JP2012080420A - Device and method for data transfer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable effective utilization of network resources and flexible scheduling of data transfer between end-to-end terminals.SOLUTION: A data transfer method includes: receiving surplus bandwidth information of each channel in an external network from a surplus bandwidth measurement device outside a router at constant time intervals; in regard to a plurality of flows output from a data transfer device, based on the surplus bandwidth information of each channel in the external network reported at constant time intervals and the transmission route of each flow, determining the data output speed of each flow at the constant time intervals, whenever necessary; adjusting a speed to output each flow data to the channel to the output speed of each flow data reported at the constant time intervals, whenever necessary; and while storing each flow data input from an input-side channel into storage means, adjusting and outputting to the output side channel the data at the output speed adjusted by an output speed control step at constant time intervals.

Description

  The present invention relates to a data transfer apparatus and method, and in particular, is arranged inside a computer network based on a packet switching method such as an IP (Internet Protocol) network, and is used for a network layer packet or a data link layer frame in an OSI reference model. The present invention relates to a data transfer apparatus and a data transfer method such as a router and a switch that control transfer.

  A standard configuration of a conventional router or switch is shown in FIG. In the case of a router, a packet in the network layer inputted from one line, and in the case of a switch, a frame in the data link layer (in the following, in order to simplify the description, it is unified in the former for the sake of simplicity) Are first stored in a buffer 123a which is a small-capacity memory on the line module 120a.

  The reason for storing the packet in the buffer 123a is to wait for the transfer until the buffer 123c on the line module 120c serving as the output destination of the packet becomes empty, the data reception speed from the line, the switch unit This is to absorb a temporary difference from the data transfer rate in the router via the network 130.

  While the packet is stored in the buffer 123a, the data transfer control unit 122a determines a transmission destination line module based on the transfer destination information obtained from the control unit 110.

  When the output destination line module is determined and it is confirmed that there is enough space to store the packet in the buffer 123c on the line module 120c, the packet in the buffer 123a of the line module 120a passes through the switch unit 130. The data is transferred to the buffer 123c on the output destination line module 120c. Then, after necessary processing is performed, the data is sent to the output side line.

  The conventional router repeats the process of transferring a packet input from each line to another line in the above flow. Unless there is intense contention for a specific output line inside the router, each traffic flow is output as it is at a speed almost the same as the data input speed at the time of input.

  A problem that arises in computer networks using conventional routers with the internal structure described above is that packets when the total amount of traffic destined for a certain line in the router exceeds the physical line capacity of that line. There is a problem that there is a possibility that a loss will occur.

  FIG. 10 shows that packet loss may occur when the total amount of traffic destined for a certain line in the router exceeds the physical line capacity of that line. This will be described with reference to the example of FIG. It is assumed that one flow from the terminal 10b connected to the router 20b flows to the line 30b and uses 70% of the physical line capacity (bandwidth) of the line 30b. Here, the physical line capacities of the line 30a and the line 30b are the same.

  In this situation, it is assumed that the terminal 10a connected to the router 20a sends a flow with a line bandwidth usage rate of 50%, and flows into the line 30b via the line 30a and the router 20b. In this case, the line 30b does not have a surplus bandwidth sufficient to flow the flows from the terminal 10a together. Therefore, if such a situation continues for a certain period of time, a small-capacity buffer inside the router 20b overflows, and there is a possibility that the packet of the flow from the terminal 10a and the terminal 10b will be lost (which flow Whether the packet is lost depends on the flow processing method in the router and the protocol used in the transport layer of the OSI reference model for each flow).

  When UDP (User Datagram Protocol) without a reliability guarantee mechanism such as a data retransmission function is used for the transport layer protocol of the OSI (Open Systems Interconnection) reference model, the above packet loss is immediately detected from end to end. Causes loss of communication data between terminals.

  When TCP (Transmission Control Protocol) having a reliability guarantee mechanism such as a rate control function to reduce packet loss probability and a data retransmission function is used for the transport layer protocol, the above packet loss is as follows: Cause problems as described.

  In TCP, the transmission side terminal transmits the amount of data (cwnd) that can be transmitted before receiving the reception confirmation response (Ack) of the data from the reception side terminal, waits for Ack, and when it receives Ack, it returns cwnd again. The process of transmitting the data amount is repeated (for example, see Non-Patent Document 1). (Strictly speaking, the cwnd value is compared with the receivable data amount rwnd notified from the receiving terminal, and data corresponding to the smaller value is transmitted).

