JP5613142B2 - Distribution router, multicast distribution system, and multicast distribution method - Google Patents

Description

  The present invention relates to a distribution router, a multicast distribution system, and a multicast distribution method.

IGMP (Internet Group Management Protocol) / MLD (Multicast Listener Discovery) exists as a protocol for controlling IP multicast distribution between an IP (Internet Protocol) router and an IP host section. Among them, both the source IP address (hereinafter referred to as S address: Source address) of the distribution flow and the destination IP address (hereinafter referred to as G address: Group address) (hereinafter referred to as (S, G)) are set as flow IDs. There are IGMPv3 (IPv4) described in Non-Patent Document 1 and MLDv2 (IPv6) described in Non-Patent Document 2 as protocols corresponding to the SSM (Source-Specific Multicast) system that performs control by designating as.
In IGMPv3 / MLDv2, the router starts / stops the delivery of the flow upon receiving a report from the host. Note that the report includes the (S, G) address of the flow and the delivery status (Join indicating delivery / Leave indicating non-delivery).

  When the IP router performs IP multicast distribution control using IGMP / MLD, there is a configuration in which interfaces of a plurality of IP routers (hereinafter abbreviated as IF (Interface)) are redundantly connected to a single shared segment. Hereinafter, IFs of a plurality of routers connected to the same shared segment are referred to as IF sets. The shared segment is configured by, for example, an Ethernet (registered trademark) VLAN (Virtual Local Area Network).

  Here, when a plurality of routers are connected to the shared segment, when all routers receive a report and start distribution, the same flow is duplicated. In this case, since the bandwidth utilization efficiency of the segment is lowered, there is a possibility that the maximum number of flows that can be simultaneously distributed may be reduced, or a host that receives the same packet twice may be adversely affected.

Therefore, IGMPv3 / MLDv2 has a querier selection mechanism. That is, distribution control is performed only for one IF (hereinafter referred to as querier) in the IF set. On the other hand, the non-querier stops the flow distribution to the shared segment. In the querier selection mechanism, the querier and non-querier replacement processing in the IF set is executed at the following timing.
(Timing 1) When the querier receives a query (GQ (General Query) or SQ (Group-Specific Query)) from another router, the querier compares the source IP address with the IP address of its own IF to determine the source address. If is smaller, demote yourself to a non-querier.
(Timing 2) When a non-querier does not receive a GQ before exceeding the timer upper limit (also called OQPT (Other Querier Present Interval)) from the querier, it assumes that a failure has occurred in the querier and acts on behalf of the querier Promote yourself to querier.

IETF, “Internet Group Management Protocol, Version 3”, [online], [searched on November 7, 2011], Internet (URL: http://www.rfc-editor.org/rfc/rfc3376.txt) IETF, “Multicast Listener Discovery Version 2 (MLDv2) for IPv6”, [online], [searched on November 7, 2011], Internet (URL: http://www.rfc-editor.org/rfc/rfc3810. TXT)

The conventional querier selection mechanism described above realizes the alternate processing of the querier and the non-querier by a simple process of (Trigger 1) and (Trigger 2). With the change, the flow distribution process becomes unstable.
Specifically, the instability means that there are 0 queriers in the IF set, a “distribution interruption” in which the flow is not distributed to the host, or 2 or more queriers in the IF set. This is “overlapping delivery” in which the same flow is delivered separately to the host.
In other words, the querier selection mechanism is not executed with precise time synchronization / sequential synchronization, but with independent timing triggered by the reception of a query sent from another IF and a report sent from the host. Therefore, the flow distribution process becomes unstable in the process of performing the alternation process between the querier and the non-querier.

Hereinafter, examples of the configuration change of the IF set will be described as (Change 1) to (Change 4), respectively.
(Change 1) When a new IF is added to an IF set, the added IF is once activated as a querier even though a querier already exists in the IF set of the addition destination. Therefore, since the querier that already exists and the querier of the added IF are duplicated in the same IF set, a new querier is selected from the added IF and the querier that already exists ( Duplicate delivery occurs in the period until the trigger 1) is executed.

  (Change 2) When a predetermined IF is removed from an IF set, when the removal target is a querier, in a period from when the querier is removed until a new querier is selected between non-queriers, The querier becomes 0 units and distribution interruption occurs.

(Change 3) When a failure occurs in a querier, in order for a non-querier to detect the failure occurrence querier, a timer upper limit value (for example, about 256 seconds) determined by a standard rule when a GQ packet is not received from the querier ) Must be waited until GQ packet is exceeded. Therefore, in the period until the non-querier acts as the failure querier, the number of queriers becomes zero and distribution interruption occurs.
Note that if the timer upper limit value is shortened, the transmission frequency of the corresponding GQ packet is increased, which causes a side effect of increasing the load of the IF set querier and other IFs.

