JP2013150291A - Device, method, and program for measuring the number of packet loss - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve measurement accuracy in the number of packet loss.SOLUTION: The device for measuring the number of packet loss detects a flow which flows between a network A and a network B and a direction in which a reception packet in the flow flows, stores information on the reception packet in flow storage means, extracts reception data packet information and ACK packet information from the reception packet, determines that a packet loss may be generated in a portion between a measurement point and a TCP receiving terminal when the reception data packet information is a retransmission packet and a timeout occurs, and turns on a time-out flag of flow recording means. Moreover, the device refers to the flow storage means on the basis of the ACK packet information, and adds the number of the packet loss when the time-out flag indicating a situation of the packet loss between the measurement point and the TCP receiving terminal is turned on.

Description

  The present invention relates to a packet loss number measuring apparatus, method, and program, and more particularly to a packet loss for measuring the number of packet losses in an IP (Internet Protocol) network that performs communication using TCP (Transmission Control Protocol). The present invention relates to a number measuring apparatus, method, and program.

  Methods for measuring the packet loss rate of an IP network can be broadly divided into two types: passive type and active type. The passive type is a method of measuring from information of packets flowing through the network, and the active type is a method of measuring by flowing test packets through the network.

  Various methods have been proposed for active and passive measurements, respectively. One of the passive measurement methods is to estimate the packet loss rate of the network based on the TCP sequence number information. There is. TCP is a bi-directional communication protocol, and consists of a data packet and an ACK packet for confirmation of reception. As a passive measurement method, focusing on TCP data packets, monitoring the sequence number of the data packet, detecting duplication of sequence numbers, skipping of the number, and original correction (the retransmitted packet is used as the parameter of the packet loss count). There is a method of estimating the packet loss rate of the network (TCP transmission terminal to measurement point, measurement point to TCP reception terminal) before and after the measurement point by adding (not added etc.) (for example, see Non-Patent Document 1).

"A Passive Method for Estimating End-to-End TCP Packet Loss", Peter Benko and Andras Veres, IEEE GLOBECOM 2002, vol.3, p. 2609-2613.

  In Non-Patent Document 1, in an environment where the packet loss rate of the network before the measurement point (TCP transmission terminal to measurement point) is larger than the packet loss rate of the subsequent network (measurement point to TCP reception terminal), the estimation error is several hundreds. It is clear from the evaluation of Non-Patent Document 1 that it becomes a percentage, which is a practical problem.

  In order to solve the problem of Non-Patent Document 1, there is a technique for improving estimation accuracy by using not only data packets but also information of ACK packets. The technology detects a flow that flows between the network A and the network B and a direction in which the received data packet flows in the IP network in which the terminal group performs TCP communication via the network A and the network B, and receives the received data packet. To receive data packet information and ACK packet information. If the extracted ACK packet information includes a SACK packet whose reception has been confirmed, the sequence number of the SACK packet is stored, and the data packet information, the flow storage means, and the stored SACK number are referred to Therefore, it is determined whether the sequence number of the received data packet is a sequence, whether the packet is a retransmission packet, whether the received packet is a duplicate packet, and a duplicate packet In this case, it is determined whether the transmission terminal is a packet retransmitted due to timeout, and the packet loss of the network corresponding to the flow direction is counted based on the determination result. This technology is hereinafter referred to as “Technology 1”.

  The problem of Non-Patent Document 1 can be solved by the technology 1 only when a timeout occurs due to the loss of the ACK packet in the network before the measurement point, and the packet loss after the measurement point is erroneously counted. is there. In addition to this, the causes of erroneous counting due to the occurrence of timeout include loss of ACK packet after the measurement point, stall of congestion window of TCP transmission terminal due to burst loss, and the like. In order to measure the packet loss with high accuracy, it is necessary to correctly discriminate these and reduce the erroneous count.

  The present invention has been made in view of the above points, and an object of the present invention is to provide a packet loss number measuring apparatus, method, and program capable of improving the estimation accuracy of the packet loss number.

In order to solve the above-described problems, the present invention (Claim 1) is configured such that a TCP transmission terminal group and a TCP reception terminal group include a network A (TCP transmission terminal to measurement point) and a network B (measurement point to TCP reception terminal). Via a TCP (Transmission Control Protocol) communication at a measurement point of an IP (Internet Protocol) network, a packet loss number measuring device for measuring the number of packet losses of the network,
At least for each flow identifier, flow storage means for storing the flow direction, the reception time of the last packet, the type of the last packet, a timeout flag, a sequence number at timeout,
The flow flowing between the network A and the network B and the direction in which the received packet of the flow flows are detected, the received packet information is stored in the flow storage means, and the received data packet information and the ACK packet are received from the received packet. Flow processing means for extracting and outputting information;
For the received data packet information acquired from the flow processing means, it is determined whether it is timed out based on the time difference between the data packet and the ACK packet sent before it, and whether it is a predetermined condition and a retransmission packet TCP data packet processing means for calculating the packet loss number based on whether or not, and updating the packet loss number storage means;
Based on the ACK packet information acquired from the flow processing means, referring to the flow information storage means, if the timeout flag indicating the status of packet loss between the measurement point and the TCP receiving terminal is on, TCP ACK packet processing means for calculating the packet loss number and updating the packet loss number storage means.