  In TCP, the cwnd value starts from a relatively small value at the start of the session, and gradually increases the cwnd value while passing traffic. In other words, the operation of reducing the influence on other traffic at the start of traffic transmission and gradually increasing the throughput is performed.

  If the above packet loss occurs in any part of the communication path during this operation, TCP operates to temporarily reduce the throughput by halving the value of cwnd. As a result, if there is no packet loss, the cwnd value is gradually increased to enter the phase of recovering the throughput, but every time a packet loss occurs, the cwnd value is halved.

  Therefore, the finally stable throughput value is that of the line portion with the narrowest usable bandwidth (hereinafter referred to as the end-to-end bottleneck usable bandwidth) among all the circuits on the path between the transmitting and receiving terminals. Almost the same value. For example, in the example of FIG. 11, a TCP session established between the terminal 40a connected to the router 50a and the terminal 40b connected to the router 50d via the lines 60a to 60c and the routers 50a to 50d. Although data communication is performed, the communication throughput between the terminals is almost the same as the surplus bandwidth of the line 60b. Although the line 60a has more surplus bandwidth than the line 60b, the terminal 40a cannot transmit data at a speed higher than the usable bandwidth of the line 60b.

  In addition, in TCP, as described above, every time packet loss occurs due to network congestion, the cwnd value is halved and then increased again, so the data transmission rate from the transmitting terminal during network congestion is There is also a problem that when the original data transmission speed is high, it takes time to recover the transmission speed after reducing the value of cwnd when the original data transmission speed is high.

  For this problem, a technique has been proposed in which the decrease rate of the cwnd value when packet loss occurs and the increase rate at the time of recovery are determined according to the original congestion window size (cwnd) value (for example, non- Patent Document 2). According to this, since the amount of cwnd reduction at the time of packet loss can be reduced and recovery can be accelerated, end-to-end throughput instability during network congestion can be prevented and high-speed communication can be prevented. The throughput can be quickly recovered.

  In addition, at the receiving terminal, based on the arrival interval of packet data, always estimate the end-to-end bottleneck available bandwidth, and when this packet loss is detected, send this estimated value to the transmitting side, A technique has been proposed in which the transmission-side terminal receiving the adjustment adjusts the data transmission speed (see, for example, Non-Patent Document 3). This technology also stabilizes end-to-end throughput during network congestion.

M. Allman, et al., "TCP Congestion Control," RFC2581, April 1999. S.Floyd, "HighSpeed TCP for Large Congestion Windows," RFC3649, Dec. 2003. Nadeem Aboobaker, David Chanady, Mario Gerla, M. Y. Sanadidi, "Streaming Media Congestion Control using Bandwidth Estimation", Proceedings of the 5th IFIP / IEEE International Conference on Management of Multimedia Networks and Services: Management of Multimedia on the Internet 2002.

  As described above, there are examples such as Non-Patent Documents 2 and 3 as techniques for improving the instability of end-to-end throughput and the difficulty of recovering throughput during network congestion. However, even with these, the speed at which the sending terminal can send data is essentially close to the end-to-end bottleneck available bandwidth at that time, and therefore end-to-end The above-described problem that the effective throughput is suppressed to a value almost equal to the end-to-end bottleneck available bandwidth at that time cannot be solved.

  Another problem that arises in computer networks using conventional routers is that there is only a small-capacity buffer inside, and traffic data that is subsequently input from the line must be quickly output to the output line. In other words, since the data of each flow that flows on the network can only be output to the line on the output side while maintaining the speed at the time of input to the router, the data output speed and the timing for starting transmission are determined by the network. The data transfer node cannot be flexibly controlled, and there is a problem that advanced services such as effective use of network resources and flexible scheduling of data transfer between end-to-end terminals cannot be realized.

  The present invention has been made in view of the above points, and can flexibly control the data output speed in accordance with surplus fluctuations in the external network. As a result, the network resources can be effectively used and end-to-end terminals can be used. It is an object of the present invention to provide a data transfer apparatus and method that enable flexible scheduling of data transfer between the two.