  (Change 4) When the querier recovers from the failure of (Change 3), the recovered querier and the querier that has been acting as the querier in (Change 3) distribute the flows, so the queriers are consolidated into one. Until then, duplicate delivery will occur.

  Therefore, the present invention limits the range in which flow distribution processing becomes unstable even if a configuration change occurs in a set of interfaces in a configuration in which interfaces of a plurality of multicast distribution routers are connected to a shared segment. The main purpose.

In order to solve the above-described problem, the present invention is configured such that a plurality of distribution routers that distribute a multicast flow to a host and one or more hosts are connected to a shared segment, and a packet transmitted to the shared segment includes: The distribution router used in a multicast distribution system that can be received by the distribution router and the host connected to the shared segment,
For each interface of the distribution router connected to the shared segment, an IF state management for calculating an IF hash value by calculating a first hash function that projects the identification information of the interface to a predetermined hash space as an argument And
For each flow distributed by the distribution router via the shared segment, a flow hash value is calculated by calculating a second hash function that projects the identification information of the flow onto the predetermined hash space as an argument. ,
A predetermined area in the hash space delimited by each calculated IF hash value is set as an assigned area of the predetermined IF hash value, and a flow corresponding to the flow hash value included in the assigned area is set as the predetermined IF hash Control that the interface of the distribution router corresponding to the value is responsible for distribution,
When newly adding the IF hash value to the predetermined hash space, the area in charge of the IF hash value to be added is an area obtained by dividing one of the areas in charge of the IF hash value,
When deleting the IF hash value from the predetermined hash space, it has a flow allocation processing unit that integrates the area in charge of the IF hash value to be deleted into another area in charge of the IF hash value. And

  As a result, it is only necessary to accommodate and change the flow that is close to the IF hash value in which the configuration change has occurred, so that it is possible to locally influence the instability of the flow distribution process. Therefore, in the configuration in which the interfaces of a plurality of multicast distribution routers are connected to the shared segment, it is possible to limit the range in which the flow distribution processing becomes unstable even when a configuration change occurs in those interface sets.

  In the present invention, the IF state management unit receives the multicast query message from the distribution router flowing in the shared segment, and calculates the IF hash value for the source IP address included in the query message. It is used as identification information of an interface.

  Thereby, the packet format (query) of the existing protocol can be used as it is.

  In the present invention, the flow allocation processing unit receives a multicast report message from the host that flows in the shared segment, and the source IP address of the distribution flow included in the report message and the destination IP address of the distribution flow At least one of the IP addresses is used as identification information of the flow for calculating the flow hash value.

  Thereby, the packet format (report) of the existing protocol can be utilized as it is.

  According to the present invention, in a configuration in which interfaces of a plurality of multicast distribution routers are connected to a shared segment, it is possible to limit a range in which flow distribution processing becomes unstable even if a configuration change occurs in those interface sets. it can.

It is a block diagram which shows the multicast delivery system regarding one Embodiment of this invention. It is a block diagram which shows the delivery router regarding one Embodiment of this invention. It is explanatory drawing which shows the ring-shaped hash space regarding one Embodiment of this invention. It is a flowchart which shows the delivery process regarding one Embodiment of this invention, and its preparatory process. It is explanatory drawing which shows that both hash values have been arrange | positioned in the ring-shaped hash space regarding one Embodiment of this invention. It is a flowchart which shows the structure change process of IF set regarding one Embodiment of this invention. It is explanatory drawing which shows the structural change example of IF set in the ring-shaped hash space regarding one Embodiment of this invention. It is a flowchart which shows the process of the flow allocation process part regarding one Embodiment of this invention.

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

FIG. 1 is a configuration diagram showing a multicast distribution system.
The multicast flow is transmitted from the distribution server 1 (S1, S2,...), Passes through the multicast distribution network 2, the link 3, the distribution router 4 (R1, R2, R3,...), The shared segment 5, and the host 6 (H1). , H2, H3, H4,. The shared segment 5 is configured as an Ethernet segment (multicast VLAN), for example.
Each device of these multicast distribution systems is configured as a computer having a CPU (Central Processing Unit), a memory, a hard disk (storage means), and a network interface, and the computer executes a program read on the memory. Thus, each processing unit is operated.

  Packets between the distribution router 4 and the host 6 are transmitted / received via the shared segment 5. The transmitted / received packet includes a query transmitted from the distribution router 4 and a report transmitted from the host 6 (Join indicating distribution / Leave indicating non-distribution). As described in RFC (Request for Comments) 3810, when “Record Type” in the ICMPv6 header of the report is set to 5 (ALLOW_NEW_SOURCES), it becomes Join for the designated (S, G), and “Record Type” is 6 When set to (BLOCK_OLD_SOURCES), it becomes Leave for the designated (S, G).

