing (in this embodiment, codec conversion processing) is expressed as a processing load vector with processing resources as coordinate axes.
The device selection unit 1044 selects an execution device that executes predetermined signal processing based on a plurality of device vectors and processing load vectors.

FIG. 7 is an explanatory diagram showing an example of a vector calculated by the vector calculation unit 1042 (vector calculation means).
In FIG. 7, for each of the server 30A and the server 30B, the CPU usable amount is taken on the X axis and the memory usable amount is taken on the Y axis, and the processing resource usable amount of each server 30A, 30B is a vector from the origin O. It is represented by a and b.
That is, the processing capability of the server 30A is indicated as a device vector a (Xa, Ya), and the processing capability of the server 30B is indicated as a device vector b (Xb, Yb).
In general, when the processing target is a single signal, the relationship between the number of signals that can be processed and the CPU usage rate is linear. Similarly, when the processing target is a single signal, the relationship between the number of signals that can be processed and the memory usage is linear. For this reason, the balance of the processing capability of the signal processing apparatus can be expressed by a vector.
Comparing the device vector a and the device vector b, it can be seen that the server 30A is in a state where the memory has a margin compared to the CPU, and the server 30B is in a state where the CPU has a margin compared to the memory.

Note that the processing capability of the processing resource may be evaluated as follows in addition to the comparison with the capability for processing the same amount of the same signal per unit time.
First, the CPU processing capacity of the signal processing device is converted into a CPU usage rate per unit time used for processing the same signal in the same amount. When a plurality of CPUs are mounted, the coefficient indicating the number is multiplied. Further, when the CPU is composed of a plurality of cores and the cores are assigned to the processing, the division is performed according to the number of cores. Further, when there is a case where the CPU resources are not used in full, such as 70% operation, division is performed according to this numerical value. That is, by setting the CPU usage rate to 70%, the execution device can be selected by changing the balance of the processing capabilities of the signal processing device, that is, changing the vector angle.
Further, the memory speed of the signal processing device is converted into the amount of memory that can be processed per unit time. When a plurality of memories are installed, the coefficient indicating the number is multiplied.

In FIG. 7, the amount of processing resources (processing load of signal processing) necessary for target signal processing (codec conversion in the present embodiment) is represented by a vector p (Xp, Yp).
The processing load of the signal processing is represented by a CPU usage rate and a memory amount per unit time necessary for executing the processing.

The device selection unit 1044 (device selection means) selects a signal processing device corresponding to the device vector having the smallest angle with the processing load vector p as an execution device.
That is, among the plurality of device vectors a and b, the smaller the angle with the processing load vector p is selected as a signal processing device with an appropriate resource balance.
In the example of FIG. 7, the angle θ ₁ formed by the device vector a and the processing load vector p is smaller than the angle θ ₂ formed by the device vector b and the processing load vector p. Therefore, the server 30A corresponding to the device vector a is selected as the execution device.
Thus, by expressing the processing capability of each signal processing device and the signal processing to be executed as a vector with the processing resources of the signal processing as coordinate axes, the characteristics of signal processing (for example, a large amount of memory is occupied) It becomes easy to select a signal processing device suitable for signal processing with a light CPU processing) as an execution device.

Note that the angle θ ₁ formed by the device vector (in the following equation (1), the device vector a) and the processing load vector p is obtained from the following equation (1).

Further, in FIG. 7, the device vectors a and b and the processing load vector p are represented by a two-dimensional vector using two axes of the CPU and the memory, but N-dimensional with three or more (N) processing resources as coordinate axes. It may be a vector. As processing resources, the amount of disk and interface speed can be used in addition to the CPU and memory.
Even in this case, the angle θ formed by the device vector and the processing load vector p can be obtained from the following equation (2).
In the following expression, X, Y, and Z are coordinate axes and correspond to processing resources used for evaluation of the signal processing device.

When performing a plurality of signal processing with a single signal processing device, the signal processing device is selected using a combined vector obtained by combining the processing load vectors pn (n = 1, 2,...) Of each signal processing.
More specifically, for example, when the signal processing Q and the signal processing R are processed by a single signal processing device, the processing load (amount of used resources) of the signal processing Q is determined by the processing load vector q (Xq, Yq), the signal processing R Is a processing load vector r (Xr, Yr), and a combined vector s (Xs, Ys) of the processing load vectors q and r is compared with a device vector of each signal processing device.

