JP5560235B2 - Arrival packet synchronization method - Google Patents

Description

  The present technology relates to a method of synchronizing arrival packets regardless of a physical distance between a client and a server regarding a function of distributing the same packet to a plurality of clients in a client server type network, and a network system including the function.

  With the spread of cloud computing, data center operators are required to provide a wide variety of services to end users. The form of the network in the data center is often server-client communication, and it can be said that there is an advantage of cloud computing in that the client can receive an equivalent service regardless of the installation location. In addition, the types of packets flowing on the network have undergone various changes from the conventional one-to-one communication to one-to-n and n-to-n. In one-to-n communication typified by multicast communication, for example, in a service that distributes information from a server to each client, the distributed packet is caused by a delay or a transmission delay in a network device arranged in the transfer network. It does not necessarily arrive at all clients at the same time.

Special table 2010-502123 Japanese Patent No. 037105098

  With the spread of cloud computing, server-client communication in the data center business is likely to continue to increase. For example, in a situation where there are a plurality of clients connected to a single server, the distributed packet is transferred in a service for distributing information from the server to each client by 1-to-n communication represented by multicast communication. Due to delays and transmission delays in the network devices deployed in the network, it does not always arrive at all clients at the same time, and delays may occur due to physical distance and data transmission speed problems. In industries such as securities and finance, the value of information transmission time is increasing.

  In the technique of Patent Document 1, in order to synchronize packet information in a distribution network, a transmission time indicator is incorporated in the packet information, and a network device having a function of holding the packet information until a time offset after the transmission time indicator is used. A method of synchronizing packet information on the receiving side is described. However, in this method, the same mechanism had to be introduced on both the distributing side and the distributing side. Another problem is that information such as a transmission time indicator has to be added to the packet information itself, and there is a possibility that the bandwidth will be tightened.

  An object of the present invention is to provide a mechanism for arriving at the same time regardless of the physical distance and transmission speed.

  In order to achieve the object of the present invention, the network system of the present invention calculates a delay time to a packet destination in advance in one-to-n communication represented by multicast communication, and considers the calculated delay time. In order to maintain the fairness of the network, a device having a function of synchronizing packet arrival times is provided. With this function, regardless of the physical distance, packets sent to multiple clients from a server-equivalent device can be transmitted on the transmission side in consideration of transmission delays, and the incoming packets can be synchronized by shifting the packet transmission timing. I am letting. A network device represented by a distribution server, a multicast packet sending router, etc. is connected, and in a network that is sent to multiple destinations, a packet synchronization device having the above functions is connected, a transmission delay is calculated in advance, and a packet is sent By shifting the timing, packets arrive at the same time regardless of the location of the base.

  Further, by adjusting the transmission timing (inter-transmission interval) of each packet based on the error of the packet delivery delay time, control is performed so that the reception times of all packets are the same regardless of the packet delivery delay time. .

  According to the present invention, simultaneous arrival of packets can be realized for all subordinate users in multicast communication, broadcast communication, and unicast communication with a plurality of destinations provided by various servers provided in cloud computing. Regardless of the physical location, the time fairness of information transmission can be guaranteed. Providing the services mentioned above is expected to be a service that is highly preferential for industries such as finance and securities, where the fairness of information transmission time is directly linked to the profits and disadvantages of companies and individuals. In addition, it is very compatible with recent cloud computing in that infrastructure can be constructed without considering the physical location, and solves the fairness of network infrastructure that will continue to be an issue in the future. It can also be expected to be a solution.

It is a figure which shows an example of the system containing a packet synchronizer. It is a figure which shows an example of a structure in a packet synchronizer. It is a sequence diagram which shows the calculation process of the delay time with respect to a certain client. It is a figure which shows an example of the delay time information table in a packet synchronizer. It is a figure which shows an example of the packet for delay time calculation exchanged between a packet synchronizer and a client. It is a figure which shows the relationship between the departure time from a server, and the arrival time to a client. It is a figure which shows the correction | amendment of the delay by the error of the delay time with respect to each client, and transmission time (DELTA) of all the data.

  Embodiments will be described below with reference to the drawings.

