METHODS, SYSTEMS, AND COMPUTER READABLE MEDIA FOR DISTRIBUTING SIGTRAN CONNECTIONS AMONG SIGNAL TRANSFER POINT (STP) MESSAGE PROCESSORS

Abstract

A method for distributing Sigtran connections among signal transfer point (STP) message processors includes providing a connection load balancer as a front end to plural message processors of an STP. The method further includes publishing, by the connection load balancer, an Internet protocol (IP) address to SS7 peers. The method further includes initializing the message processors of the STP to listen on the IP address published by the connection load balancer. The method further includes receiving, at the connection load balancer, a Sigtran message addressed to the IP address. The method further includes determining, by the connection load balancer, whether the Sigtran message is an initial message for a Sigtran connection or a subsequent message for a Sigtran connection and whether the Sigtran connection has been assigned to one of the message processors. The method further includes forwarding the message to one of the message processors or dropping the message based on whether the message is an initial message or a subsequent message and based on whether the connection has been assigned to one of the message processors.

Claims

1. A method for distributing Sigtran connections among signal transfer point (STP) message processors, the method comprising: providing a connection load balancer as a front end to plural message processors of an STP; publishing, by the connection load balancer, an Internet protocol (IP) address to SS7 peers; initializing the message processors of the STP to listen on the IP address published by the connection load balancer; receiving, at the connection load balancer, a Sigtran message addressed to the IP address; determining, by the connection load balancer, whether the Sigtran message is an initial message or a subsequent message for a Sigtran connection; determining, by the connection load balancer, whether the Sigtran message is for a Sigtran connection to which one of the message processors has been assigned; and forwarding the message to one of the message processors or dropping the message based on whether the message is an initial message or a subsequent message and whether the Sigtran message is for a Sigtran connection to which one of the message processors has been assigned.

2. The method of claim 1 wherein the message processors comprise physical message processors of a physical STP.

3. The method of claim 1 wherein the message processors comprise virtual message processors of a virtual STP.

4. The method of claim 1 comprising associating the IP address with loopback interface of the message processors.

5. The method of claim 1 wherein publishing the IP address to the SS7 peers includes broadcasting, by the connection load balancer, gratuitous address resolution protocol (ARP) or Internet control management protocol version 6 (ICMPv6) neighbor discovery messages that associate the IP address with a medium access control (MAC) address of the connection load balancer.

6. The method of claim 1 wherein the Sigtran message is an initial message for a Sigtran connection to which one of the message processors has not been assigned and wherein the method further comprises assigning one of the message processors to the Sigtran connection for which the initial message is intended to initiate establishment using a load balancing algorithm.

7. The method of claim 6 wherein assigning one of the message processors to the Sigtran connection using a load balancing algorithm includes a maintaining a group count for each of the message processors, where the group count is indicative of a number of Sigtran connections assigned to each message processor, calculating, for each message processor, a group count difference between the group count for the message processor and a lowest group count of the message processors, and including, as connection distribution candidates, message processors having a group count difference that is less than a link distribution threshold.

8. The method of claim 1 comprising detecting failure and reconnection of one the message processors and, in response, using the reconnected message processor as a candidate for Sigtran connection distribution.

9. The method of claim 1 wherein whether the message is an initial message or a subsequent message includes determining that the message is a subsequent message associated with a Sigtran connection, determining whether one of the message processors has been assigned to the connection includes determining that one of the message processors has been assigned to the connection and wherein forwarding or dropping the message includes forwarding the message to the message processor assigned to the connection.

10. The method of claim 1 comprising forwarding egress traffic from the message processors of the STP to the SS7 peers in a manner that bypasses the connection load balancer.

11. A system for distributing Sigtran connections among signal transfer point (STP) message processors, the system comprising: a signal transfer point (STP) including a plurality of message processors for routing Sigtran messages received on Sigtran connections; a connection load balancer that operates as a front end to the message processors of the STP, wherein: the connection load balancer is configured to publish an Internet protocol (IP) address to SS7 peers; the message processors are initialized to listen on the IP address published by the connection load balancer; the connection load balancer is configured to receive a Sigtran message addressed to the IP address, determine whether the Sigtran message is an initial message or a subsequent message for a Sigtran connection, determine whether the Sigtran message is for a Sigtran connection to which one of the message processors has been assigned, and forward the message to one of the message processors or drop the message based on whether the message is an initial message or a subsequent message and whether the Sigtran message is for a Sigtran connection to which one of the message processors has been assigned.