FIG. 2 is a configuration diagram showing the distribution router 4.
The distribution router 4 includes an MC distribution network connection interface 41 connected to the multicast distribution network 2, an MC distribution network frame transfer unit 42 that performs frame transmission / reception with the multicast distribution network 2, and a shared segment frame transfer unit 44 that performs frame transmission / reception with the shared segment 5. , A shared segment connection interface 45 connected to the shared segment 5, and a transfer control circuit 43 that controls flow distribution.

The MC distribution network connection interface 41 and the shared segment connection interface 45 control a general physical layer, and correspond to a PHY layer in the case of Ethernet.
The MC distribution network connection interface 41 includes a reception IF unit 411 and a transmission IF unit 412.
The shared segment connection interface 45 includes a transmission IF unit 451 and a reception IF unit 452.

The MC distribution network frame transfer unit 42 and the shared segment frame transfer unit 44 serve as a general data link layer, and correspond to a MAC (Media Access Control) layer in the case of Ethernet.
The MC distribution network frame transfer unit 42 includes a frame reception unit 421 and a frame transmission unit 422. The shared segment frame transfer unit 44 includes a frame transmission unit 441 and a frame reception unit 442.

The transfer control circuit 43 includes a distribution state management unit 431, an MC distribution network cooperation unit 432, an IF state management unit 433, an IF management table 434, a flow allocation processing unit 435, a flow distribution table 436, and a distribution filter unit 437.
When the report is received from the host 6, it is notified to the flow allocation processing unit 435 via the shared segment connection interface 45, the shared segment frame transfer unit 44, and the distribution state management unit 431.
The distribution state management unit 431 corresponds to a conventional IGMP / MLD router or the like, and manages the access distribution state of the IP multicast flow. Here, the delivery status management unit 431 performs a linkage process related to delivery via the MC delivery network linkage unit 432 according to the content of the report received from the host 6 and the current flow delivery status.
The MC distribution network cooperation unit 432 corresponds to a conventional PIM router or the like. IP multicast routing is controlled in cooperation with the multicast distribution network 2.
In response to the instruction from the flow allocation processing unit 435, the distribution filter unit 437 opens and closes the distribution filter for each flow. The flow whose filter is released transfers the flow packet received from the MC distribution network frame transfer unit 42 to the shared segment frame transfer unit 44. On the contrary, the flow closed by the filter is discarded without being transferred.

The IF state management unit 433 manages information on all IFs connected to the shared segment. Checks the source IP address and reception time of a query from another IF received via the shared segment connection interface 45, the shared segment frame transfer unit 44, and the distribution status management unit 431, and calculates an IF hash value from the identification information of the IF (Details are described in Table 1 to be described later), and other IFs are monitored for life and death.
The IF management table 434 is a database in which IF information checked by the IF state management unit 433 is registered and managed.
The IF state management unit 433 reads the interface information registered in the IF management table 434 in accordance with an instruction from the flow allocation processing unit 435 and delivers it to the flow allocation processing unit 435.

The flow allocation processing unit 435 manages IF allocation information for flows in the flow distribution table 436. The flow allocation processing unit 435 calculates a flow hash value based on the flow identification information read from the report (details will be described later in Table 2). The flow allocation processing unit 435 acquires all IF hash values from the IF state management unit 433, calculates the directional distance between the flow hash value and all IF hash values, and determines the distribution IF allocation for the flow. Control delivery status. Here, these pieces of processing information are registered in the flow distribution table 436.
The flow distribution table 436 is a database in which allocation information and distribution states for interfaces for each flow processed in the flow allocation processing unit 435 are registered and managed.

FIG. 3 is an explanatory diagram showing a ring-shaped hash space used by both tables (IF management table 434 and flow distribution table 436).
FIG. 3A shows the hash space when a constant n is set for determining the upper limit of the hash value, which is the size of the hash space. The constant n is the maximum hash value + 1. The hash space is configured so that the hash value increases clockwise (clockwise), the direction at 12 o'clock (directly above) is the hash value “0” and the hash value “n”, and the direction at 6 o'clock (directly below). The hash value “n / 2” is indicated.
FIG. 3B shows an example when n = 1000 as the hash space of FIG.

Table 1 shows an IF management table 434 generated for each shared segment 5. The IF management table 434 includes an IF ID that is identification information of the IF, an IP address of the IF, an IF hash value (Hash) of the IF, an IF flag indicating whether the IF is its own, and an IF activity A querier alive monitoring timer for monitoring whether or not it is being stored is stored in association with each other.
For example, the IFID “R1” is an IF accommodated by itself.
The IFIDs “R2, R3” are IFs accommodated by other devices, and are active because the time elapsed since the query was received (the time of the querier alive monitoring timer) is less than or equal to a predetermined time.
The IFID “R4” is an IF accommodated by another device and is not active because the time elapsed since the query was received has exceeded a predetermined time (timed out).