  In this embodiment, the server 101 that sends packets to a plurality of destinations has a function of simultaneously arriving packets to the client A 103, the client B 104, and the client C 105 in the network that sends the packets to the client A 103, the client B 104, and the client C 105. An example of the packet synchronization apparatus 102 will be described. FIG. 1 is an example of a system configuration diagram of a client-server system including the packet synchronization apparatus 102 of the present embodiment. In the figure, 101 is a server that distributes various information to each client, and 102 is a copy of the packet that causes the information distributed from the server 101 to arrive at each client at the same time, and considers delay information for each client described later. In addition, a packet synchronization apparatus having a function of copying and sending packets, 103 is a client A located at a medium distance from the server 101, and 104 is a client B located at a short distance from the server 101. , 105 is a client C that is located at a long distance from the server 101.

  FIG. 2 is an example of a configuration diagram of a packet synchronization apparatus that implements the present embodiment. The packet synchronization apparatus shown in FIG. 2 is configured as follows. In FIG. 2, 201 is a delay information control unit that controls the delay information for each client to shift the packet transmission timing calculated by the method described later, and 202 holds a packet delay information table and a multi-destination packet storage table. , 203 is a packet delay information table holding packet delay information, 204 is a multi-destination packet storage table storing each destination of packets whose transmission is shifted, 205 is a CPU for calculating delay time, 206 is A multi-destination packet storage unit (a kind of data buffer) that stores packets for which transmission is shifted for each destination, and 207 is an input / output device that is an interface part with an external device.

  In the system configured in FIG. 1, a server 101 is connected to a client A 103, a client B 104, and a client C 105 via a certain network. At this time, the network does not depend on a protocol such as VPN (Virtual Private Network) regardless of L2 / L3 (L2 and L3 are network switches). Moreover, although it is desirable that they are networks of the same carrier, communication via other carriers may be used. Here, it is assumed that each client A103, client B104, and client C105 is a switch or router that is aggregated in the previous stage even in a company or individual PC terminal. In this system, the packet synchronization apparatus 102 is connected under the server, and calculates the transmission delay via the client and the network connected under the server. Hereinafter, a transmission delay calculation method and a mechanism for simultaneously arriving at a client from packet transmission will be described with reference to the sequence diagrams of FIGS.

  When the server 101 sends a packet to each client A 103, client B 104, and client C 105, the input / output device 207 of the packet synchronization apparatus 102 recognizes that the packet received from the server 101 is for a plurality of addresses. The packet synchronizer 102 sends a packet in which the current time information that is the transmission time is applied to the packet shown in FIG. 5 to each of the client A 103, the client B 104, and the client C 105 that are the destinations of the received packet (step 301). . Each client A103, client B104, and client C105 that received the packet applies the time at which it received the packet to the packet and returns it to the packet synchronizer 102 (step 302). When the destination of this packet is unknown, the packet synchronizer 102 sends a multicast packet to the network 106 and waits for a reply from a client under its control and sends a packet to which its own time information is applied. Although this sequence may be performed at the timing when a plurality of destination packets arrive, destination confirmation may be performed in advance for all clients to which this packet may be sent. In addition, an administrator may set destinations that can be stored in a packet delay information table 203 described later, or may automatically set destinations.

  Each device (server 101, packet synchronizer 102, client A103, client B104, client C105) appearing in this embodiment is a network time protocol (NTP) or a high precision time protocol (PTP). ) Etc., and the same time synchronization is assumed in advance, and it is assumed that the time in each device is the same. The time reference is not limited to the above description, and time synchronization may be performed by any method.

  The packet synchronization apparatus 102 that has received packets from the clients A 103, B 104, and C 105 stores the received packet information in the delay time information table 401. The CPU 205 calculates a delay time 407 from the transmission time 405 applied by the packet synchronizer 102 in the stored delay time information table 401 and the reception time 406 applied by each client A 103, client B 104, and client C 105. The delay information control unit 201 refers to the delay time 407 in the delay time information table 401 and determines the transmission time of each packet with reference to the destination with the longest delay among the destinations to be sent in the packet. A method of determining the transmission delay time of each client packet for the client having the longest delay time will be described later. The corresponding packet is generated by the packet synchronizer 102 to be sent to each destination IP address without changing the payload (payload data body) 504 portion. The generated packet is sent to the client according to the value stored in the delay time 407 in the delay information table 401 (step 303).