12. The system of claim 11 wherein the message processors comprise physical message processors of a physical STP.

13. The system of claim 11 wherein the message processors comprise virtual message processors of a virtual STP.

14. The system of claim 11 wherein the IP address is associated with a loopback interface of the message processors.

15. The system of claim 11 wherein the connection load balancer is configured to broadcast gratuitous address resolution protocol (ARP) or Internet control management protocol version 6 (ICMPv6) neighbor discovery messages that associate the IP address with a medium access control (MAC) address of the connection load balancer.

16. The system of claim 11 wherein the connection load balancer determines that the Sigtran message is an initial message for a Sigtran connection to which one of the message processors has not been assigned and assigns one of the message processors to the Sigtran connection for which the initial message is intended to initiate establishment using a load balancing algorithm.

17. The method of claim 6 wherein the connection load balancer is configured to maintain a group count for each of the message processors, where the group count is indicative of a number of Sigtran connections assigned to each message processor, and wherein the connection load balancer is further configured to calculate, for each message processor, a group count difference between the group count for the message processor and a lowest group count of the message processors, and include, as connection distribution candidates, message processors having a group count difference that is less than a link distribution threshold.

18. The system of claim 11 wherein the connection load balancer is configured to detect failure and reconnection of one the message processors and, in response, using the reconnected message processor as a candidate for Sigtran connection distribution.

19. The system of claim 11 wherein the message processors are configured to forward egress traffic from the STP to the SS7 peers in a manner that bypasses the connection load balancer.

20. A non-transitory computer readable medium having stored thereon executable instructions that when executed by a processor of a computer control the computer to perform steps comprising: providing a connection load balancer as a front end to plural message processors of an STP; publishing, by the connection load balancer, an Internet protocol (IP) address to SS7 peers; initializing the message processors of the STP to listen on the IP address published by the connection load balancer; receiving, at the connection load balancer, a Sigtran message addressed to the IP address; determining, by the connection load balancer, whether the Sigtran message is an initial message or a subsequent message for a Sigtran connection; determining, by the connection load balancer, whether the Sigtran message is for a Sigtran connection to which one of the message processors has been assigned; and forwarding the message to one of the message processors or dropping the message based on whether the message is an initial message or a subsequent message and whether the Sigtran message is for a Sigtran connection to which one of the message processors has been assigned.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The subject matter described herein will now be explained with reference to accompanying drawings of which:

[0025] FIG. 1 is a block diagram illustrating conventional distribution of Sigtran connections to message processors of an STP;

[0026] FIG. 2 is a network diagram illustrating conventional distribution of Sigtran connections among message processors at an STP when a message processor goes down;

[0027] FIG. 3 is a network diagram illustrating distribution of Sigtran connections among message processors at an STP using a connection load balancer;

[0028] FIG. 4 is a network diagram illustrating routing of egress traffic from an STP in a manner that bypasses the connection load balancer;

[0029] FIG. 5 is a message flow diagram illustrating connection load balancer initialization;

[0030] FIG. 6 is a message flow diagram illustrating STP message processor initialization;

[0031] FIG. 7 is a message flow diagram illustrating distribution of Sigtran connections among STP message processors using a connection load balancer;

[0032] FIG. 8 is a message flow diagram illustrating processing of messages associated with existing Sigtran connections at an STP using a connection load balancer;

[0033] FIG. 9 is a network diagram illustrating routing of messages using a connection load balancer after reconnection of a failed message processor;

[0034] FIG. 10 is a block diagram of a connection load balancer and a signal transfer point including multiple message processors; and

[0035] FIG. 11 is a flow chart illustrating an exemplary process for distributing Sigtran messages among STP message processors using a connection load balancer.