The IF hash value of the IF management table 434 is a result obtained by performing a remainder calculation (mod) on the IP address of the IF by a constant n. The IP address of each IF can be obtained by reading the source IP address (IPv6 link local address) described in the IPv6 header of the query.
In the following, the IP address of the IF is used as an argument of the hash function. However, the identification information of other IF (part of the source IP address, the ID of the IF set by a predetermined rule, the shared segment 5 The IF querier order (the smallest IP address in the shared segment 5) may be used instead as an argument.

Hereinafter, the size of the hash space is “n”, the IP address of the IF that is the hash argument is “IP”, the hash function is “H ()”, the hash argument conversion function is “f ()”, Hexadecimal representation is expressed as “() h”.
H (IP, n) = mod [f (IP), n]
For example, when “fe80 :: 1122 :: 3344 :: 5566 :: 789a” where n = 1000 and IP = R1
H (IP, n)
= mod [(fe80 + 1122 + 3344 + 5566 + 789a) h, 1000]
= mod [(65152 + 4386 + 13124 + 21862 + 30874), 1000]
= mod [135398,1000] = 398

Table 2 shows a flow distribution table 436 generated for each shared segment 5. The flow distribution table 436 stores a flow ID, a flow S address, a flow G address, a flow hash value (Hash), an IF in charge of flow distribution, and a flow distribution state in association with each other. . The distribution state of the flow takes either the value of distribution (Join) or non-distribution (Leave).
For example, the flow with the flow ID “F1” is distributed from the IF “R3” and is not distributed from other IFs.

The flow hash value of the flow distribution table 436 is a result of the remainder (mod) calculation of the (S, G) address information of the flow by a constant n. The S address can be read from “Multicast Address” of the ICMPv6 header in the report, and the G address can be read from “Source Address [1] to [N]” of the ICMPv6 header in the report.
In the following, the flow (S, G) address information is used as the argument of the hash function, but other flow identification information (S address alone, G address alone, etc.) is used instead. May be used as an argument.

In the following, the size of the hash space is “n”, the (S, G) address information of the flow that is the hash argument is “S, G”, the hash function is “H ()”, and the hash argument conversion function is “F ()” is expressed as a hexadecimal expression “() h”.
H (S, G, n) = mod [f (S, G), n]
For example, n = 1000, S = F1, "2001 :: 1000 :: 2000 :: 3000 :: 4000", and G = F1, "ff3e :: 0001 :: 0002 :: 0003 :: 0004"
H (S, G, n)
= mod [(2001 + 1000 + 2000 + 3000 + 4000) h + (ff3e + 0001 + 0002 + 0003 + 0004) h, 1000]
= mod [(8193 + 4096 + 8192 + 12288 + 16384) + (65342 + 1 + 2 + 3 + 4), 1000]
= mod [(49153 + 65352), 1000] = 114

For each of the IF hash value and flow hash value described above, the network administrator may manually set a hash key and a hash value.
Note that the hash function does not necessarily need to be a residue calculation, and other commonly used hash functions may be used. However, in order to maintain consistency of distribution control within the same shared segment 5, it is better to use a hash value calculation method that is common among the distribution routers 4 connected to the same shared segment 5.

  In the calculation of the flow hash value, even if two IFs (two arguments) are different from each other, a hash collision may occur in which the hash values in the hash space projected by the hash function coincide by chance. . When a hash collision occurs, a network administrator is warned of the collision by an SNMP (Simple Network Management Protocol) command or the like.

Alternatively, the flow allocation processing unit 435 may execute the hash collision avoidance process shown in the following (1) to (4) when a hash collision occurs.
(1) All IF hash values that do not collide with each other are placed in the hash space as they are.
(2) Among a plurality of conflicting IF hash values, the IF hash value with the smallest IP address used for calculation of the IF hash value is arranged in the hash space as it is. The collision determination (2) is executed every time the configuration of the IF set is changed (IF addition or IF failure recovery).
(3) Among the plurality of IF hash values that collide in (2) above, the IF hash value that is not arranged in the hash space in (2) is located at a location different from the IF hash value (for example, IF hash value From the starting point to the place moved in order n / 8 minutes clockwise) and placed in the hash space.
(4) When an IF hash value already placed in the hash space is deleted due to a change in the configuration of the IF set (IF reduction or IF failure occurs), the deleted IF hash value is changed to the hash space in (2). If the IF hash value has been arranged, the IF hash value corrected and arranged in (3) is returned to the IF hash value before correction. In order to realize (4), a data structure (for example, a list structure) for managing a plurality of conflicting IF hash values is prepared.