  That is, when performing a plurality of signal processing using a single signal processing device, the vector calculation unit 1042 calculates a processing load vector for each signal processing, and combines the calculated processing load vectors. The vector is calculated, and the device selection unit 1044 selects a signal processing device corresponding to the device vector having the smallest angle with the combined vector among the plurality of device vectors as the execution device.

Further, when a plurality of signal processing devices are virtually regarded as one signal processing device, device vectors corresponding to the plurality of signal processing devices may be synthesized. In this case, when the signal processing devices having the same specifications are regarded as one, the angle between the signal processing devices is almost equal because the vector does not change in each signal processing device. “Almost” means that it is necessary to consider the overhead when using middleware technology such as a hypervisor when considering a plurality of signal processing devices as one unit. It means that there is.
Further, when combining originally virtualized signal processing apparatuses, there is little overhead at the time of virtualization, and it is only necessary to consider the overhead for functions to be regarded as one unit.

  That is, when performing a single signal processing using a predetermined signal processing device group among a plurality of signal processing devices, the vector calculation unit 1042 calculates a device vector for each signal processing device, The combined vector of the device vectors (device vector group) of the signal processing devices included in the processing device group is calculated, and the device selection unit 1044 performs signal processing corresponding to the combined vector having the smallest angle with the processing load vector among the combined vectors. A device group is selected as an execution device.

In addition, since the device selection unit 1044 selects the signal processing device corresponding to the device vectors a and b having the smallest angle with the processing load vector p as the execution device, the device selection unit 1044 performs other processing even while the execution device is executing signal processing. There may be remaining processing resources that can perform signal processing.
That is, when the processing load vector p and the device vectors a and b completely match, it is considered that the execution device performs signal processing using the processing resources completely, but the processing load vector p and the device vector a , B and the device vectors a, b are long, there is a possibility that any processing resource is not used during signal processing.
In this case, the device vector indicating the remaining resources of the signal processing device and a _2, the device vector indicating the processing resources being used as a _1, the synthesis vector is original equipment with a device vector a ₂ a device vector a ₁ The vector a is set.
That is, the above determination is performed using a device vector a ₂ having an angle different from that of the original device vector a.

FIG. 5 is an explanatory diagram illustrating an example of a combination of codecs and an execution device.
Table 50 in FIG. 5 shows the codec 502 before conversion, the codec 504 after conversion, and an execution device 506 corresponding to a combination thereof.
The first line combination is G. A media signal that has been codec in 711 a-Law 711 is converted to u-Law. The server 30A (device A) is designated as the execution device for this conversion process.
The second line combination is G. 711 A media signal that has been codec in u-Law 711 is converted to a-Law. In this case, the server 30A (device A) is designated as the execution device.
The combination on the third line is G. 711 A media signal that has been codec in u-Law is converted to AMR-WB. In this case, the server 30B (device B) is designated as the execution device.

For example, G. 711 a-Law to G.G. The conversion from 711 u-Law to A.G. Compared with the conversion to 711 u-Law, the processing load is small.
That is, when the codec methods are close (eg G.711 a-Law and G.711 u-Law), the processing load is relatively small, such as the sampling rate and bandwidth in the codec before and after conversion are the same. Become. On the other hand, G. For example, conversion between 711 u-Law and AMR-WB increases processing load.
The processing device selection unit 104 selects the server 30 that is most suitable for the target signal processing based on these processing loads, particularly the balance of the usage of each processing resource.

FIG. 6 is an explanatory diagram illustrating an example of information for transmission to the execution apparatus.
The table 60 in FIG. 6 shows an execution device 602, a setting protocol 604, a setting address 606, a setting port number 608, a transmission destination address 610, and a transmission destination port number 612.

The execution device 602 corresponds to the execution device 506 in FIG.
The setting protocol 604 is a protocol used when transmitting a media signal to the execution apparatus.
When the setting protocol 604 is selected, the terminal 40 on the transmission side transmits to the processing resource using a protocol stack provided in advance.
As the setting protocol, OpenFlow used in SDN (Software-Defined Network), NFV (Network Functions Virtualization), or the like can be used. By using these protocols, by setting the address and port number of the receiving terminal of the packet in the network controller 22 in the virtualized network 20, the receiving terminal 40B can be set via the desired execution device. Can be sent.