Here, the procedure for obtaining the departure delay time of data transmission to each client will be described with reference to FIG. 6, the starting time T _S and arrival time T _A of the packet to each client packet from the _server, T _B, is a diagram showing the relationship of T _C, the vertical axis a distance from the server for each client L is shown, and the horizontal axis indicating the arrival time of each client is shown at a position corresponding to each distance.

  The inclined double arrows indicate that it takes a finite time to send all data from the server to the client. If the total amount of data to be sent is S (bit), and the transfer rate, which is the amount of data sent or received in unit time, is v (bit / sec), the time required for sending or receiving all data Δ (sec) Becomes S / v. However, v is the same for the server and each client.

  The distance L (km) of the transmission path from the server to each client and the data transmission speed P (km / sec) are different, and the data transmission time in the transmission path is L / P (sec). (The transmission rate P may be not only a physical transmission rate but also an apparent transmission rate in consideration of the degree of congestion in the transmission path.) The angle of inclination of the inclined double arrow shown in FIG. Then, the data transmission rate P is expressed as P = tan θ. That is, the higher the data transmission speed P, the larger the inclination angle θ of the inclined double arrow. FIG. 6 illustrates that the transmission speed of the transmission path from the server to each client is different.

The difference between the departure time T _S of the packet from the server and the arrival times T _A , T _B , and T _C of the packet to each client is the delay time T _K of each client k, and this value includes the distance from the server The influence of L and each transmission rate P is included. When the longest delay time T _m of the longest among the delay times T _K, after a delay time has sent the packet to the longest of the client, the delivery delay time tau _k when sending packets to other clients k It is given by τ _k = _T m -T _K. By delaying the departure time of the packet transmission from the server to each client by this transmission delay time, the arrival time of the packet to each client becomes the same.

  In the first embodiment, in accordance with the delay time 407 of the delay time information table 401, the destination IP address having the longest delay time is No. 3.1.1 with a destination IP address of 3.1.1.75. 1.0 seconds after sending the destination packet No. 3, the destination IP address is 20.5.5.101. 1 packet, and 0.5 seconds later, the destination IP address of 10.5.5.112 No. By sending two packets, it is possible to synchronize the arrival time to each client. That is, the packet arrival time can be made the same by delaying the packet transmission time by the time difference between the delay time and the longest delay time.

  If the packet arrives, the arrival notification packet is exchanged in the same manner as the previous time (steps 301 and 302), the arrival confirmation result is fed back, the delay time information table 401 is updated, and the delay time 407 for each destination is recalculated. Thus, the accuracy of the delay time 407 can be improved. Further, in order to confirm whether or not the actually distributed packet has arrived at the same time, the delay time may be calculated for the packet. The contents described in the delay time information table can be confirmed from an external terminal, can be manually confirmed by a network administrator, and can be automatically transmitted to the external terminal. Further, as described above, the transmission rate P apparently changes depending not only on the physical transmission rate but also on the degree of congestion of the transmission path, so the delay time is measured several times, and the average value is obtained. By using, the reliability of the delay time 407 is improved. In order to perform packet transmission according to the delay time 407, the packet is stored in the packet storage unit 206, and the packet is sent according to the instruction of the delay control instruction unit 201.

  Note that the method for controlling the sending of the packet may be in the form of scheduling, a method of accumulating in a queue, or the like. An event in which a failure occurs in the network between the client and the user by executing and confirming the above-described series of processing by the packet synchronization apparatus 102, and the packet cannot be synchronized if the delay time 407 for each destination set once is transmitted. Can be avoided.

  In the method of incorporating the transmission time indicator into the packet information described in Patent Document 1 and holding the packet information until the time offset after the transmission time indicator, it is necessary to set several stages of transmission time indicators from the source to the end user accommodation base. There is also a possibility that the receiving side must keep the packet unnecessarily. For this reason, many parameters are required for overall optimization for reducing the delay, and the processing is complicated. This method has the advantage that the delay situation at the end-end can be grasped and the system can be constructed without depending on the network. In the present embodiment, an example of a transmission delay calculation method and a packet synchronization method will be described without being limited to the packet synchronization method shown in the first embodiment.