DETAILED DESCRIPTION

[0036] Methods, systems, and computer readable media for distributing Sigtran connections among STP processors are disclosed. FIG. 1 illustrates a conventional system for distributing Sigtran connections among STP message processors. In FIG. 1, a signal transfer from STP 100 includes multiple message processors 102.sub.1-102.sub.N. Each message processor 102.sub.1-102.sub.N has dedicated connections, referred to as links, to external devices, such as SS7 peer 104. In the illustrated example, the SS7 links are numbered 1-K. Links 1-K form a logical grouping, referred to as a linkset. The message processors represent a single SS7 point code. Each of the message processors 102.sub.1-102.sub.N has two Ethernet devices 106a and 106b. Multi-homed stream control transmission protocol (SCTP) associations 110 and 112 are established between the message processors of STP 100 and signaling devices 108 of SS7 peer 104. A multi-homed SCTP association is a connection between processes where the process at each endpoint has more than one IP address. In the illustrated example, the IP addresses IP-STP-1-a and IP-STP-N-a are IP addresses associated with one end of path a of a multi-homed STP association. Similarly, IP addresses IP-STP-1-b and IP-STP-N-b are IP addresses associated with one end of path b of a multi-homed SCTP association. The IP addresses IP-STP-1-a, IP-STP-N-a, IP-STP-1b, and IP-STP-N-b are published to SS7 peer 104. As a result, when one of message processors 102.sub.1-102.sub.N fails, or a new message processor is added, reconfiguration of SS7 peer 104 is required.

[0037] Multi-homed SCTP associations 110 and 112 each have endpoints on two different message processors of STP 100. Accordingly, when one of message processors 102.sub.1-102.sub.N fails, traffic can be automatically switched to the other message processor associated with the multi-homed connection or association. However, the remaining path of the SCTP association will operate at reduced capacity. When a message processor becomes available, either after failure or during a scale up operation, new links have to be configured with an IP address for the new message processor. Such configuration can be labor or processor intensive for remote STP peers.

[0038] FIG. 2 illustrates a process that occurs when one of the message processor fails. In FIG. 2, in step 1, message processor 102.sub.1 fails. In step 2, all of the links connected to message processor 102.sub.1 are dropped. Reconnection attempts to message processor 102.sub.1, also fail. For remote peer nodes connected to STP 100, links may be on a single message processor or distributed across multiple message processors based on a redundancy mode. When the failure occurs, SS7 peer 104 attempts to reconfigure the affected links with fixed IPs of another message processor, which can be labor intensive.

[0039] Another issue with the scenario illustrated in FIG. 2 is that automatic connection distribution to the least loaded message processor may not be possible. For example, if message processor 102.sub.1 is loaded disproportionately with Sigtran connections load balancing may not be possible because SS7 peer 104 does not know the relative loading of message processors 102.sub.1-102.sub.N. Another issue with the scenario illustrated in FIG. 2 is that when an STP message processor goes down, links connected to the failed message processor are not automatically distributed to other STP message processors, leaving the linkset operating with reduced capacity unless manual intervention is performed to reconfigure links with other STP message processors. As described above, using multi-homed Sigtran connections, when the message processor associated with one path fails, messages may be transmitted to the STP over the alternate link associated with the multi-homed Sigtran connections. However, as illustrated in FIG. 2, when one message processor fails, half of the signaling links associated with a linkset fails, which reduces the capacity of the linkset. In light of these difficulties, there exists a need for improved distribution of Sigtran connections among message processors at a signal transfer point.

