JP6022948B2 - Management device and program for control device of communication network - Google Patents

Description

  The present invention relates to a network control technique.

  Open flow technology has attracted attention as a network control. When a network using OpenFlow technology receives a packet belonging to a new flow and a switch device that is a communication device that processes packets, the route of this flow is determined, and each route on the determined route is determined. And a control device that notifies the communication device of the output interface of the packet and updates the flow table. The flow table is a table showing the correspondence between information for specifying which flow a packet belongs to and the output interface of the packet belonging to the flow.

  Here, when the control device becomes faulty, the switch device notified of the output interface by the control device cannot transfer the packet belonging to the new flow. For this reason, Patent Document 1 has a configuration in which a plurality of control devices independently calculate a flow route and notify a switch device on each route of an output interface of a packet belonging to a new flow. Disclosure. In the method described in Patent Document 1, the switch device may receive an instruction of an output interface from each of one or more control devices, but the switch device transfers a packet in accordance with the first received instruction.

JP 2011-160363 A

  In the method described in Patent Document 1, since the switch device receives instructions for updating the flow table from a plurality of control devices, the load on the switch device temporarily increases. In addition, since each control device independently determines a flow route, information on unused flows remains in the flow table in the switch device on the unselected route.

  Here, unlike Patent Document 1, a plurality of control devices are provided. However, instead of a plurality of control devices controlling the same switch device, a configuration in which one switch device is controlled by only one control device is considered. . In this case, when a certain control device becomes an obstacle, it is necessary to promptly take over control of the switch device controlled by this control device to another control device. At this time, the problem is how to monitor the failure of the control device and how to determine the control device that takes over the control of the switch device controlled by the failed control device.

  In view of the above problems, the present invention provides a management device that can monitor a failure of a control device, and a management device and program that can cause another control device to take over control of a communication device controlled by the failed control device Is to provide.

According to one aspect of the present invention, communication devices belonging to a network are grouped into a plurality of groups, a plurality of control devices corresponding to the plurality of groups are provided, and each control device controls a communication device of a corresponding group. in the system, the management apparatus that manages the plurality of control devices indicates that a fault condition of said plurality of control devices, said plurality of control apparatus transmits to the management apparatus, the transmission source of the control device is normal The determination means for determining based on the first signal, and one or more control devices that take over control of the communication devices of the group corresponding to the control device in the failed state are set to load values indicating the load states of the plurality of control devices. A selection unit that selects based on the control unit that is selected by the selection unit, and controls the communication devices of the group corresponding to the control unit that has entered the failure state. And notifying means for notifying that splicing can, and wherein the determination means does not receive the first signal from the first controller of the plurality of control devices, among the plurality of control devices The first control device is in a failure state using a second control device different from the first control device or a host device connected to a communication device of a group corresponding to the first control device. It is characterized by determining whether or not there is .

According to one aspect of the present invention, communication devices belonging to a network are grouped into a plurality of groups, a plurality of control devices corresponding to the plurality of groups are provided, and each control device controls a communication device of a corresponding group. In the system, a management device that manages the plurality of control devices, and if a signal indicating normality is not received from a first control device among the plurality of control devices, The second control device other than the first control device is requested to confirm the existence of the first control device, and the first control device indicated in the report received from the second control device is normal. It is characterized by comprising determination means for determining whether or not the first control device is in a failure state based on the number that is present and the number that is abnormal.

According to one aspect of the present invention, communication devices belonging to a network are grouped into a plurality of groups, a plurality of control devices corresponding to the plurality of groups are provided, and each control device controls a communication device of a corresponding group. In the system, if the management device manages the plurality of control devices and does not receive a normal signal from the first control device among the plurality of control devices, the first control device The first control device transmits a survival confirmation signal for requesting a response to the host device connected to the communication device of the corresponding group, and the first control device is based on the transmitted survival confirmation signal and the number of responses to the survival confirmation signal. It is characterized by comprising determination means for determining whether or not it is in a fault state.