In order to solve the above problems, the present invention (Claim 1) is arranged inside a computer network based on a packet switching system such as an IP network, and is a router or data for managing packet transfer in the network layer in the OSI reference model. A data transfer apparatus including a switch that controls transfer of a link layer frame,
Storage means realized by a large capacity storage device such as a large capacity hard disk drive (HDD) or a large capacity solid state disk (SSD) inside,
Surplus bandwidth information receiving means for receiving surplus bandwidth information of each line of the external network from the surplus bandwidth measuring device outside the router at a constant time interval and outputting the received surplus bandwidth information;
For a plurality of flows output from the data transfer apparatus, based on the surplus bandwidth information of each line of the external network notified from the surplus bandwidth information receiving means at the above-mentioned fixed time interval, and the transmission path of each flow , Output speed calculation means having a function of determining and notifying the output speed of the data of each flow at any given time interval as described above,
An output speed control means for adjusting the speed at the time of outputting the data of each flow to the line at any time to the output speed of the data of each flow notified at the predetermined time interval from the output speed calculation means;
Have
Each flow data input from the input side line is temporarily stored in the storage means, and output to the output side line while being adjusted at regular intervals at the output speed adjusted by the speed control means. .

Further, the present invention (Claim 2) is arranged inside a computer network based on a packet switching method such as an IP network, and transfers a frame of a router or a data link layer that controls a packet of a network layer in an OSI reference model. A data transfer method in a data transfer device including a switch to control,
Storage means realized by a large-capacity storage device such as a large-capacity hard disk drive (HDD) or large-capacity solid state disk (SSD) inside, surplus bandwidth information reception means, output speed calculation means, output speed control Means and a device comprising:
The surplus bandwidth information receiving means receives surplus bandwidth information of each line of the external network from a surplus bandwidth measuring device outside the router at a constant time interval, and notifies the output speed calculation means of the received surplus bandwidth information. An information receiving step;
The output speed calculation means, for a plurality of flows output from the data transfer device, surplus bandwidth information of each line of the external network that is notified from the surplus bandwidth information receiving means at the predetermined time interval, and each An output speed calculation step for determining the output speed of the data of each flow at any given time interval based on the transmission path of the flow, and notifying the output speed control means;
The output speed at which the output speed control means adjusts the speed at which the data of each flow is output to the line at any time to the output speed of the data of each flow notified from the output speed calculation means at the predetermined time interval. Control steps;
And
Data of each flow input from the input side line is temporarily stored in the storage means, and is output to the output side line while being adjusted at regular intervals at the output speed adjusted in the output speed control step.

  As described above, according to the present invention, the data transfer device temporarily stores the input data in the large-capacity storage means in the device, and controls the transmission rate according to the surplus bandwidth, so that the surplus fluctuation of the external network Accordingly, the data output speed can be flexibly controlled in accordance with the above, and as a result, effective utilization of network resources and flexible scheduling of data transfer between end-to-end terminals are possible.

It is a block diagram of a router in an embodiment of the present invention. It is a figure for demonstrating the relationship between surplus zone | band information and the output speed of each flow. It is a figure for demonstrating the function of the storage part in one embodiment of this invention. It is a figure for demonstrating the function of the surplus zone | band information receiving part in one embodiment of this invention, an output speed calculation part, and an output speed control part. It is an example of the network configuration and available bandwidth transition in one embodiment of the present invention. It is a block diagram of the router of one Example of this invention. It is a figure for demonstrating the operation | movement at the time of the traffic data input from the line | wire in one Example of this invention. It is a figure for demonstrating the operation | movement at the time of the traffic data output to the circuit | line in one Example of this invention. It is a standard block diagram of a conventional router or switch. It is FIG. (1) for demonstrating the conventional subject. It is FIG. (2) for demonstrating the conventional subject.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 shows a configuration of a router according to an embodiment of the present invention.

  The router 200 shown in the figure includes a control unit 210, a line module 220, a switch unit 230, and a storage unit 240. The line module 220 includes a line interface unit 221, a data transfer control unit 222, and a buffer unit 223.

  In the above configuration, the present invention includes the storage unit 240. This is because the above-described problem is that traffic that requires a bandwidth greater than the physical line capacity of the output line flows into the input line of the router. In this case, the existing router having only a small-capacity buffer inside is easily caused by packet loss, and is used to avoid this phenomenon.