The present invention (Claim 2) provides the TCP data packet processing means,
For the received data packet information acquired from the flow processing means, if it has timed out, determine that there is a possibility that a packet loss has occurred between the measurement point and the TCP receiving terminal,
As the predetermined condition, it is determined whether the type of the final packet of the flow storage means is ACK, the duplication number of the ACK number is 0, and the flag of the retransmission packet is off. Packet loss number N
N = LOG2 ((packet reception time-last ACK packet reception time of flow storage means) / estimated RTO value) + 1
And a first packet loss number calculation means for turning on the retransmission packet flag of the flow storage means.

The present invention (Claim 3) provides the TCP data packet processing means,
If the received data packet information acquired from the flow processing means has not timed out and the received sequence number storage means does not include the sequence number of the data packet, the received sequence number storage means Stores the upper limit value and lower limit value of the sequence number, turns off the retransmission packet flag of the flow storage means, and if included, estimates the time obtained by subtracting the reception time of the last packet of the flow storage means from the current time If it is equal to or greater than the RTO value, it is determined that a timeout has occurred. If the final packet type of the flow storage means is a data packet, the timeout flag of the flow storage means is turned on and the final packet type is an ACK packet In this case, it is added to the packet loss number storage means according to the flow direction, and the flow recording is performed. It means for turning on the retransmission packet flag means,
Means for setting an upper limit value of the sequence number of the packet to the timeout sequence number of the flow storage means;
If the time obtained by subtracting the reception time of the last packet of the flow storage means from the current time is smaller than the estimated RTO value, a second packet loss number calculation means for setting the retransmission packet flag of the flow storage means off;
Means for setting the current time to the reception time of the final packet of the flow storage means and setting the type of the final packet as a data packet.

The present invention (Claim 4) is characterized in that in the TCP ACK packet processing means,
As a result of referring to the flow information storage means, when the timeout flag is on and the value obtained by adding 1 to the timeout sequence number is equal to the sequence number of the ACK packet, depending on the flow direction, A third packet loss number calculating means for adding 1 to the packet loss number in the packet loss number storage means and turning on a retransmission packet flag of the flow of data packets in the flow information storage means;
Means for turning off the timeout flag of the flow information storage means.

The present invention (Claim 5) provides the TCP ACK packet processing means,
As a result of referring to the flow information storage means, when the timeout flag is OFF and the sequence number of the ACK packet and the sequence number of the final ACK packet of the flow information storage means are the same value, the flow information storage 1 is added to the duplication number of the means, and if not the same value, means for setting 0 to the duplication number of the flow information storage means,
Means for setting the sequence number of the received ACK packet to the sequence number of the final ACK packet in the flow information storage means.

In the present invention (Claim 6), in the first packet loss number calculating means,
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, the packet loss number N is added to the packet loss number of the network B of the packet loss number storage means. When the network B flows from the network B to the network A, the packet loss number storage means includes means for adding the packet loss number N to the packet loss number of the network A.

In the present invention (Claim 7), in the second packet loss number calculating means,
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, 1 is added to the packet loss number of the network B of the packet loss number storage means, and the network In the case of flowing from B to the network A, a means for adding 1 to the packet loss number of the network A in the packet loss number storage means is included.

The present invention (Claim 8) is characterized in that in the third packet loss number calculating means,
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, 1 is added to the packet loss number of the network B of the packet loss number storage means, and the network In the case of flowing from B to the network A, a means for adding 1 to the packet loss number of the network A in the packet loss number storage means is included.

According to the present invention, measurement of an IP network in which a TCP transmission terminal group and a TCP reception terminal group perform TCP communication via a network A (TCP transmission terminal-measurement point) and a network B (measurement terminal-TCP reception terminal). At the point, by selecting the received packet information with the possibility of packet loss and setting a flag in the flow table (flow storage means),
-All packet loss of window at measurement point-TCP receiving terminal (network A-> network B);
-Packet loss of more than half of the window at the measurement point-TCP receiving terminal (network A-> network B) (but not all packet loss);
-All ACK packet loss of window from TCP receiving terminal to measurement point (network B ⇒ network A);
Can be detected, and the accuracy of packet loss measurement can be improved.