  The failure of the control device can be monitored, and the control of the switch device controlled by the failed control device can be taken over by another control device.

1 is an exemplary system configuration diagram according to one embodiment. FIG. The figure which shows the example management information which the management apparatus by one Embodiment hold | maintains. The sequence diagram of the takeover process by one Embodiment. The sequence diagram of the takeover process by one Embodiment. The flowchart of the failure determination process by one Embodiment. The flowchart of the failure determination process by one Embodiment. The flowchart of the takeover destination determination process by one Embodiment. The schematic block diagram of the management apparatus by one Embodiment.

  Hereinafter, exemplary embodiments of the present invention will be described with reference to the drawings. In the following drawings, components that are not necessary for the description of the embodiments are omitted from the drawings.

  FIG. 1 is an exemplary system configuration diagram according to the present embodiment. The network includes a plurality of switch devices 1, and each switch device 1 is classified into one of switch groups 10-1 to 10-n which are n groups (n is an integer of 2 or more). In FIG. 1, in order to avoid complication of the drawing, the switch device 1 is expressed as having a link only with the switch device 1 of the same switch group. The switch device 1 is classified into the switch group and the connection configuration between the switch devices 1 is irrelevant.

  In the present embodiment, n control devices 2-1 to 2-n are provided corresponding to the switch groups 10-1 to 10-n, and the control devices 2-1 to 2-n are respectively corresponding. The switch device 1 included in the switch group to be controlled is controlled. That is, at the normal time, the control device 2-1 recognizes that only the control of the switch device 1 in the switch group 10-1 is performed, and the switch device 1 in the switch group 10-1 also controls the control device 2-2. It is recognized that it is controlled only by one. Therefore, when receiving a notification of reception of a packet belonging to a new flow from the switch device 1 belonging to the switch group 10-1, the control device 2-1 determines the path of the flow and normally selects the switch group 10- Among the switch devices 1 of 1, the output interface is notified to the switch device 1 on the determined path, and the flow table is updated. Each of the control devices 2-1 to 2-n is configured to be accessible to all the switch devices 1 in case of a failure.

  Furthermore, in the present embodiment, it is determined whether the control devices 2-1 to 2-n are in a normal state or a failure state. A management device 3 is provided for determining one or more control devices that take over control of the switch device 1 controlled by the device.

  FIG. 2 is an explanatory diagram of information held by the management device 3 according to the present embodiment. In the example of FIG. 2, the management device 3 holds the number of switch groups 10-1 to 10-n in the normal state and the number of switch devices 1 in each of the switch groups 10-1 to 10-n. Although omitted in FIG. 2, information for identifying each switch device 1 of each of the switch groups 10-1 to 10 -n and information for accessing, for example, an IP address, are also included in the management device 3. Is holding. In FIG. 2, “master” indicates a control device that controls the switch device 1 of the corresponding switch group in a normal time, and “slave” indicates a control device that is not controlled in a normal time. For example, in FIG. 2, the switch device 1 of the switch group 10-1 is normally controlled only by the control device 2-1, and the switch device 1 of the switch group 10-2 is controlled only by the control device 2-2. It has been shown that.

  Moreover, in this embodiment, the management apparatus 3 acquires the load information of each control apparatus 2-1 to 2-n. In this embodiment, the CPU load factor of each control device 2-1 to 2-n is used as load information. The management device 3 acquires the CPU load factor from each control device at a relatively short cycle of, for example, several seconds to several minutes, but for a predetermined period from the CPU load factor acquired at the relatively short cycle, For example, an average value of several days such as one week is obtained and recorded. In the present embodiment, not only the form using the CPU load factor but also the load of the control devices 2-1 to 2-n using both the memory use amount and the CPU load factor and the memory use amount. It is also possible to calculate the load value shown. Further, as shown in FIG. 2, the management device 3 holds setting information set for each of the control devices 2-1 to 2-n. Specifically, for example, it is described that the setting information set in the control device 2-1 is the setting information A, and the setting information set in the control device 2-2 is the setting information B.