  The storage unit 240 is not a memory such as an expensive and small-capacity DRAM such as a buffer of a conventional router, but a large-capacity storage device such as a hard disk drive (HDD) or a solid state disk (SSD). is there. In order to improve the speed by parallel processing, a plurality of the storage units 240 may be mounted.

  Further, according to the present invention, the line interface unit 221 includes the surplus bandwidth information receiving unit 2211, the data transfer control unit 222 includes the output rate control unit 2221, and the control unit 210 includes the output rate calculation unit 211.

  As shown in FIG. 2, the surplus bandwidth information receiving unit 2211 receives surplus bandwidth information of each line of the external network from the surplus bandwidth measuring device 400 of the external network at a constant time interval, and outputs the received value, which will be described later. It has a function of notifying the speed calculation unit 211.

  The surplus bandwidth measuring device 400 is a device having a function of measuring a surplus bandwidth of a line in an external network composed of a conventional router other than the router according to the present invention. When a network is constructed by combining a conventional router that does not have a large-capacity storage device inside and a router according to the present invention, in the network portion constructed by the conventional router that connects the routers according to the present invention, In order to avoid packet loss due to excessive inflow of traffic, or to realize advanced services that flexibly control the data transfer rate and transmission timing according to the usage status of the network resources of each part, Therefore, it is necessary to measure the surplus bandwidth and adjust the data transfer rate within the range. The surplus bandwidth measuring device 400 is a device that performs this surplus bandwidth measurement. The surplus bandwidth measuring device 400 may be diverted from the existing technology, and the details are omitted because it is not related to the gist of the present invention. Further, the format of the surplus bandwidth information notified to the surplus bandwidth information receiving unit 2211 is not particularly specified.

  The output speed calculation unit 211 of the control unit 210 includes a transmission path of each flow output from the router 200 of the present invention, and each line of the external network notified from the surplus bandwidth information reception unit 2211 at the predetermined time interval. The data output speed at the time of output to the line of each flow is calculated and determined at any given time interval based on the surplus bandwidth information.

  In order to facilitate understanding, an example of output speed adjustment of flow data using surplus bandwidth information will be described with reference to FIG. In the figure, routers 200a, 200b, 200c, and 200d of the present invention having the configuration shown in FIG. 1 are connected via a network constituted by routers 100a, 100b, 100c, and 100d based on the prior art. Here, a control example for preventing packet loss due to an inflow of an excessive amount of traffic in a portion configured by the router 100 based on the prior art will be shown.

  In a state in which data of flow a destined for the router 200c is flowing from the router 200a via the lines 300a, 300b, and 300c, the router 200b is destined for the router 200c via the lines 300d, 300e, 300b, and 300c. It is assumed that data of the flow b to be newly flowed. In this case, the router 200b is based on the surplus bandwidth information notified from the surplus bandwidth measuring devices 400b, 400e, 400c, and 400g of the lines 300d, 300e, 300b, and 300c that are the transmission paths of the flow b. In all the lines on the transmission path, the data is output while controlling the data output speed of the flow b so that the traffic amount exceeding the surplus bandwidth does not always flow.

  Specifically, if the data output speed of the flow b is always controlled to be equal to or less than the end-to-end bottleneck available bandwidth from the router 200b to the router 200c, the two flows a and b are used in common. In each of the connected lines 300b and 300c, the sum of the bands used by the two flows can be always kept within the range of the physical line capacity, and packet loss can be prevented.

  The details of the method of deriving the transmission path of each flow and the details of the final determination method of the output speed of the data of each flow are not related to the gist of the present invention, and thus the details are omitted.

  Next, the output speed control unit 2221 of the data transfer control unit 222 has a function of adjusting the actual sending speed when sending the data of each flow to the output side line. The data transfer control unit is controlled at the predetermined time interval so that the data transmission speed to the output side line of each flow becomes the value notified from the output speed calculation unit 211 at the predetermined time interval. .

  In the router 200 of the present invention having the above-described configuration, all data input from the input side line is temporarily written in the storage unit 240 and then output to the output side line.

  When data is output to the output side line, the output speed of each flow is controlled so as to be the value determined by the output speed calculation unit 211 for each predetermined time interval, regardless of the data input speed on the input line side. While outputting the data.