It is a figure which shows the measurement point in one embodiment of this invention. Packet loss occurrence point and occurrence pattern. It is a block diagram of the network assumed by this invention. It is a block diagram of the packet loss number measuring apparatus in one embodiment of this invention. It is an example of the flow table in one embodiment of this invention. It is an example of the packet loss number table in one embodiment of this invention. It is an example of the received sequence number table in one embodiment of this invention. It is a flowchart of the TCP data packet processing part in one embodiment of the present invention. 4 is a flowchart of a TCP ACK packet processing unit in an embodiment of the present invention.

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

  In the present invention, similarly to Non-Patent Document 1 and Technique 1 above, as shown in FIG. 1, in the data packet transfer direction (the communication direction from the TCP transmitting terminal) to the TCP receiving terminal) Measure the number of packet losses in the network. In the figure, the c and d packet losses in the reverse direction need to be considered in order to distinguish them from the packet losses in the sections a and b, and are mentioned in this specification, but actually count the number of packet losses. Are only points in the sections a and b in the direction from the TCP transmitting terminal to the TCP receiving terminal.

  The “measurement point” is the position of the packet loss number measuring apparatus of the present invention.

  FIG. 2 shows packet loss occurrence points and occurrence patterns.

  In the figure, a section a indicates between a TCP transmission terminal and a measurement point in communication from a TCP transmission terminal to a TCP reception terminal. The interval b indicates between the measurement point and the TCP receiving terminal in the communication from the TCP transmitting terminal to the TCP receiving terminal. The section c indicates between the TCP receiving terminal and the measurement point in the communication from the TCP receiving terminal to the TCP transmitting terminal. The interval d indicates between the measurement point and the TCP transmission terminal in the communication from the TCP reception terminal to the TCP transmission terminal.

  A state in which packets less than half the window are lost in the interval a (hereinafter referred to as “a- (2)”), and a state in which packets less than half the window are lost in the interval b (hereinafter, “ “b- (2)” can be detected by the method of Non-Patent Document 1.

  However, in the case where more than half of the packets are lost and the TCP transmission terminal times out, in the interval a, all the packets in the window are lost (hereinafter referred to as “a- (1)”), A state in which more than half of the windows are lost in a (hereinafter referred to as “a- (3)”), or a state in which all packets in the window are lost in interval b (hereinafter, “b- (1)”). ), A state in which more than half of the windows are lost (hereinafter referred to as “b- (3)”), or a state in which all ACK packets in the window are lost in the interval c (hereinafter referred to as “c- (1)”. ) ”) Or a state in which all ACK packets of the window are lost in the interval d (hereinafter referred to as“ d- (1) ”), the behavior of TCP is very similar and difficult to distinguish. These correctly Inability another has become a cause for a large error in the non-patent document 1 approach.

  Further, in the technique 1, it is possible to detect d- (1) in which all ACK packets in the window are lost in the interval d. However, in the present invention, “a- (1)”, “b- ( 1) "," b- (3) ", and" c- (1) "can be detected to improve measurement accuracy.

  FIG. 3 is a configuration diagram of a network assumed in the present invention.

  In the system shown in the figure, a plurality of terminal groups 100 and 200 perform TCP communication via a network A and a network B. A device for copying communication data between the network A and the network B (a port mirroring function of a network switch, a tap device, etc.) is placed, and a copy of a packet exchanged between the TCP transmitting terminal 100 and the TCP receiving terminal 200 is the number of packet losses. Connection is made so as to be transferred to the measuring apparatus 300. The packet loss number measuring apparatus 300 measures the number of packet losses in the network A and the network B by analyzing packets flowing between the network A and the network B.

  Each case will be described.

<Case of a- (1)>
In this case, all packets are lost before the packet transmitted from the TCP transmission terminal 100 arrives at the measurement point, and after the TCP transmission terminal 100 times out, the packet is retransmitted and arrives at the measurement point. .

  In this case, the measurement point does not know that the first packet has been transmitted, and only the retransmission packet arrives. However, it is not possible to determine whether the retransmission packet is the first transmission packet or the retransmission packet because there is no such information in the TCP packet. For this reason, since the time difference between two packets is not known, it is difficult to even determine whether a timeout has occurred.

In the present invention, in order to detect the timeout, the timeout is detected from the time difference between the arrived packet pkt _i and the ACK packet ACK _i−1 sent before that.

T _i −TA _i−1 ≧ (RTOestimated + α)
If the following formula is established, it is considered that a timeout has occurred. In this method, since it is necessary to consider the propagation delay until the ACK packet arrives at the TCP transmitting terminal 100, for example, a measured value of RTT (Round Trip Time) between the TCP transmitting terminal 100 and the measurement point is used. And
α = (RTT between TCP transmission terminal and measurement point) / 2
It is also possible to correct as follows.