  Next, a takeover process executed by the management apparatus 3 according to the present embodiment will be described with reference to FIG. The management device 3 polls the control devices 2-1 to 2-n, for example, and the management device 3 indicates that the control devices 2-1 to 2-n are operating normally by responding to the polling. (S1). Instead of polling by the management apparatus 3, a signal indicating that the control apparatuses 2-1 to 2 -n are normally operating periodically may be transmitted to the management apparatus 3.

  For example, assume that the control device 2-1 becomes a failure (S2), and the management device 3 detects that there is no response from the control device 2-1 (S3). At this point, when viewed from the management device 3, it is determined whether the control device 2-1 is in trouble or only communication between the management device 3 and the control device 2-1 is in trouble. I can't. Therefore, in S4, the management device 3 requests the other control devices 2-2 to 2-n to confirm the existence of the control device 2-1. In response to receiving the survival confirmation request signal from the management apparatus 3, the control apparatuses 2-2 to 2-n transmit a survival confirmation signal to the control apparatus 2-1 in S5. In the present embodiment, each control device is configured to transmit a survival report signal to the transmission source of the survival confirmation signal when receiving the survival confirmation signal. In S6, the control devices 2-2 to 2-n report the survival confirmation result of the control device 2-1 to the management device 3. That is, when the control devices 2-2 to 2-n receive the survival report signal from the control device 2-1, the control device 2-1 reports to the management device 3 that the control device 2-1 is normal. If the survival report signal is not received, the control device 2-1 reports to the management device 3 that there is an abnormality.

  The management device 3 determines whether or not the control device 2-1 is faulty when a report has been received from all of the control devices 2-2 to 2-n or when a predetermined period has elapsed since the confirmation request in S4. (S7). Specifically, the ratio of reported abnormalities to all reports indicating abnormalities or normality received from the control devices 2-2 to 2-n at the time of determination is obtained, and if the obtained ratio is equal to or greater than a threshold value, control is performed. It is determined that the device 2-1 has a failure. When the management device 3 determines that the control device 2-1 is in failure, in S7, the management device 3 controls all the switch devices 1 belonging to the switch group 10-1 controlled by the control device 2-1. Decide which control device to take over.

  Thereafter, according to the determination in S7, the management device 3 transmits a takeover instruction signal to the control devices 2-2 to 2-n in S8. The takeover instruction signal includes information for specifying the switch device 1 to which the control device of the transmission destination takes over control and setting information necessary for controlling the switch device 1. It is not necessary to send a takeover instruction signal to a control device that does not take over control of the control device 2-1. In S9, the control device that has received the takeover instruction signal changes the control device that controls the switch device 1 to be controlled by the own device among the switch devices 1 of the switch group 10-1. Notify that it has become. In addition, the process of S9 can utilize "Roll Request" prescribed | regulated by the open flow, for example. When the takeover process in S9 is completed, the control apparatus that has taken over the process of the control apparatus 2-1 transmits a completion report signal to the management apparatus 3 in S10, and the management apparatus 3 adds a switch to the information shown in FIG. Information on which control device is used to control each of the switch devices 1 belonging to the group 10-1 is stored.