  In the present invention, processing for temporarily storing the data of the input traffic enters the storage unit 240. For this reason, as shown in FIG. 3, the sum of the traffic volume of flows that are input via one or more input lines and whose output destination is a specific line can be output from those output destination lines. A state in which the data rate continues to be greater than the data rate (determined by the physical line capacity of the output line or the surplus bandwidth value notified from the surplus bandwidth measuring device 400 of the external network connected to the output line) continues. In addition, data corresponding to the speed difference between the two can be stored in the storage unit 240.

  Also, as shown in FIG. 4, when the surplus bandwidth information receiving unit 2211 detects that the surplus bandwidth of the external network on the output side has increased, the output speed is increased if a sufficient amount of data is accumulated in the accumulating unit 240. The calculation unit 211 increases the output speed of the data of the flow, and the output speed control unit 2221 of the data transfer control unit 222 outputs the data of the flow from the storage unit 240 to the line at a speed reflecting this. Can be controlled.

  These functions eliminate the problem described in the section of the problem to be solved by the above-mentioned invention that the data transmission speed is governed by the end-to-end bottleneck usable bandwidth between the transmitting and receiving terminals, and the data between the transmitting and receiving terminals The first effect is that it becomes possible to send the message earlier.

  A specific example of this effect will be described with reference to FIG. The upper part of the figure shows the network configuration used for explanation. A data transmitting terminal and a receiving terminal are connected via routers A, B, C, and D according to the present invention. The routers according to the present invention are connected to each other via networks a, b, and c, which are conventional routers.

  In the lower part of the figure, a graph showing the transition of the end-to-end bottleneck usable bandwidth between routers according to the present invention connected to both ends of the networks a, b, and c is shown.

  In the network shown in the figure, packet loss does not occur as long as control is performed so that data exceeding the bottleneck usable bandwidth does not flow in each network portion composed of a conventional router that does not have a storage unit inside. This effect will be described by taking as an example the case of transmitting 30 Gbits data from the transmitting terminal to the receiving terminal.

As in the prior art, when the data transmission speed from the transmission terminal is always transmitted according to the end-to-end bottleneck bandwidth of the data transfer path from the transmission terminal to the reception terminal, the transmission of the 30 Gbits data is completed. The time it takes until the end-to-end bottleneck bandwidth from the sending terminal to the receiving terminal
0 to 5 minutes: 3Gbps
5 to 10 minutes: 5Gbps
So, the first 5 minutes can be transmitted at 3Gbps, and the remaining data 15Gbits can be transmitted in the next 5 minutes.
5 + (30-5 x 3) / 5 = 8 (minutes)
It becomes.

On the other hand, using the configuration shown in FIG. 1 of the router according to the present invention, while constantly observing the local bottleneck available bandwidth of each network portion composed of conventional routers, On the other hand, the data transmission speed is adjusted so that the data does not flow at a speed higher than the above available bandwidth, and the data input at a higher speed from the input side network is stored in the storage section 240. In this way, when the data transmission rate is controlled so as to satisfy only the condition that data exceeding the local bottleneck available bandwidth does not flow to the networks a, b, and c,
・ Router A sends data for 6 minutes at 5Gbps:
・ Router B sends data at 3Gbps for the first 5 minutes, then sends 10Gbits of data stored in its storage 240 for the first 5 minutes (at a speed of 10Gbps) for the first 5 minutes, Send the remaining 5Gbits data (at 10Gbps speed) in the last 0.5 minutes:
For this reason, the time required to complete the data transfer is
5 + 1 + 0.5 = 6.5 (minutes)
Thus, the time required to complete the data transfer can be shortened compared to the former case.

  For simplicity, data transmission delay is not considered in both cases. In addition, when a protocol having window control such as TCP is used as the transport layer protocol, the data transmission rate control function by the transport layer itself works, so that the effect of the present invention can be directly enjoyed. However, since it is not the gist of the present invention, the protocol used for the transport layer is not defined here.

  Furthermore, according to the present invention, the timing and speed at which the data stored in the storage unit 240 is transmitted in each router can be freely adjusted while taking into account the available bandwidth information of the external network. For this reason, it is possible to realize a flexible service such as transferring a large amount of data with loose delivery time constraints at high speed when the network usage rate is low by looking at the network usage status. There is also the effect.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

  FIG. 6 shows the configuration of a router according to an embodiment of the present invention.

The router 200 shown in the figure includes N line modules 220 _{1 to} 220 _N , a control unit 210, M storage units 240 ₁ to 240 _M , and a switch unit 230.