  If the occurrence of a timeout is detected, an evaluation is performed to determine whether the following conditions are met in order to prevent erroneous determination.

• pkt _i and the previous ACK packet have the same sequence number.

  ・ No duplicate ACK has occurred.

The previous data packet pkt _i-1 is not a retransmission packet.

  In summary, the following equation is obtained. When this formula is applied, it is considered that the packet loss pattern a- (1) has occurred.

However,

TA _i : Arrival time at the measurement point of ACK _i
ACK _i : Acknowledgment number of ACK _i (next expected sequence number
RTO _estimated : Estimated value of RTO
Dupack: Number of consecutive ACK packets with the same ACK number
RX _i : pkt _i is a flag indicating the type of retransmission packet (retransmission packet = 1, not retransmission packet = 0)
It is.

  This pattern is a pattern in which all packets are lost and the packet is retransmitted after a timeout occurs, but there is a possibility that the retransmission packet may be lost once more or a plurality of times.

  TCP specifies an exponential backoff that doubles the retransmission interval each time a retransmission occurs due to a timeout (RFC2988). Considering this exponential backoff, the number of packet retransmissions (= number of packet losses) is specified as follows. As described above, α is a correction considering the propagation delay until the ACK packet reaches the TCP transmitting terminal 100.

Number of packet loss = Log ₂ ((T _i −TA _i−1 −α / RTO _estimated ) +1
<Case b- (1)>
This case is a case where all packets in the TCP window are lost in the interval b after the measurement point. In this case, since ACK does not return to the TCP transmission terminal 100, the TCP transmission terminal 100 times out.

Since a timeout occurs, the interval between the retransmitted data packet pkt _i and the previous data packet pkt _i−1 is equal to or greater than the RTO (Retransmission Time Out) value. The TCP receiving terminal 200 returns an ACK after receiving the retransmission packet pkt _i , but since all the packets in the window are lost before arriving at the TCP receiving terminal 200, packets other than the retransmitted pkt _i are not received. Not receiving. For this reason, the returned ACK can be expected to indicate the next sequence number of pkt _i . Therefore, as a detection method of the case, it is possible to detect a retransmission packet (a packet that has already been observed at the measurement point) at the measurement point using the following determination formula.

However,

It is.

<Case b- (3)>
This case is a case where more than half of the TCP window packets are lost in the interval b after the measurement point. In this case, when the TCP transmitting terminal 100 detects Triple Duplicated ACK, the congestion avoiding mode is entered and the window is halved. However, since more than half of the packets are lost, the TCP transmitting terminal 100 causes a timeout. .

Since a timeout occurs, the interval between the retransmitted data packet pkt _i and the immediately preceding data packet pkt _{i i−1} is equal to or greater than the RTO value. Although a timeout occurs, ACK has normally reached the TCP transmission terminal 100, so that the retransmission packet pkt _i becomes a retransmission packet according to the acknowledgment information of the ACK packet. That is, the lower limit of the sequence number of pkt _i is the same as the sequence number indicated by the previously received ACK packet ACK _i-1 . Since TCP receiving terminal 200 which has already received for some packets other than packets lost, ACK packet ACKi the TCP receiving terminal 200 which has received the pkt _i sends back the the next pkt _i as acknowledgment number It can be expected to indicate a number larger than the sequence number. Therefore, as a detection method of b- (3), it is possible to detect a retransmission packet (a packet that has already been observed at the measurement point) at the measurement point using the following determination formula.

However,

It is.

  Moreover, it is also possible to detect only by detecting a timeout and confirming that the previous packet is ACK, as in the following judgment formula.

<C- (1) Case>
This case is a case where all ACK packets in the TCP window are lost in the interval c after the measurement point. In this case, since the ACK does not return to the TCP transmission terminal 100, the TCP transmission terminal 100 causes a timeout. Since a timeout occurs, the interval between the retransmitted data packet pkt _i and the immediately preceding data packet pkt _i-1 is equal to or greater than the RTO value. After receiving the retransmission packet pkt _i , the TCP receiving terminal 200 returns an ACK, but the lost packet is the ACK packet, and the data packet has already reached the TCP receiving terminal 200. For this reason, the returned ACK can be expected to indicate a larger number than the sequence number next to pkt _i . Therefore, as a detection method of c- (1), a retransmitted packet (a packet that has already been observed at the measurement point) can be detected by the following determination formula at the measurement point.

The packet loss number measuring apparatus of the present invention will be described below.

  FIG. 4 shows the configuration of the packet loss number measuring apparatus according to the embodiment of the present invention.

  A packet loss measuring apparatus 300 shown in FIG. 1 includes a flow processing unit 1, a TCP data packet processing unit 2, a TCP ACK packet processing unit 3, a flow table 4, a received sequence number table 5, and a packet loss number table 6. . The flow table 4, the received sequence number table 5, and the packet loss number table 6 are storage media such as a memory and a hard disk device.