  FIG. 4 is a sequence diagram of the takeover process executed by the management device 3 according to another embodiment. Hereinafter, the takeover process of FIG. 4 will be described focusing on differences from the takeover process of FIG. 3. S1 to S3 are the same as the takeover process of FIG. In the embodiment of FIG. 4, when the management device 3 detects that there is no response from the control device 2-1, the management device 3 itself executes a status check of the control device 2-1 (S 4-1). Specifically, the management device 3 transmits a survival confirmation signal to a plurality of host devices (not shown) connected to the switch device 1 included in the switch group 10-1. It is assumed that the host device that has received the survival confirmation signal transmits a survival report signal to the management device 3. For example, an ICMP (Internet Control Notification Protocol) echo request message can be used as the existence confirmation signal. In this case, the existence report signal is an ICMP echo response message. Since the signal does not reach the host device connected to the switch device 1 included in the switch group 10-1 without the control of the control device 2-1, in this embodiment, the survival report signal corresponding to the number of transmission of the survival confirmation signal is transmitted. It is determined whether or not the control device 2-1 has a failure based on the ratio of the number of receptions. The subsequent processing is the same as the takeover processing in FIG. 3, and the description thereof is omitted.

  Next, the control device failure determination process performed by the management apparatus 3 in S7 of the takeover process of FIG. 3 will be described with reference to FIG. When the management apparatus 3 receives a report from all the control apparatuses that have made the confirmation request after a predetermined time has elapsed after the confirmation request in S4 of FIG. 3 (S50), the management apparatus 3 calculates an abnormality rate in S51. Specifically, at the time of calculation, the ratio of reporting an abnormality to the number of reports received from the control device is calculated. If the abnormality rate is equal to or greater than the threshold (S52), the management device 3 determines that the control device is a failure in S53, and otherwise determines that the control device is normal in S54.

  Subsequently, the failure determination processing of the control device performed by the management device 3 in S7 of the takeover processing of FIG. 4 will be described with reference to FIG. When the management apparatus 3 receives a survival report signal from all the host apparatuses that have passed a predetermined time after the confirmation processing in S4-1 of FIG. 4 or transmitted a survival confirmation signal (S60), an abnormality is detected in S61. Calculate the percentage. Specifically, the ratio of the difference between the number of transmitted survival confirmation signals and the number of received survival report signals at the time of calculation is calculated with respect to the number of transmitted survival confirmation signals. If the abnormality rate is equal to or greater than the threshold (S62), the management device 3 determines that the control device is a failure in S63, and otherwise determines that the control device is normal in S64. It is also possible to obtain a normal ratio that is a ratio of the number of received survival report signals at the time of calculation to the number of transmitted survival confirmation signals, and to determine that the control device is faulty if the normal ratio is equal to or less than a threshold.

  Next, the takeover destination determination process performed by the management apparatus 3 in S7 of the takeover process of FIGS. 3 and 4 will be described with reference to FIG. When the management device 3 determines that a certain control device has failed by the failure determination processing described with reference to FIG. 5 or FIG. 6, the load value of the control device that has been determined to have failed in the past predetermined period in S70, this embodiment Then, the average value n (%) of the load factor of the CPU is determined or calculated. Subsequently, even if the average value n (%) of the CPU load factor of the failed control device is added, there is a control device in which the average value of the CPU load factor does not exceed the threshold value X (%). Whether or not to do so is determined in S71.

  In S71, when there is a control device that does not exceed the threshold value X (%), the management device 3 takes over any one of the control devices in S73 as the failed control device in S73. Decide first. On the other hand, if there is no control device that does not exceed the threshold value X (%) in S71, the management device 3 has a plurality of CPU load factors so that the average value of the CPU load factor does not exceed X (%) in S72. Is selected as the takeover destination. Specifically, for example, assuming that the threshold value X (%) is 80%, the average values of the CPU load factors of the control devices A, B, and C are 50%, 50%, and 50%, respectively. The average value of the CPU load factor of the control device D is 60%. In this case, for example, the management device 3 can distribute the average value of 60% of the CPU load factor of the control device D to the control devices A and B by 30% and 30%, respectively. Further, the management device 3 can distribute the average value of 60% of the CPU load factor of the control device D to the control devices B and C by 30% and 30%, respectively. Furthermore, the management device 3 can distribute the average value of 60% of the CPU load factor of the control device D to the control devices A, B, and C by 20%, 20%, and 20%, respectively.