Each of the N line modules 220 _{1 to} 220 _N includes a data transfer control unit 222, a line interface unit 221, and a buffer unit 223.

  The line interface unit 221 includes a surplus bandwidth information receiving unit 2211 and a packet processing unit 2212. The line interface unit 221 accommodates one line, and the packet processing unit 2212 performs data input / output via the line in units of packets.

  The surplus bandwidth information receiving unit 2211 receives information on the surplus bandwidth of the external network to which the line is connected from an external surplus bandwidth measuring device at regular time intervals, and every time the value is received, the surplus bandwidth information receiving unit 2211 Notify the output speed calculator 211.

  The buffer unit 223 absorbs the difference between the data transmission / reception speed of the line and the data transfer speed of the storage unit 240, and prevents data loss in the router. The buffer unit 223 is a high-speed and small-capacity storage device composed of a DRAM or the like. Each traffic flow is divided into areas, and each area is used as a FIFO.

  The data transfer control unit 222 controls data transfer between the line interface unit 221 and the buffer unit 223 and between the buffer unit 223 and the storage unit 240.

  The output speed control unit 2221 in the data transfer control unit 222 has a function of controlling the actual output speed when sending data of each flow to the output side line. Each time the output speed value of each flow is notified from the output speed calculation unit 211 of the control unit 210 described later at a constant time interval, the data of each flow is transferred from the buffer unit to the line interface unit 221 at that rate.

  The above are the components included in each line module 220.

  The control unit 210 is composed of a general-purpose microprocessor, and exchanges various types of information with other components, controls them, and performs various arithmetic processes.

The output speed calculation unit 211 of the control unit 210 is software that realizes arithmetic processing executed by the control unit 210. For a plurality of flows output from this router, based on the surplus bandwidth information of each line of the external network notified from the surplus bandwidth information receiving unit 2211 at the predetermined time interval and the transmission path of each flow, The storage unit 240 has a function of determining the output speed of the data of each flow at any given time interval as described above and notifying the output speed control unit, such as a hard disk drive (HDD) or a solid state disk (SSD). It is a storage device of capacity. Traffic data input from the line is temporarily stored until it is output to the output side line.

  The switch unit 230 switches the transfer of traffic data between each line module and each storage unit.

  These components are connected to each other by a control signal line and a data line as shown in FIG. A data line is a path through which traffic data flows, and a control signal line is a signal line that transmits and receives control signals and various information between the components.

  Among the constituent elements described above, the surplus bandwidth information receiving unit 2211, the output speed control unit 2221, the output speed calculation unit 211, and the storage unit 240 indicated by bold lines in the drawing realize the functions unique to the present invention.

  Next, the operation of the router of this embodiment will be described.

  FIG. 7 is a diagram for explaining the operation at the time of traffic data input from the line in the embodiment of the present invention. FIG. 8 shows the operation at the time of traffic data output to the line in the embodiment of the present invention. It is a figure for demonstrating.

  Hereinafter, the processing until the data of the flow (a) input from the line A is output from the line module 220B will be described with reference to FIGS.

  First, the operation at the time of traffic data input from the line will be described with reference to FIG.

  The packet processing unit 2212A of the line module 220A to which packet data is input from the line A identifies a flow consisting of a set of packets (source IP address, destination IP address, source port number, destination port number). If it is a packet of a new flow, information relating to the flow (a) as “new flow information” to the data transfer control unit 222 on the input side line module 220A and the output side line module 220B. Is transmitted ((1) in the figure).

  The data transfer control unit 222A that has received this information of the line module 220A generates a new FIFO queue ((2) in the figure) for storing the data of the flow (a) in the buffer unit 223A and A new entry for the flow (a) is generated for the “flow / queue map” which is a correspondence table between the flow and the FIFO queue in the buffer ((3) in the figure).

  With the processing so far, the operation of storing the packet data of the flow (a) successively input from the line in the FIFO queue for the flow (a) in the buffer can be started ((4) in the figure). ). Note that the packet data of the flow other than the flow (a) (indicated as flow (b) in the figure) input from the same line A is for the flow (a) in the buffer unit 223A as shown in the figure. Is stored in a separate FIFO queue.

  In parallel with the operation (4) in the figure, the data transfer control unit 222A transmits the new flow information to the control unit 210 ((5) in the figure).