  In the figure, the packets flowing from the network A to the network B and the packets flowing from the network B to the network A are connected to the flow processing unit 1 by, for example, connecting them with separate lines so that the flow direction of the packets can be identified. Shall be transferred.

  FIG. 5 is an example of a flow table in one embodiment of the present invention.

  The flow table 4 shown in the figure includes a flow ID, a transmission source IP address, a transmission destination IP address, a transmission source TCP port number, a transmission destination TCP port number, a flow direction, a reception time of the final packet, and a reception time of the final ACK packet. The sequence number of the last ACK packet, the overlapping number of ACK numbers, the type of last packet (data or ACK), the retransmission packet flag, the timeout flag, and the sequence number at timeout are stored. The retransmission packet flag is a flag that is turned on when there is a possibility of packet and loss (however, not all packet loss) of half or more of the windows in the interval b. The time-out flag is a flag that is turned on when there is a possibility of a total packet loss of the window in the period b or a possibility of a total ACK packet loss of the window in the period c.

  FIG. 6 is an example of a packet loss number table in one embodiment of the present invention.

  The packet loss number table 5 shown in the figure stores the flow ID, the network A packet loss number, and the network B packet loss number.

  FIG. 7 is an example of a received sequence number table in one embodiment of the present invention.

  The received sequence number table 5 shown in the figure stores the flow ID, the lower limit value and the upper limit value of the sequence number of the packet.

  In the flow processing unit 1, from the IP header and the TCP header of the received packet, those that can be uniquely identified by the source IP address, destination IP address, source TCP port number, destination TCP port number, and flow direction If it is detected as a flow and the flow table 4 is referenced and the flow is already registered in the flow table 4, a flow ID is obtained. If it is not registered, a unique flow ID is newly assigned. Register in Table 4. It is assumed that “0” (off) is set as a default for the retransmission packet flag and the timeout flag at this time.

  The flow direction set in the flow table 4 is, for example, “1” when flowing from the network A to the network B and “0” when flowing from the network B to the network A with reference to the direction in which the data packet flows. An identification number like this is set.

  Since TCP is a bidirectional communication protocol, one packet includes both a sequence number as a data packet and an ACK sequence number for confirmation of reception.

  The flow processing unit 1 extracts data packet information from the received packet from the network A to the network B and passes it to the TCP data packet processing unit 2. In addition, ACK packet information is extracted from the packet from network B to network A and passed to TCP ACK packet processing unit 3.

The TCP data packet processing unit 2
a) Confirmation of retransmission packet;
b) confirmation of timeout;
c) Confirmation of ACK before timeout;
I do.

  a) For confirmation of a retransmitted packet, it is checked whether the sequence number of the packet is included in the received sequence number table 6.

  b) For confirmation of timeout, it is checked whether the reception time of the last packet in the flow table 4 from the current time is equal to or greater than the estimated (Retransmission TimeOut) value.

  c) Confirmation of ACK before time-out is whether the type of the final packet in the flow table 4 is a data packet when (current time−reception time of the final packet in the flow table 4 ≧ estimated RTO value) in b) above. Check for ACK packet.

  Hereinafter, the processing of the TCP data packet processing unit 3 will be described.

  FIG. 8 is a flowchart of the TCP data packet processing unit in one embodiment of the present invention.

  Step 101) The TCP data packet processing unit 2 acquires a flow ID and packet data from the flow processing unit 1.

  Step 102) If the current time minus the reception time of the last ACK packet in the flow table 4 ≧ estimated RTO value + α, the process proceeds to Step 103; otherwise, the process proceeds to Step 106. Α is a correction value.

  Step 103) If the following condition is satisfied, the process proceeds to Step 104. If not, the process proceeds to Step 106.

conditions)
-Last packet type of flow table 4 = ACK;
-Duplicate number of ACK numbers in flow table 4 = 0;
-Retransmission packet flag of flow table 4 = 0;
Step 104) The packet loss number is calculated and stored in the packet loss number table 5.

The number of packet loss is
N = LOG ₂ (packet reception time−flow table last ACK packet reception time−α) / estimated TRO value) +1
Ask for.

  When the number of packet losses is stored in the packet loss number table 5, it differs depending on the flow direction as follows.

-When the flow direction is network A → network B:
N is added to the packet loss number of the network B in the packet loss number table 5.

・ When the flow direction is network B → network A:
N is added to the packet loss number of the network A in the packet loss number table 5.

  Step 105) The retransmission packet flag of the flow table 4 is set to 1, and the process proceeds to Step 116.