  Note that the selection of a plurality of control devices in S72 can be performed according to various criteria. For example, the selection can be made so that the number of control devices of the takeover destination is minimized. As a result, the number of control devices that increase the load can be minimized. On the other hand, it is possible to select so that the number of control devices as takeover destinations is maximized. As a result, it is possible to minimize an increase in the load of the control device that is operating normally. Further, the installation position of the control device can be taken into consideration. For example, a control device installed at a position as close as possible can be selected. On the other hand, the installation positions of the control devices can be dispersed as much as possible. Further, the CPU load factor margin, that is, the average value of the difference between the threshold value X (%) and the CPU load factor can be selected as the maximum value, or the minimum value can be selected.

  When the management device determines a plurality of control devices as takeover destinations in S72, the management device determines the switch device 1 controlled by each of the takeover destination control devices in S74. When only one control device as the takeover destination is selected in S73, the selected control device controls all the switch devices 1 controlled by the control device that has failed.

  For example, in S74, the management device 3 determines the number of switch devices controlled by each control device as the takeover destination in accordance with the ratio of the CPU load factor margins. For example, if the control device that has failed controls 100 switch devices 1 and the margins of the control devices A and B as takeover destinations are 30% and 10%, respectively, the control device A has 75 switches. The control of the device 1 is taken over, and the control device B takes over the control of the 25 switch devices 1. The fraction is adjusted by any method such as rounding off.

  At this time, the management device 3 can determine a takeover destination for each switch capacity of each switch device. For example, of the 100 switch devices 1 controlled by the control device that has failed, there are 40 switches having a capacity of 100 Gbps, 40 switches having a capacity of 40 Gbps, and a capacity of 10 Gbps. There are 20 switches. In this case, assuming that the margins of the control devices A and B as takeover destinations are 30% and 10%, respectively, the control device A includes 30 switches having a capacity of 100 Gbps and 30 switches having a capacity of 40 Gbps. The control of 15 switches having a capacity of 10 Gbps will be taken over. A switch with a large switch capacity is involved in many flows, and the frequency at which the control device performs control depends on the switch capacity. Therefore, it is possible to suppress the bias of the control load of the switch device by considering the switch capacity. it can.

  FIG. 8 is a schematic configuration diagram of the management apparatus according to the present embodiment. The storage unit 34 holds the management information described with reference to FIG. The determination unit 31 performs the failure determination process described with reference to FIGS. The selection unit 32 performs the takeover destination determination process described with reference to FIG. As described in S8 of FIGS. 3 and 4, the notification unit 33 takes over the control of the communication device of the group corresponding to the control device in the failed state to the control device of the takeover destination selected by the selection unit 32. To be notified.

  The management apparatus 3 according to the present invention can be realized by a program that causes a computer to operate as the management apparatus 3. These computer programs can be stored in a computer-readable storage medium or distributed via a network. The management device 3 may be realized by a single computer or a plurality of computers that can communicate with each other.

Claims (15)