Upon receiving this information, the control unit 210 newly adds an entry for the flow (a) to the list of flows passing through the router 200 ((6) in the figure), and then stores the flow ( A data storage file of a) is generated ((7) in the figure). In the present invention, the method for associating the flow with the storage unit 240 for storing the data is not particularly defined because it is not the gist of the invention. However, in the present embodiment, in order to speed the transfer of data from the buffer section 223A to the storage unit 240, data storage file of the flow to all of the storage portion of the storage portion 240 ₁ ~240 _M (a) When data is transferred from the buffer unit 223A to the storage unit 240, the data of the flow (a) is striped and written in parallel to these storage units. The data of the flow other than the flow (a) input from the same line (flow (b) in the figure) is stored in a file different from that for the flow (a) as shown in the figure. .

  When the generation of the file is completed, the location information (storage unit number) of the generated file is transmitted to the data transfer control unit 222 on the input side line module 220A and the output side line module 220B of the flow (a). , And the position information of the file in the storage unit) ((8) in the figure).

The data transfer control unit 222A on the input side line module 220A that has received this information stores the data of the flow (a) stored in the FIFO queue for the flow (a) of the buffer unit 223A. An operation of writing to the file in 240 is started ((9) in the figure). As described above, data in the buffer unit 223A is written in parallel to the write files generated in all the storage units 240 ₁ to 240 _M , thereby transferring data from the buffer unit 223A to the storage unit 240. Speed up.

  This file write operation does not necessarily need to be performed in units of packets, and may be performed in larger units in order to improve write efficiency. In FIG. 7, the unit is described as chunk. However, since this device is a router that handles packet transfer, the packet boundary and packet header information of the flow (a) can be held and handled until the data output to the output side line is completed. The data write operation to the FIFO queue of the buffer unit 223A and the data write operation to the file of the storage unit 240 must be performed in a data format that can hold such information. However, the details of this format are not the gist of the present invention, and will be omitted.

  The above is the description of the operation when the traffic data is input from the line. Next, the operation at the time of outputting traffic data to the line will be described with reference to FIG.

  The data transfer control unit 222222B on the output side line module 220B that has received the above-mentioned new flow information ((1) in the figure) from the packet processing unit 2212A on the input side line module 220A is first similar to that on the input side. In addition, a new FIFO queue for storing the data of the flow (a) is generated ((2) in the figure) in the buffer unit 223B, and the flow and the FIFO queue in the buffer that are included in the buffer unit 223B are also generated. A new entry for the flow (a) is generated for the “flow / queue map” as the correspondence table ((3) in the figure). Thereafter, it waits until the location information of the file opened as the data storage file of the flow (a) in the storage unit 240 is notified from the control unit 210.

When this information is notified from the control unit 210 ((8) in the figure), the data of the flow (a) written in the file of the storage unit 240 is used as the FIFO for the flow (a) in the buffer unit 223B. The operation of storing in the queue can be started (FIG. 9). By transferring the data of the flow (a) written across all the storage units 240 ₁ to 240 _M at the time of data input to the buffer unit 223B in parallel, the data transfer of this part is speeded up. . The data transfer operation between the buffer unit 223B and the storage unit 240 is performed in the same data unit (chunk) as that on the input side. Note that the data of the flow (b) in the figure is also stored in a FIFO queue different from that for the flow (a) in the buffer unit as shown in the figure.

  When the data of the flow (a) starts to be stored in the buffer unit 223B through the processing so far, it is ready to output the data of the flow (a) to the output line B.

  Next, processing until the output of the data to the output line B is started will be described.

  The surplus bandwidth information of each line of the external network connected to the line B transmitted from the surplus band measuring device installed outside the router 200 via the line B is received ((10) in the figure). The surplus bandwidth information receiving unit 2211B transmits this surplus bandwidth information to the control unit 210 ((11) in the figure).

  The control unit 210 that has received the surplus bandwidth information uses the calculation function of the output speed calculation unit 211 while referring to the list of flows that pass through the router 200 ((12) in the figure), Based on the excess bandwidth information of the line and the transmission path of the flow (a), the output speed of the flow (a) to the output side line is determined. A specific method for determining the output speed is omitted because it is not the gist of the present invention.

  When the output speed of the flow (a) is determined, the output speed calculator 211 of the controller 210 transmits the information to the output speed controller 2221B of the data transfer controller 222b of the line module 210B ((14 in the figure)). )).