FIG. 4 is a flowchart showing the distribution process and its preparation process.
As S101, each distribution router 4 initializes both tables (an IF management table 434 generated by the IF state management unit 433 and a flow distribution table 436 generated by the flow allocation processing unit 435) in response to power-on to itself. Turn into. Note that both tables initialized in S101 are deleted, for example, when the distribution router 4 is shut down.
The IF state management unit 433 adds a new IF entry to the IF management table 434. The entry contents are an IP address of the own IF, an IF hash value of the IP address, an IF flag of “own IF” indicating the own IF, and invalidation of the querier timer (GQ reception alive monitoring).

As S102, the distribution router 4 (R1) transmits a query to the shared segment 5 and receives a query from another IF.
In S103, the IF state management unit 433 of the distribution router 4 (R2) registers the IF described in the query received in S102 in its own hash space.
Specifically, the IF state management unit 433 uses the transmission source IP address read from the IP header of the received packet as a search key, and the search key entry (hereinafter referred to as a query entry) exists from the IF management table 434. Search whether or not.
When there is a query entry, the querier alive monitoring timer of the query entry is set to zero.
When there is no query entry, a new query entry is added to the IF management table 434. The contents of the entry are the query source IP address, the IF hash value of the source IP address, the “other IF” IF flag indicating the other IF, and the querier alive monitoring timer = 0.

FIG. 5 is an explanatory diagram showing that both hash values (IF hash value, flow hash value) are arranged in the ring-shaped hash space shown in FIG. FIG. 5A is a diagram in which three IF hash values are arranged in the hash space.
As an example of the hash space registration process of S103, the IF hash value “136” of the distribution router 4 “R3”, the IF hash value “398” of the distribution router 4 “R1”, and the IF hash of the distribution router 4 “R2” The value “767” is arranged.
Then, a flow is assigned based on the direction distance in the hash space. For example, the hash space that is in charge of distribution of the distribution router 4 “R1” is clockwise from the IF hash value “136” +1 of the distribution router 4 “R3” that is located one before (clockwise). The range is up to its own IF hash value “398”.

Returning to FIG. 4, in S111, the host 6 returns a report for the query. Here, it is assumed that the report includes a Join indicating a distribution request for the flows F1 to F4.
As S112, the flow allocation processing unit 435 of each distribution router 4 refers to the report of S111 and registers F1 to F4 in its own hash space (refer to FIG. 8 for details). As a result, the distribution router 4 (R1) is in charge of distributing the flows F2 and F3, and the distribution router 4 (R2) is not in charge of distributing the flow.

FIG. 5B shows an example in which flow hash values are arranged in the hash space of FIG. For example, the flow hash value “114” of the flow F1 is arranged in the vicinity of the IF hash value “136” of the distribution router 4 “R3”.
The flow allocation processing unit 435 determines an IF in charge of each flow based on the direction distance between the IF hash value and the flow hash value. For example, as shown in FIG. 5A, the hash space in charge of distribution of the distribution router 4 “R1” starts from the IF hash value “136” +1 of the distribution router 4 “R3” in the clockwise direction. Therefore, the flow F5 of the flow hash value “177” included in this range is in charge of distribution of the distribution router 4 “R1”.

Hereinafter, a method for determining an IF hash value for each flow hash value from the hash space in which the IF hash value “R” and the flow hash value “F” are calculated will be described.
The hash distance D (R, F) = RF, and for each F, if there is one or more D (R, F) ≧ 0, R is the minimum value of D (R, F) ≧ 0 Is assigned as the person in charge of F distribution. On the other hand, when D (R, F) ≧ 0 does not exist, R having the minimum value of D (R, F) is assigned as the person in charge of F distribution.

  Returning to FIG. 4, as S <b> 113, the distribution server 1 distributes the data of the flow F <b> 2 specified by Join in S <b> 111 to the shared segment 5. This data can be received by all the distribution routers 4 connected to the shared segment 5, but this time the distribution router 4 (R1) in charge of distribution of the flow F2 sends the flow F2 to the host 6 in S112. To deliver.

In step S121, the host 6 returns a report for the query in the same manner as in step S111. Here, it is assumed that the report includes a Join indicating a distribution request for the flow F5.
As S122, the flow allocation processing unit 435 of each distribution router 4 refers to the report of S121 and registers the flow F5 in its own hash space (see FIG. 8 for details) as in S112. As a result, the distribution router 4 (R2) is in charge of distribution of the flow F5.
In S123, as in S113, the distribution router 4 (R2) distributes data of the flow F5 that is in charge of distribution.