In a communication system in which communication devices belonging to a network are grouped into a plurality of groups, a plurality of control devices respectively corresponding to the plurality of groups are provided, and each control device controls a communication device of a corresponding group, the plurality of control devices A management device for managing
Determining means for determining a failure state of the plurality of control devices based on a first signal transmitted from the plurality of control devices to the management device and indicating that a transmission source control device is normal ;
Selection means for selecting one or more control devices that take over control of a communication device of a group corresponding to a control device that has failed, based on a load value indicating a load state of the plurality of control devices;
A notifying means for notifying the takeover destination control device selected by the selection means that the control of the communication device of the group corresponding to the control device in the failed state is taken over;
Equipped with a,
If the determination means does not receive the first signal from the first control device of the plurality of control devices, the second control device is different from the first control device of the plurality of control devices. Alternatively, the management apparatus determines whether or not the first control apparatus is in a failure state by using a host apparatus connected to a communication apparatus of a group corresponding to the first control apparatus.   When the selection unit selects a plurality of control devices as a takeover destination control device, the selection unit further includes a determination unit that determines a communication device to which the takeover destination control device takes over control based on a load value of the takeover destination control device. The management apparatus according to claim 1, wherein:   The determining means sets the difference between the load value of the control device at the takeover destination and a predetermined threshold as a margin, and the control device at the takeover destination performs control according to a ratio of the margin of the control device at the takeover destination. The management apparatus according to claim 2, wherein a communication apparatus to be taken over is determined.   The determination unit classifies communication devices of a group corresponding to the control device in the failed state for each switch capacity, and performs the takeover for each switch capacity according to a margin ratio of the control device of the takeover destination. The management apparatus according to claim 3, wherein the previous control apparatus determines a communication apparatus to take over control.   The selection means distributes the load value of the control device that has become the failure state to the control device of the takeover destination, and adds the assigned load value to the load value of the control device of the takeover destination. The management apparatus according to claim 1, wherein the control apparatus of the takeover destination is selected so as not to exceed a threshold value. The determination unit, when the pre-Symbol first controller does not receive the first signal, prior Symbol second controller performs survival confirmation request of said first control device, from said second control device It is determined whether or not the first control device is in a failure state based on the number that the first control device indicated in the received report is normal and the number that the first control device is abnormal. The management apparatus according to any one of claims 1 to 5. The determination unit, when the pre-Symbol first controller does not receive the first signal, the presence confirmation signal to determine a response to the host apparatus connected to the communication device of the group corresponding to the first control device 6. The method according to claim 1, further comprising: determining whether the first control device is in a failure state based on the transmitted survival confirmation signal and the number of responses to the survival confirmation signal. The management device according to any one of the above.   The management device according to claim 1, wherein the load value is a load factor of a processor of the control device. In a communication system in which communication devices belonging to a network are grouped into a plurality of groups, a plurality of control devices respectively corresponding to the plurality of groups are provided, and each control device controls a communication device of a corresponding group, the plurality of control devices A management device for managing
If a signal indicating normality is not received from a first control device among the plurality of control devices, the first control device other than the first control device among the plurality of control devices receives the first control device. Based on the number that the first control device is normal and the number that is abnormal as indicated in the report received from the second control device. A management device comprising: a determination unit that determines whether or not the control device is in a failure state. The determination means compares the ratio of the number of abnormalities with respect to the total of the number of normal and abnormal numbers of the first control device with a threshold value, so that the first control device The management apparatus according to claim 9, wherein it is determined whether or not a failure state occurs. In a communication system in which communication devices belonging to a network are grouped into a plurality of groups, a plurality of control devices respectively corresponding to the plurality of groups are provided, and each control device controls a communication device of a corresponding group, the plurality of control devices A management device for managing
If a signal indicating normality is not received from a first control device among the plurality of control devices, a response is requested to a host device connected to a communication device of a group corresponding to the first control device. A determination unit that transmits a survival confirmation signal and determines whether or not the first control device is in a failure state based on the transmitted survival confirmation signal and the number of responses to the survival confirmation signal; Management device characterized by. The said determination means determines whether the said 1st control apparatus is in a failure state by comparing the ratio of the number of responses with respect to the number of transmitted survival confirmation signals with a threshold value. The management device described. If the determination means determines that the first control device is in a failure state, one or more control devices that take over control of the communication devices of the group corresponding to the first control device are assigned to the plurality of control devices. Selecting means for selecting based on a load value indicating a load state;
A notification means for notifying the control device of the takeover destination selected by the selection means that the control of the communication device of the group corresponding to the first control device is taken over;
The management apparatus according to claim 9, further comprising:   The management apparatus according to any one of claims 1 to 13, wherein the management apparatus is realized by one or more computers.   15. A program that causes a computer to function as the management device according to claim 1.

Images