  Upon receiving this, the output speed control unit 2221B receives the packet data of the flow (a) from the FIFO queue for the flow (a) at the speed notified from the output speed calculation unit 211, and performs packet processing of the output side line interface unit 221B. The process of transferring to the unit 2212B is started ((15) in the figure).

  The packet processing unit 2212B performs IP header processing of the packet data transferred from the buffer unit 223B and outputs the packet data to the line B. By performing this processing at a rate equal to the rate at which data is transferred from the buffer unit 223B, the data of the flow (a) is transferred in the output band assigned to the flow (a) by the control unit 210. It is output to the line B.

  Thereafter, the surplus bandwidth information receiving unit 2211B, the output rate calculating unit 211 of the control unit 210, and the output rate control unit 2221B of the data transfer control unit 222B perform the processes (10) to (15) in the figure at regular time intervals. Repeatedly, each time, by adjusting the transfer rate of the data of the flow (a) from the buffer unit 223B to the packet processing unit 2212B, the output corresponding to the change in the surplus bandwidth of the external network connected to the line B Realize speed control.

  The present invention is not limited to the description of the above embodiments and examples, and various modifications and applications can be made within the scope of the claims.

100 router 110 control unit 120 line module 121 line interface 122 data transfer control unit 123 buffer 130 switch unit 200 router 210 control unit 211 output speed calculation unit 220 line module 221 line interface unit 222 data transfer unit 223 buffer unit 230 switch unit 240 accumulation Unit 300 line 400 surplus bandwidth measuring device 2211 surplus bandwidth information receiving unit 2212 packet processing unit 2221 output rate control unit

Claims (2)

A router that is located inside a computer network based on a packet switching method such as an IP (Internet Protocol) network, and that is responsible for forwarding network layer packets in the OSI (Open Systems Interconnection) reference model, or a switch that forwards data link layer frames A data transfer device including:
Storage means realized by a large capacity storage device such as a large capacity hard disk drive (HDD) or a large capacity solid state disk (SSD) inside,
Surplus bandwidth information receiving means for receiving surplus bandwidth information of each line of the external network from the surplus bandwidth measuring device outside the router at a constant time interval and outputting the received surplus bandwidth information;
For a plurality of flows output from the data transfer apparatus, based on the surplus bandwidth information of each line of the external network notified from the surplus bandwidth information receiving means at the above-mentioned fixed time interval, and the transmission path of each flow , Output speed calculation means having a function of determining and notifying the output speed of the data of each flow at any given time interval as described above,
An output speed control means for adjusting the speed at the time of outputting the data of each flow to the line at any time to the output speed of the data of each flow notified at the predetermined time interval from the output speed calculation means;
Have
Each flow data input from the input side line is temporarily stored in the storage means, and output to the output side line while being adjusted at regular intervals at the output speed adjusted by the speed control means. A data transfer device. A router that is located inside a computer network based on a packet switching method such as an IP (Internet Protocol) network, and that is responsible for forwarding network layer packets in the OSI (Open Systems Interconnection) reference model, or a switch that forwards data link layer frames A data transfer method in a data transfer device including:
Storage means realized by a large-capacity storage device such as a large-capacity hard disk drive (HDD) or large-capacity solid state disk (SSD) inside, surplus bandwidth information reception means, output speed calculation means, output speed control Means and a device comprising:
The surplus bandwidth information receiving means receives surplus bandwidth information of each line of the external network from a surplus bandwidth measuring device outside the router at a constant time interval, and notifies the output speed calculation means of the received surplus bandwidth information. An information receiving step;
The output speed calculation means, for a plurality of flows output from the data transfer device, surplus bandwidth information of each line of the external network that is notified from the surplus bandwidth information receiving means at the predetermined time interval, and each An output speed calculation step for determining the output speed of the data of each flow at any given time interval based on the transmission path of the flow, and notifying the output speed control means;
The output speed at which the output speed control means adjusts the speed at which the data of each flow is output to the line at any time to the output speed of the data of each flow notified from the output speed calculation means at the predetermined time interval. Control steps;
And
The data of each flow input from the input side line is temporarily stored in the storage means, and output to the output side line while being adjusted at regular intervals at the output speed adjusted in the output speed control step. A featured data transfer method.

Images