In S131, the host 6 returns a report for the query. Here, it is assumed that the report includes Leave indicating a withdrawal request for the flow F5.
In S132, the flow allocation processing unit 435 of each distribution router 4 determines that when the host 6 (host 6 that continues to request distribution of the flow F5) other than the host 6 that has received Leave this time does not exist in the shared segment 5, The flow F5 is deleted from the hash space of the flow distribution table 436.
Instead of deleting an entry in the flow distribution table 436, the entry shifts to a state of being inactive (Leave), and when a Join for the entry is received later, the entry from the inactive state (Leave) to active (Join State transition to).
As described above, storing the suspended entry in the flow distribution table 436 increases the size of the flow distribution table 436, but the state transition is performed without re-registering the suspended entry. As a result, distribution control can be resumed early.

FIG. 6 is a flowchart showing the IF set configuration change processing.
As S201, the flow allocation processing unit 435 calculates the IF of the distribution router 4 in charge of each joined flow, as described in S112, S122, and S132 in FIG. Here, the distribution router 4 (R1) is in charge of the flow F5, the distribution router 4 (R2) is in charge of the flows F2 and 3, and the distribution router 4 (R3) is in charge of the flows F1 and 4.

Assume that the network administrator adds a distribution router 4 (R4) to the shared segment 5 (or recovers from a failure) in S211.
In S212, the distribution router 4 (R4) notifies the other distribution routers 4 (R1 to R3) of its own addition by transmitting a query including an IP address indicating its own IF to the shared segment 5.
As S213, the IF state management unit 433 of the other distribution router 4 (R1 to R3) registers the IF of the distribution router 4 (R4) notified in S212 in the hash space (the same processing as S103).

As S214, the flow allocation processing unit 435 of each distribution router 4 calculates the flow charge in the updated hash space in accordance with the configuration change of the IP set. As a result, as the change from S201, the charge of the flow F4 is changed from R3 to R4, but the charge of the other flows F1 to F3 is not changed.
As described above, the reason for the small change in charge of the flow is that the IF hash value is arranged in the hash space on the ring as shown in FIG. That can be done.

FIG. 7 is an explanatory diagram illustrating an example of a configuration change of an IF set in a ring-shaped hash space. FIG. 7A shows a hash space when the IF (IF hash value = 965) of the distribution router 4 (R4) is further expanded (or trouble recovered) from the state of FIG. 5B.
The IF range (R2 → R3) of the previous distribution router 4 (R3) is the IF range (R2 → R4) of the current distribution router 4 (R4) and the IF of the current distribution router 4 (R3). The assigned range (R4 → R3).

Returning to FIG. 6, it is assumed that, in S221, a failure occurs in the distribution router 4 (R4) (or is removed), and a query from the distribution router 4 (R4) becomes impossible to be transmitted.
In S222, the IF state management unit 433 of the other distribution router 4 (R1 to R3) detects a timeout of query reception from the distribution router 4 (R4), and deletes the IF of the distribution router 4 (R4) from the hash space. To do.
Specifically, the IF status management unit 433 receives a query from the interface even if the upper limit value of the querier alive monitoring timer set in advance is reached for each entry of the router interface registered in the IF management table 434. It is detected and determined as a timeout. Then, the entry exceeding the timer upper limit value is deleted.
Here, instead of deleting the entry in the IF management table 434 that has exceeded the timer upper limit value, the IF state management unit 433 may shift the entry to a dormant state (the same state transition process as in S132).
In S223, the flow allocation processing unit 435 calculates the flow assignment in the updated hash space in accordance with the IP set configuration change (IF deletion) in S222. As a result, when the person in charge of the flow F4 returns from R4 to R3, the process returns to the state in charge of distribution in S201.

  FIG. 7B shows a hash space when a failure occurs in IF “R4” (or when it is removed) from the state of FIG. By deleting the IF hash value (965) of the distribution router 4 (R4), the IF range (R2 → R4) of the previous distribution router 4 (R4) and the IF of the previous distribution router 4 (R3) The assigned range (R4 → R3) is integrated into the IF assigned range (R2 → R3) of the current distribution router 4 (R3).

  FIG. 8 is a flowchart showing flow allocation processing (S112, S122 in FIG. 4) executed by the flow allocation processing unit 435. Since the operating subject of this flowchart is the flow allocation processing unit 435, the description of the operating subject is omitted in the description of each processing step.

In S301, it is searched whether or not an existing entry for the flow (S, G) indicated by Join received from the host 6 exists in the flow distribution table 436. If Yes in S301, the process proceeds to S303, and if No, the process proceeds to S302.
In S302, the flow (S, G) is registered as a new entry in the flow distribution table 436. The entry contents are the address of the flow (S, G), the flow hash value calculated from the (S, G) address, and the distribution responsible IF in the hash space corresponding to the flow hash value.