  Description of the same reference numerals and configurations already described in the first embodiment is omitted. In the first embodiment, the delay time 407 in the packet delay information table 203 for each destination is calculated by applying the NTP time synchronization sequence between the client and the server, and the timing for sending the packet for each destination is determined. However, not only this system but also a system that realizes synchronization of a plurality of destination packets in the entire network by gradually delaying packets in all network devices located between the server and the client can be considered. At this time, as a method for calculating the delay time, as in this method, the time information is transmitted from a device equivalent to a certain packet synchronizer, and the packet is actually transmitted by returning the time information received by the opposite network device. However, the accuracy cannot be improved, but the delay time of each client can be fixedly determined by the sum of the distance and the route network device. Any method may be taken.

Above, description of the processing flow for synchronizing the timing which the packet sent from the server 101 arrives at each client A103, client B104, and client C105 via the packet synchronizer 102 and the network 106 is complete | finished. The packet synchronization apparatus 102 sends a time synchronization packet to each client A 103, client B 104, and client C 105, and repeats the delay information 407 to keep track of the delay situation so that multiple destination packets from the server 101 arrive at all clients simultaneously. It is possible.
(Correction of delay by error of delay time for each client and sending time Δ of all data)
In the above embodiment, it is assumed that there is no error in the value measured in advance for the delay time when the packet is transmitted from the server 101 to each client. However, the delay time depends on the state of the transmission path from the server to the client. Variations and delay time errors occur. It is considered that the variation in delay time of packet transmission is more easily affected by the condition of the transmission path as the length of the transmission path is longer and the transmission speed is lower, and the error in the delay time is considered to be larger. As a result, even if the departure time from the packet transmission server is delayed in advance as described above, the time for receiving the packet on each client side is not uniform.

  On the other hand, in FIG. 6, the time Δ required to send or receive all (packet) data is the same for each client, but the error in delay time and the sending time Δ for all data are as follows. In consideration, by further adjusting the delay time of the departure time from the packet transmission server, the time to receive the beginning and the end of all data can be made substantially the same at each client.

  The outline of the above assumption and delay correction is as follows.

  (1) For packets with a long delay time, the transmission time interval of each packet from the server is shortened. In this case, due to a large error in the delay time, there is a possibility that the time interval between the time for receiving the start of all data and the time for receiving the end on the client side may be widened.

  (2) For packets with a short delay time, the transmission time interval of each packet from the server is increased. In this case, since the error of the delay time is small, the time interval between the time for receiving the start of all data and the time for receiving the end on the client side is not so wide.

(3) such that the sum of twice the error sigma _k of the delay time and the transmission time delta _k of all data is the same for each client k, adjust the time interval of transmission of packets from the server to each client In addition, the delay time at the start of transmission from the packet transmission server described above is adjusted so that the time when the central packet of all data (packet string) is received matches. FIG. 7 is a diagram for explaining the states (1) and (2).

Based on FIG. 7, a procedure for correcting a packet transmission delay will be described.
(1) The delay time T _k is measured several times for each client (k), and the average value <T _k > and error (standard deviation) σ _k are obtained and stored in the delay time information table 401 of FIG. Considering that the situation of the transmission path changes according to the time zone, further holding the average value and error of the delay time for each time zone for each client, and correcting for the packet transmission delay according to the time zone, The accuracy of correction is improved.

Here, if there is a positive correlation between the average value <T _k > and the error σ _k for a plurality of clients, that is, if σ _k ∝ <T _k >, the correction described below uses <T _It may be based on σ _k of the client with the largest _k >.