  Step 106) It is determined whether the sequence number of the packet is included in the received sequence number table 6. If included, the process proceeds to Step 107. If not included, the process proceeds to Step 114.

Step 107) As confirmation of timeout,
If the current time—the reception time of the last packet in the flow table ≧ the estimated RTO value, “there is a possibility of timeout”, the process proceeds to step 108; otherwise, the process proceeds to step 113.

  Step 108) As a confirmation of the ACK packet before the timeout, it is determined whether the type of the final packet in the flow table 4 is a data packet or an ACK packet. If YES, go to Step 110.

  Step 109) “1” is set in the timeout flag of the flow table 4 indicating the possibility of the case of b- (1) or c- (1), and the routine proceeds to Step 112.

  Step 110) Record the packet loss in the packet loss number table 5. At this time, the data is stored as follows depending on the flow direction.

-When the flow direction is network A → network B:
1 is added to the packet loss number of the network B in the packet loss number table 5.

・ When the flow direction is network B → network A:
1 is added to the packet loss number of the network A in the packet loss number table 5.

  Step 111) The retransmission packet flag of the flow table 4 is set to “1”.

  Step 112) The upper limit value of the sequence number of the packet is recorded in the time-out sequence number of the flow table 4, and the process proceeds to Step 116.

  Step 113) The retransmission packet flag of the flow table 4 is set to 0, and the process proceeds to Step 116.

  Step 114) In step 106, if the packet sequence number is not included in the received sequence number table 6, the upper limit value and the lower limit value of the sequence number of the packet are recorded in the received sequence number of the flow table 4.

  Step 115) The retransmission packet flag of the flow table 4 is set to 0.

  Step 116) The current time is set as the reception time of the last packet in the flow table 4, and the type of the last packet (data packet / ACK packet) is set.

  The TCP ACK packet processing unit 3 acquires ACK packet information extracted from the network B to the network A via the flow processing unit 1, and the number of packets assumed to have lost the flow in the network A and the network B. The packet loss rate (= packet loss number / transfer packet number) can be calculated.

  Next, the processing of the TCP ACK packet processing unit 3 will be described.

  FIG. 9 is a flowchart of the TCP ACK packet processing unit in one embodiment of the present invention.

  Step 201) The TCP ACK packet processing unit 3 acquires a flow ID and ACK packet data as flow information from the flow table 4.

  Step 202) It is determined whether the timeout flag of the flow table 4 indicating the possibility of the case of b- (1) or c- (1) is “1”. If “0”, the process proceeds to step 207.

  Step 203) If the sequence number at timeout in the flow table 4 + 1 = the sequence number of the ACK packet, it is determined that there is a packet loss, and the process proceeds to Step 204. Otherwise, the process proceeds to Step 206.

  Step 204) The number of packet losses is obtained as follows, recorded in the packet loss number table 5, and the process proceeds to Step 205.

-When the flow direction is (Network A ⇒ Network B):
Add 1 to the packet loss number of network B in the packet loss number table.

・ When the flow direction is (Network B⇒Network A):
1 is added to the packet loss number of the network A in the packet loss number table.

  Step 205) The retransmission packet flag of the reverse flow (data packet flow) in the flow table 4 is set to 1.

  Step 206) The timeout flag of the flow table 4 is set to “0”.

  Step 207) It is determined whether the sequence number of the ACK packet and the sequence number of the final ACK packet in the flow table 4 are the same value. If they are the same value, the process proceeds to Step 209, and if they are different values, the process proceeds to Step 208. Transition.

  Step 208) Set 0 to the overlap number of the ACK number in the flow table 4 and proceed to Step 210.

  Step 209) 1 is added to the overlap of the ACK number in the flow table 4.

  Step 210) The sequence number of the received ACK packet is set in the sequence number of the final ACK packet in the flow table 4.

  Since the present invention is aimed at counting the loss of data packets, the case of c- (1) is a loss of ACK packets, and therefore recording to the packet loss number table 5 is not performed.

  The operation of each component of the packet loss number measuring apparatus shown in FIG. 4 can be constructed as a program, installed in a computer used as the packet loss number measuring apparatus and executed, or distributed via a network. Is possible.