In S303, it is determined whether or not the flow (S, G) distribution state in the flow distribution table 436 is “Leave” to determine whether or not the flow is in an undistributed state. If Yes in S303, the process proceeds to S304, and if No, the process of FIG.
In S304, it is determined whether or not it is in charge of distribution by referring to whether or not the IF column of the flow (S, G) in the flow distribution table 436 is accommodated. If Yes in S304, the process proceeds to S305, and if No, the process of FIG.
As S305, distribution of the flow (S, G) is started. Specifically, after setting the distribution state in the flow distribution table 436 to “Join”, the distribution filter unit 437 is instructed to open the filter, and the distribution server 1 is requested to distribute via the MC distribution network cooperation unit 432 ( PIM-Join) is transmitted.

  In the present embodiment described above, as shown in FIG. 1, a plurality of distribution routers 4 are connected to a single shared segment 5, and a distribution person is assigned in units of flow among the plurality of distribution routers 4. In this way, access distribution in the multicast distribution process is realized. In other words, not only is the failure risk distributed by the redundant configuration of the entire system, but also each flow in charge of distribution is distributed between the distribution routers 4 even when a failure has not occurred, thereby improving access accommodation efficiency.

On the other hand, if the person in charge of distribution is not assigned to each flow unit, the computing resources of the distribution router 4 necessary for packet copy processing and the communication resources of the distribution path (bandwidth) on the shared segment 5 necessary for packet distribution processing are limited. Therefore, the load is concentrated on a specific resource.
As a result, the resource utilization efficiency of the entire system is reduced, and deterioration of service quality such as an increase in packet loss and delay (fluctuation) due to congestion reduces the convenience for the user. Here, if the load concentration is easily relaxed by system enhancement, an excessive cost is generated.

Each distribution router 4 defines a ring-type hash space as shown in FIG. 5 and calculates both an IF hash value and a flow hash value so as to fit in the hash space. Based on the direction distance between the IF hash value and the flow hash value, the person in charge of distribution for each flow is specified.
Therefore, even if the configuration change of the distribution router 4 connected to the shared segment 5 occurs due to each event such as increase / decrease / failure switching / recovery switchback of the distribution router 4, Since the flow allocation based on the directional distance is recalculated, it is only necessary to accommodate and change the flow close to the IF hash value in which the configuration change has occurred. it can. Therefore, it is possible to provide an IP multicast distribution service that can flexibly cope with fluctuations in facility demand while maintaining the quality and reliability of distribution.

  It should be noted that by using IF identification information (IP address, etc.) and flow identification information ((S, G) address information, etc.) as an argument for calculating a hash value, the number of IFs is used as an argument for hash calculation. May not be included. In other words, even if the configuration change of the distribution router 4 occurs, the IF hash values of many distribution routers 4 that have not been changed in configuration do not change, and thus the influence of instability of flow distribution processing may be locally limited. it can.

  Furthermore, identification information that is an argument of hash calculation can be easily obtained by detecting and monitoring all the IFs connected and the control packet (multicast / broadcast) exchanged between the hosts. Therefore, since the packet format (query and report) of the existing protocol can be used as it is, each router can autonomously distribute each flow in a scalable and simple manner without exchanging special control information between the router / host. The person in charge of distribution can be identified.

  As described above, since a plurality of distribution routers 4 connected to a single shared segment 5 can flexibly cope with each distribution flow in response to a configuration change of the distribution router 4, the distribution quality Multicast distribution control that can withstand fluctuations in traffic demand while maintaining reliability is possible, and high-quality and highly reliable IP multicast distribution services such as terrestrial digital broadcast / BS digital broadcast IP retransmission services are realized. It becomes possible.

DESCRIPTION OF SYMBOLS 1 Distribution server 2 Multicast distribution network 3 Link 4 Distribution router 5 Shared segment 6 Host 41 MC distribution network connection interface 42 MC distribution network frame transfer part 43 Transfer control circuit 44 Shared segment frame transfer part 45 Shared segment connection interface 411 Reception IF part 412 Transmission IF unit 421 Frame reception unit 422 Frame transmission unit 431 Distribution state management unit 432 MC distribution network cooperation unit 433 IF state management unit 434 IF management table 435 Flow allocation processing unit 436 Flow distribution table 437 Distribution filter unit 441 Frame transmission unit 442 frame Reception unit 451 Transmission IF unit 452 Reception IF unit

Claims (6)