Actually, in the virtualized network, a device that serves as an entrance to the communication network is provided with a determination unit that discriminates the inflow packet to monitor the inflow packet, and is selected as the codec conversion processing unit 12 when the inflow packet is a media signal. Guide to the execution device.
For the guidance, a MPLS (Multi Protocol Label Switching) path, a VPN (Virtual Private Network), a tag VLAN (Virtual Local Area Network), or the like is used.

In the setting address 606, an IP address assigned to an interface used by the general-purpose transfer device 204 for connection to the network controller 22 is designated. In the setting port number 608, a reception port when the general-purpose transfer device 204 receives a control signal from the network controller 22 is designated.
In the transmission destination address 610, the IP address of the execution device (server 30A or server 30B) selected as the codec conversion processing unit 12 is designated. In the transmission destination port number 612, the reception port of the media signal in the execution device is designated.

  When a network address port translation (NAPT) device is relaying to the execution device, it is necessary to designate a device in front of the NAPT device as the transmission destination address 610 and the transmission destination port number 612.

Returning to the description of FIG. 4, the transmission destination control unit 106 controls to transmit the media signal from the terminal 40 </ b> A on the transmission side of the media signal to the processing resource determined by the processing device selection unit 104.
The transmission destination control unit 106 edits the SIP signal returned from the receiving side terminal 40B to the calling side terminal 40A, and the IP address or port number of the execution device determined as the codec conversion processing unit 12 as the transmission destination of the media signal. Is described.
The control signal output unit 108 transmits a control signal to the server 30 determined as the codec conversion processing unit 12 so as to execute the codec conversion processing of the media signal.

FIG. 8 and FIG. 9 are flowcharts showing a processing procedure in the communication relay unit 10.
In the flowchart of FIG. 8, the communication relay unit 10 acquires a SIP signal from the upstream SIP server 212 (step S800).
The communication relay unit 10 refers to the acquired SIP signal to determine whether a communication route is specified (step S802).
If the route is not specified (step S802: No), the communication relay unit 10 selects an error response (step S804) and transmits a SIP signal describing the error response to the upstream SIP server 212. (Step S816).

On the other hand, when the route is specified (step S802: Yes), the communication relay unit 10 acquires a scenario for each route (step S806) and determines whether codec conversion of the media signal is necessary. (Step S808).
When codec conversion of the media signal is not necessary (step S808: No), the communication relay unit 10 designates a specified transmission destination as the transmission destination of the media signal (step S810). The prescribed transmission destination is, for example, a receiving terminal or a server that performs processing other than codec conversion. The designation of the transmission destination means that the SIP signal is rewritten by reflecting various settings shown in FIG.

If codec conversion of the media signal is necessary (step S808: Yes), the communication relay unit 10 selects an execution device that performs codec conversion (step S812). The execution device selection process will be described in detail with reference to FIG.
When the execution device is selected, the communication relay unit 10 designates the selected execution device as the transmission destination of the media signal (step S814).
Thereafter, the communication relay unit 10 transmits a SIP signal reflecting the designated execution device to the downstream SIP server 212 (step S816), and ends the processing of this flowchart.

Next, the execution device determination process in step S812 will be described with reference to FIG.
First, the communication relay unit 10 acquires the processing load information of the target signal processing (step S900). The processing load information is information indicating the amount of processing resources necessary for target signal processing. For example, the communication relay unit 10 may associate a combination of codecs before and after conversion with processing load information in advance.
Next, resource information indicating each resource state of the servers 30A and 30B (signal processing devices) that are candidates for execution devices is acquired (step S902).

Subsequently, the communication relay unit 10 calculates a processing load vector using the processing load information acquired in step S900 and a device vector using the resource information acquired in step S902 (step S904).
Based on the calculated processing load vector and device vector, an execution device is selected from the servers 30A and 30B (step S906). Specifically, the server 30 corresponding to the device vector having the smallest angle with the processing load vector is selected as the execution device, and the processing according to this flowchart ends.
If there is another signal process to be executed by the server 30, the process returns to step S900 and the subsequent processes are repeated.

  As described above, according to the communication relay unit 10 according to the embodiment, the balance between the processing resources of each signal processing device and the processing resources required for the target signal processing is considered. Thus, an appropriate signal processing device can be selected as the execution device. In addition, since the processing capability of the signal processing device is expressed as a vector with each processing resource as a coordinate axis, arithmetic processing can be easily performed even when the number of processing resources to be evaluated is large.