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

DESCRIPTION OF SYMBOLS 1 Flow processing part 2 TCP data packet processing part 3 TCP ACK packet processing part 4 Flow table 5 Packet loss number table 6 Received sequence number table 100 TCP transmitting terminal 200 TCP receiving terminal 300 Packet loss number measuring device

Claims (15)

IP (Internet Protocol) in which TCP transmission terminal group and TCP receiving terminal group perform TCP (Transmission Control Protocol) communication via network A (TCP transmission terminal to measurement point) and network B (measurement point to TCP reception terminal) A packet loss number measuring device for measuring the number of packet losses in a network at a measurement point of the network,
At least for each flow identifier, flow storage means for storing the flow direction, the reception time of the last packet, the type of the last packet, a timeout flag, a sequence number at timeout,
The flow flowing between the network A and the network B and the direction in which the received packet of the flow flows are detected, the received packet information is stored in the flow storage means, and the received data packet information and the ACK packet are received from the received packet. Flow processing means for extracting and outputting information;
For the received data packet information acquired from the flow processing means, it is determined whether it is timed out based on the time difference between the data packet and the ACK packet sent before it, and whether it is a predetermined condition and a retransmission packet TCP data packet processing means for calculating the packet loss number based on whether or not, and updating the packet loss number storage means;
The flow storage means is referred to based on the ACK packet information acquired from the flow processing means, and when the timeout flag indicating the state of packet loss between the measurement point and the TCP receiving terminal is on, the packet TCP ACK packet processing means for calculating the number of losses and updating the packet loss number storage means;
A device for measuring the number of packet losses. The TCP data packet processing means includes
For the received data packet information acquired from the flow processing means, if it has timed out, determine that there is a possibility that a packet loss has occurred between the measurement point and the TCP receiving terminal,
As the predetermined condition, it is determined whether the type of the final packet of the flow storage means is ACK, the duplication number of the ACK number is 0, and the flag of the retransmission packet is off. Packet loss number N
N = LOG2 ((packet reception time-last ACK packet reception time of flow storage means) / estimated RTO value) + 1
2. The packet loss number according to claim 1, further comprising: a first packet loss number calculation unit that calculates the packet loss number according to a flow direction and adds the packet loss number storage unit to the retransmission packet flag of the flow storage unit. measuring device. The TCP data packet processing means includes
If the received data packet information acquired from the flow processing means has not timed out and the received sequence number storage means does not include the sequence number of the data packet, the received sequence number storage means Stores the upper limit value and lower limit value of the sequence number, turns off the retransmission packet flag of the flow storage means, and if included, estimates the time obtained by subtracting the reception time of the last packet of the flow storage means from the current time If it is equal to or greater than the RTO value, it is determined that a timeout has occurred. If the final packet type of the flow storage means is a data packet, the timeout flag of the flow storage means is turned on, and the final packet type is an ACK packet. In this case, it is added to the packet loss number storage means according to the flow direction, and the flow recording is performed. It means for turning on the retransmission packet flag means,
Means for setting an upper limit value of the sequence number of the packet to the timeout sequence number of the flow storage means;
If the time obtained by subtracting the reception time of the last packet of the flow storage means from the current time is smaller than the estimated RTO value, a second packet loss number calculation means for setting the retransmission packet flag of the flow storage means off;
2. The packet loss number measuring apparatus according to claim 1, further comprising means for setting a current time as a reception time of the final packet of the flow storage unit and setting a type of the final packet as a data packet. The TCP ACK packet processing means is
As a result of referring to the flow storage means, when the timeout flag is on and the value obtained by adding 1 to the timeout sequence number is equal to the sequence number of the ACK packet, depending on the flow direction, the packet A third packet loss number calculating means for adding 1 to the packet loss number of the loss number storage means and turning on the retransmission packet flag of the flow of data packets in the flow storage means;
The packet loss number measuring apparatus according to claim 1, further comprising: means for turning off the timeout flag of the flow storage means. The TCP ACK packet processing means is
As a result of referring to the flow storage means, if the timeout flag is OFF and the sequence number of the ACK packet and the sequence number of the final ACK packet of the flow storage means are the same value, the flow storage means is duplicated. 1 is added to a plurality, and if they are not the same value, means for setting 0 to the overlap number of the flow storage means,
Means for setting the sequence number of the received ACK packet to the sequence number of the final ACK packet of the flow storage means;
2. The packet loss number measuring apparatus according to claim 1, comprising: The first packet loss number calculating means includes
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, the packet loss number N is added to the packet loss number of the network B of the packet loss number storage means. 3. A packet loss number measuring device according to claim 2, further comprising means for adding the packet loss number N to the packet loss number of the network A in the packet loss number storage means when flowing from the network B to the network A. . The second packet loss number calculating means is
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, 1 is added to the packet loss number of the network B of the packet loss number storage means, and the network 4. The packet loss number measuring apparatus according to claim 3, further comprising means for adding 1 to the packet loss number of the network A in the packet loss number storage means when flowing from B to the network A. The third packet loss number calculating means,