Further, in the correction described below, the average value <T _k > and the error σ _k are treated as independent. However, when the relative error σ _k / <T _k > is large, it depends on the length of the transmission path distance. It shows that there are more uncertain factors such as fluctuations in the operation of network components such as routers in the middle of the transmission path than disturbance factors. In such a case, when the data is actually received, there is a possibility that the fluctuation of the time required for receiving all the data on the client side becomes large.
(2) Using <T _m > of the client (M) having the largest average delay time <T _k > as a reference, as shown in FIG. 6, <T _m > and <T of each client k _From the difference from _k >, the departure delay time [tau] _k = < _Tm >-< _Tk > is _obtained for the central packet of all data. The diagonally dashed arrows in FIG. 7 represent the delay with respect to the central packet of all data, and each client indicates that the reception time is the same for the central packet of all data as in FIG. Show.
(3) If the total number of packets is n and the transmission (time) interval of each packet from the server to the client M is δ _m (δ _m is the minimum value of the packet transmission interval), all data is transmitted from the server to the client M. time delta _m comprises _{Δ m = (n-1)} δ m , and the delay time error sigma _m, required reception time _{W m} of all the data (packet) on the client M is a _{W m} = _{Δ m} + _{2σ m} Become. Similarly, for other clients k, Δ _k = (n−1) δ _k and W _k = Δ _k + 2σ _k . In general, the longer the delay time T _k is, the larger the error σ _k is (σ _m > σ _k ). However, in the correction for the packet transmission delay, the error σ _{k of the} delay time is not related to the length of the delay time. The maximum value may be σ _m .
(4) W _k = Δ _k + 2σ _k = (n−1) δ _k + 2σ _k is expressed as W _m so that the time required for receiving all data (all packets) is the same in the client M and each client k. _{= Δ m + 2σ m = (} n-1) from the condition that is the same as the δ _m + 2σ _m, the transmission interval [delta] _k of packets from the server for each client _{k, δ k = δ m +2} (σ m -σ k) / (N-1). Here, since σ _m > σ _k , δ _k > δ _m , and the smaller the delay time error, the longer the packet transmission interval, the statistical reception of all data on each client side It is expected that the required time will be the same, and the reception time of the first packet and the reception time of the last packet of all data (all packets) will be the same.
(5) The correction amount for the departure delay time τ _k of data from the server to each client described with reference to FIG. 6 is −Δ _k / 2, and the departure delay time for the first packet of all data is τ _k → It is corrected to τ _k −Δ _k / 2. Here, the correction amount of the departure delay time τ _m for the reference client M is −Δ _m / 2, and the correction amount of the departure delay time τ k for the other clients k is −Δ _k / 2 = − (Δ _m / 2+ (σ _m −σ _k )). However, it is (DELTA) _k > (DELTA) _m .

  As described above, the departure delay time of the packet and the transmission interval of each packet are corrected for each client based on the error of the delay time.

  The above correction increases the possibility (probability) that the time for receiving the start of all data and the time for receiving the end on the client side are almost the same. A sense of fairness is reduced.

  According to the present embodiment, a packet sent to a plurality of destinations in a certain network has a function of arriving at each client at the same time regardless of the physical distance of the base, so that the time of the packet There is an effect of ensuring fairness and enabling communication carriers to provide more attractive services.

101 Server 102 Packet Synchronizer 103 Client A
104 Client B
105 Client C
201 Delay Information Control Unit 202 Memory 203 Packet Delay Information Table 204 Packet Storage Table 205 CPU
206 Packet storage unit 207 I / O device

Claims (3)

An arrival packet synchronization method in a network system in which a server and a plurality of clients are connected via a network, and the same data consisting of a plurality of packets is transmitted from the server to each of the plurality of clients. ,
When the transmission time when data is transmitted a plurality of times to each of the plurality of clients and the average value of the delay time calculated from the reception time when each of the plurality of clients receives the data and when the data is transmitted a plurality of times And an error of the delay time in correspondence with the plurality of clients,
Based on the average value of the calculated delay times, obtain a packet departure delay time for each of the plurality of clients for each of the plurality of clients to receive the packet at a predetermined time;
The time required for the client having the maximum error to receive the plurality of packets based on the number of the plurality of packets, the minimum value of each packet transmission interval, and the error for the client having the maximum error. Seeking
For each client other than the client having the maximum error, obtain each packet transmission interval based on the required time and the delay time error of the client,
Based on the determined packet transmission interval, the departure delay time of the packet is corrected,
For the plurality of clients, the corrected departure delay time from a transmission time for the client having the maximum average delay time to receive a packet at the center of the plurality of packets at the predetermined time. An arrival packet synchronization method, wherein transmission of the plurality of packets is started by delaying a transmission time by a predetermined amount. When there is a positive correlation between the average value of the delay time and the error for each of the plurality of clients, the required time is obtained for the client having the maximum delay time. The arrival packet synchronization method according to claim 1 . For each of the plurality of clients, the average value of the delay time obtained for each time zone and the error are held, and the departure delay time of the packet is corrected according to the time zone. The arrival packet synchronization method according to claim 1 .

Images