In the present embodiment, the example in which the present invention is applied to the case of selecting a signal processing apparatus that performs codec conversion of a media signal has been described, but the application of the present invention is not limited to this and is applied to various processes. Is possible.
The application of the present application is such that the type of the next packet to be transmitted can be predicted by SIP negotiation or the like such as communication between SIP terminals, the type of the packet to be transmitted next can be manipulated, or the equipment such as routing A particularly high effect can be obtained by targeting a route that can be assigned to which route at the time of design.
That is, it is preferable to employ the case where the execution device can be selected immediately before receiving the first packet to be processed.

  In this embodiment, the case where SIP is used as a signaling protocol has been described. However, the present invention is not limited to this, and for example, XMPP (extensible messaging and presence protocol) can be used. In this case, for communication of media signals, STUN (Simple Traversal of UDP through NATs), ICE (Interactive Connectivity Establishment), TURN (Traverse Using Relay NAT), etc. are used as the destination address. it can. In the case of a home network, the WAN address and port number correspond.

10 Communication relay unit (signal processing control device)
102 Conversion Determination Unit 104 Processing Device Selection Unit 1042 Vector Calculation Unit (Vector Calculation Unit)
1044 Device selection unit (device selection means)
DESCRIPTION OF SYMBOLS 106 Transmission destination control part 108 Control signal output part 12 Codec conversion process part 20 Network 22 Network controller 202 General-purpose server 204 General-purpose transfer equipment 212 SIP server 214 Edge router 216 Core router 30 (30A, 30B) Server 40A, 40B Terminal

Claims (8)

A device selection method for selecting an execution device that executes predetermined signal processing from a plurality of signal processing devices,
The usable amount of processing resources in each of the plurality of signal processing devices is represented as a plurality of device vectors with the processing resources as coordinate axes, and the usage amount of the processing resources necessary for the signal processing is expressed as the processing resources. A vector calculation step represented as a processing load vector as coordinate axes;
A device selection step of selecting an execution device that executes the signal processing based on a plurality of the device vectors and the processing load vector;
A device selection method comprising: In the device selection step, the signal processing device corresponding to the device vector having the smallest angle with the processing load vector among the plurality of device vectors is selected as the execution device.
The apparatus selection method according to claim 1. When performing a plurality of the signal processing using a single signal processing device,
In the vector calculation step, the processing load vector is calculated for each of the signal processes, and a combined vector of the plurality of calculated processing load vectors is calculated.
In the device selection step, the signal processing device corresponding to the device vector having the smallest angle formed with the combined vector among the plurality of device vectors is selected as the execution device.
The apparatus selection method according to claim 1. When performing the single signal processing using a predetermined signal processing device group among the plurality of signal processing devices,
In the vector calculation step, the device vector is calculated for each of the signal processing devices, and a combined vector of device vector groups corresponding to the signal processing devices included in the signal processing device group is calculated.
In the device selection step, the signal processing device group corresponding to the composite vector having the smallest angle with the processing load vector among the composite vectors is selected as the execution device.
The apparatus selection method according to claim 1. A signal processing control device that selects an execution device that executes predetermined signal processing from a plurality of signal processing devices,
The usable amount of processing resources in each of the plurality of signal processing devices is represented as a plurality of device vectors with the processing resources as coordinate axes, and the usage amount of the processing resources necessary for the signal processing is expressed as the processing resources. Vector calculation means represented as a processing load vector as coordinate axes;
Device selection means for selecting an execution device that executes the signal processing based on a plurality of the device vectors and the processing load vector;
A signal processing control device comprising: The device selection means selects, as the execution device, the signal processing device corresponding to the device vector having the smallest angle with the processing load vector among the plurality of device vectors.
The signal processing control apparatus according to claim 5. When performing a plurality of the signal processing using a single signal processing device,
The vector calculation means calculates the processing load vector for each of the signal processes, calculates a combined vector of the plurality of calculated processing load vectors,
The device selection means selects, as the execution device, the signal processing device corresponding to the device vector having the smallest angle with the combined vector among the plurality of device vectors.
The signal processing control apparatus according to claim 5. When performing the single signal processing using a predetermined signal processing device group among the plurality of signal processing devices,
The vector calculation means calculates the device vector for each of the signal processing devices, calculates a combined vector of device vector groups corresponding to the signal processing devices included in the signal processing device group,
The device selection means selects, as the execution device, the signal processing device group corresponding to the synthesized vector having the smallest angle with the processing load vector among the synthesized vectors.
The signal processing control apparatus according to claim 5.

Images