A plurality of distribution routers for distributing a multicast flow to a host and one or more of the hosts are connected to a shared segment, and a packet transmitted to the shared segment includes the distribution router connected to the shared segment and the The distribution router used in a multicast distribution system that can be received by each host,
For each interface of the distribution router connected to the shared segment, an IF state management for calculating an IF hash value by calculating a first hash function that projects the identification information of the interface to a predetermined hash space as an argument And
For each flow distributed by the distribution router via the shared segment, a flow hash value is calculated by calculating a second hash function that projects the identification information of the flow onto the predetermined hash space as an argument. ,
A predetermined area in the hash space delimited by each calculated IF hash value is set as an assigned area of the predetermined IF hash value, and a flow corresponding to the flow hash value included in the assigned area is set as the predetermined IF hash Control that the interface of the distribution router corresponding to the value is responsible for distribution,
When newly adding the IF hash value to the predetermined hash space, the area in charge of the IF hash value to be added is an area obtained by dividing one of the areas in charge of the IF hash value,
When deleting the IF hash value from the predetermined hash space, it has a flow allocation processing unit that integrates the area in charge of the IF hash value to be deleted into another area in charge of the IF hash value. And delivery router. The IF state management unit receives the multicast query message from the distribution router flowing in the shared segment, and identifies the interface identification information for calculating the IF hash value for the source IP address included in the query message The distribution router according to claim 1, wherein the distribution router is used as a distribution router. The flow allocation processing unit receives a multicast report message from the host that flows in the shared segment, and includes at least one of a source IP address of a distribution flow and a destination IP address of the distribution flow included in the report message The distribution router according to claim 1, wherein one IP address is used as identification information of the flow for calculating the flow hash value. The flow allocation processing unit
With respect to the interface group of the distribution router in which the calculated IF hash value is the same, one interface of the interface group is associated with the IF hash value calculated as the same value and the predetermined hash space Assigned to
The other interfaces in the interface group are assigned to the predetermined hash space by associating the IF hash value moved by a predetermined amount with the IF hash value calculated as the same value as a starting point,
When the IF hash value of the one interface is deleted from the predetermined hash space, the IF hash value associated with the other interface is returned to the IF hash value before movement in the predetermined hash space. The distribution router according to any one of claims 1 to 3, wherein the distribution router is characterized in that: A plurality of distribution routers for distributing a multicast flow to a host and one or more of the hosts are connected to a shared segment, and a packet transmitted to the shared segment includes the distribution router connected to the shared segment and the A multicast distribution system in which each host can receive,
Each of the distribution routers used in the multicast distribution system is
For each interface of the distribution router connected to the shared segment, an IF state management for calculating an IF hash value by calculating a first hash function that projects the identification information of the interface to a predetermined hash space as an argument And
For each flow distributed by the distribution router via the shared segment, a flow hash value is calculated by calculating a second hash function that projects the identification information of the flow onto the predetermined hash space as an argument. ,
A predetermined area in the hash space delimited by each calculated IF hash value is set as an assigned area of the predetermined IF hash value, and a flow corresponding to the flow hash value included in the assigned area is set as the predetermined IF hash Control that the interface of the distribution router corresponding to the value is responsible for distribution,
When newly adding the IF hash value to the predetermined hash space, the area in charge of the IF hash value to be added is an area obtained by dividing one of the areas in charge of the IF hash value,
When deleting the IF hash value from the predetermined hash space, it has a flow allocation processing unit that integrates the area in charge of the IF hash value to be deleted into another area in charge of the IF hash value. Multicast distribution system. A plurality of distribution routers for distributing a multicast flow to a host and one or more of the hosts are connected to a shared segment, and a packet transmitted to the shared segment includes the distribution router connected to the shared segment and the A multicast distribution method executed by a multicast distribution system that can be received by each host,
Each of the distribution routers used in the multicast distribution system has an IF state management unit and a flow allocation processing unit,
The IF state management unit calculates an IF hash for each interface of the distribution router connected to the shared segment by calculating a first hash function that projects the identification information of the interface onto a predetermined hash space as an argument. Calculate the value
The flow allocation processing unit
For each flow distributed by the distribution router via the shared segment, a flow hash value is calculated by calculating a second hash function that projects the identification information of the flow onto the predetermined hash space as an argument. ,
A predetermined area in the hash space delimited by each calculated IF hash value is set as an assigned area of the predetermined IF hash value, and a flow corresponding to the flow hash value included in the assigned area is set as the predetermined IF hash Control that the interface of the distribution router corresponding to the value is responsible for distribution,
When newly adding the IF hash value to the predetermined hash space, the area in charge of the IF hash value to be added is an area obtained by dividing one of the areas in charge of the IF hash value,
When deleting the IF hash value from the predetermined hash space, the area in charge of the IF hash value to be deleted is integrated into another area in charge of the IF hash value.

Images