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, 1 is added to the packet loss number of the network B of the packet loss number storage means, and the network 5. The packet loss number measuring device according to claim 4, further comprising means for adding 1 to the packet loss number of the network A in the packet loss number storage means when flowing from B to the network A. IP (Internet Protocol) in which TCP transmission terminal group and TCP receiving terminal group perform TCP (Transmission Control Protocol) communication via network A (TCP transmission terminal to measurement point) and network B (measurement point to TCP reception terminal) A packet loss number measuring method for measuring the number of packet losses in a network at a measurement point of the network,
At least, for each flow identifier, in an apparatus having flow storage means for storing the flow direction, the last packet reception time, the last packet type, the timeout flag, and the timeout sequence number,
The flow processing means detects the flow flowing between the network A and the network B and the direction in which the received packet of the flow flows, stores the received packet information in the flow storage means, and receives received data from the received packet. A flow processing step for extracting and outputting packet information and ACK packet information;
The TCP data packet processing means determines whether the received data packet information acquired from the flow processing means has timed out based on the time difference between the data packet and the ACK packet sent before the data packet, and a predetermined condition And a TCP data packet processing step of calculating a packet loss number based on whether the packet is a retransmission packet and updating the packet loss number storage means;
TCP ACK packet processing means refers to the flow storage means based on the ACK packet information, and when the timeout flag indicating the state of packet loss between the measurement point and the TCP receiving terminal is on, the packet A TCP ACK packet processing step of calculating a loss number and updating the packet loss number storage means;
A method for measuring the number of packet losses. In the TCP data packet processing step,
For the received data packet information, if a timeout has occurred, determine that there is a possibility that a packet loss has occurred between the measurement point and the TCP receiving terminal,
As the predetermined condition, it is determined whether the type of the final packet of the flow storage means is ACK, the duplication number of the ACK number is 0, and the flag of the retransmission packet is off. Packet loss number N
N = LOG2 ((packet reception time-last ACK packet reception time of flow storage means))
/ Estimated RTO value) +1
A first packet loss number calculating step of turning on the retransmission packet flag of the flow storage means, and adding to the packet loss number storage means according to the flow direction;
If the received data packet information has not timed out and the received sequence number storage means does not include the sequence number of the data packet, the received sequence number storage means stores an upper limit value of the sequence number and Store the lower limit value, turn off the retransmission packet flag of the flow storage means, if included, if the time obtained by subtracting the reception time of the last packet of the flow storage means from the current time is greater than or equal to the estimated RTO value If it is determined that a time-out has occurred and the type of the final packet of the flow storage means is a data packet, the time-out flag of the flow storage means is turned on. If the type of the final packet is an ACK packet, Accordingly, the packet loss count storage means is added, and the retransmission packet flag of the flow storage means is turned on. To,
A step of setting an upper limit value of the sequence number of the packet to the timeout sequence number of the flow storage means;
If the time obtained by subtracting the reception time of the last packet of the flow storage means from the current time is smaller than the estimated RTO value, a second packet loss number calculation step of setting the retransmission packet flag of the flow storage means off;
The packet loss number measuring method according to claim 9, further comprising: setting a current time as a reception time of the final packet of the flow storage unit, and setting a type of the final packet as a data packet. In the TCP ACK packet processing step,
As a result of referring to the flow storage means, when the timeout flag is on and the value obtained by adding 1 to the timeout sequence number is equal to the sequence number of the ACK packet, depending on the flow direction, the packet A third packet loss number calculation step of adding 1 to the packet loss number of the loss number storage means, turning on the retransmission packet flag of the data packet flow in the flow storage means, and turning off the timeout flag of the flow storage means When,
As a result of referring to the flow storage means, if the timeout flag is OFF and the sequence number of the ACK packet and the sequence number of the final ACK packet of the flow storage means are the same value, the flow storage means is duplicated. A step of adding 1 to a plurality and setting 0 to the overlap number of the flow storage means if they are not the same value, and setting the sequence number of the received ACK packet to the sequence number of the final ACK packet of the flow storage means When,
The method for measuring the number of packet losses according to claim 9. In the first packet loss number calculating step,
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, the packet loss number N is added to the packet loss number of the network B of the packet loss number storage means. 11. The packet loss number measuring method according to claim 10, wherein, when flowing from the network B to the network A, the packet loss number N is added to the packet loss number of the network A of the packet loss number storage means. In the second packet loss number calculating step,
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, 1 is added to the packet loss number of the network B of the packet loss number storage means, and the network 11. The packet loss number measuring method according to claim 10, wherein when the packet flows from B to the network A, 1 is added to the packet loss number of the network A in the packet loss number storage means. In the third packet loss number calculating step,
When adding to the packet loss number storage means, if the flow direction flows from the network A to the network B, 1 is added to the packet loss number of the network B of the packet loss number storage means, and the network 12. The packet loss number measuring method according to claim 11, wherein when the packet flows from B to the network A, 1 is added to the packet loss number of the network A in the packet loss number storage means. Computer
A packet loss number measurement program for causing each of the means of the packet loss number measurement apparatus according to any one of claims 1 to 8 to function.