[0040] FIG. 3 is a network diagram illustrating distribution of Sigtran connections at an STP using a connection load balancer. Referring to FIG. 3, a connection load balancer 150 functions as a front end to message processors 102.sub.1-102.sub.N of STP 100. Connection load balancer 150 publishes an IP address on each network to SS7 peer 104. In the illustrated example, the IP addresses that are published are IP-STP-a for network A and IP-STP-b for network B. These IP addresses are public IP addresses that connection load balancer 150 recognizes for messages that need to be distributed among message processors of STP 100. Connection load balancer 150 includes Ethernet interfaces 152A and 152B. Publishing the IP addresses may include broadcasting gratuitous ARP messages to SS7 peer 104 that associate the IP addresses IP-STP-a and IP-STP-b with the MAC addresses of Ethernet interfaces 152A and 152B. Gratuitous ARP messages may be used to publish the IP addresses used by the connection load balancer for load sharing among STP processors if IP version 4 is being used in the network. If IP version 6 is being used, rather than broadcasting gratuitous ARP messages, ICMPv6 neighbor discovery messages may be used to broadcast the IP addresses and corresponding MAC addresses to SS7 peers. Typically, the IP addresses would be published to a border gateway adjacent to connection load balancer 150, and the border gateway would communicate the IP addresses to other SS7 nodes using border gateway protocol (BGP) or other suitable IP route discovery mechanism.

[0041] As with the examples illustrated in FIGS. 1 and 2, multi-homed SCTP associations may be established between SS7 peer 104 and message processors 102.sub.1 and 102.sub.N. SS7 traffic may be transmitted to and from STP 100 over the multi-homed SCTP associations using an adaptation layer protocol, such as M3UA.

[0042] Even though the example illustrated in FIG. 3 illustrates multi-homed SCTP associations between SS7 peer 104 and message processors 102.sub.1-102.sub.N, the subject matter described herein is not limited to such an implementation. Connection load balancer 150 may be used to load balance connections among uni-homed SCTP associations, i.e., those with only a single IP address associated with each endpoint, between SS7 peers and STP message processors.

[0043] As used herein, the term Sigtran connection will be used to refer to a connection established between SS7 endpoints over an IP network to carry Sigtran traffic. An example of a Sigtran connection is an SCTP association.

[0044] Each message processor 102.sub.1-102.sub.N also configures itself to listen to the IP addresses published by connection load balancer 150. For example, message processor 102.sub.1 may listen on IP addresses IP-STP-a and IP-STP-b. Configuring a message processor to listen may include executing sockets a listen command on a given IP address and port.

[0045] Connection load balancer 150 addresses at least some of the problems described above. By exposing only one set public IP addresses used to connect to STP message processors and continuing to use the same set of IP addresses despite message processors being taken into and out of service, changes in STP message processors will not require manual reconfiguration of links and linksets. STP 100 can scale up or scale down without changing the public IP address information. Changes in configuration will be transparent to SS7 peers 104 because new message processors will use IP addresses from the existing set of IP addresses published by connection load balancer 150. In FIG. 3, connection load balancer 150 only exposes IP addresses IP-STP-a and IP-STP-b to SS7 peer 104. STP message processors 102.sub.1-102.sub.N each recognize the set of IP addresses published by connection load balancer 150. New physical or virtual message processors will be configured to recognize the same set of IP addresses. As a result, when message processors are added or removed from service, such a change in configuration is transparent to SS7 peer 104.

[0046] Connection load balancer 150 may also balance Sigtran connections among SS7 message processors 102.sub.1-102.sub.N. If message processing capacity at STP 100 is lost or decreases, connection load balancer 150 may redistribute links among available SS7 message processors and thereby avoids service disruption with remote SS7 peer 104. Similarly, if processing capacity is added at STP 100, connection load balancer 150 may rebalance Sigtran connections among available message processors in a manner that is transparent to remote SS7 peers.

[0047] FIG. 3 can also be used to illustrate the routing of ingress traffic using connection load balancer 150. In FIG. 3, traffic from signaling devices 108 on SS7 peer 104 is addressed to the public IP addresses IP-STP-a or IP-STP-b published by connection load balancer 150. Connection load balancer 150 receives the message traffic and determines whether, for each message, the message is an initial message associated with a Sigtran connection or a subsequent message associated with a Sigtran connection. An initial message is a connection request message for which a message processor has not been assigned. For an SCTP association, an example of an initial message is an SCTP INIT message. In the SCTP protocol, an INIT message is the message used to initiate establishment of an SCTP association with a remote endpoint. The INIT message carries a list of IP addresses associated with the endpoint sending the INIT message.

[0048] A subsequent message is a message, such as an SCTP cookie-echo message, selective acknowledgement (SACK), heartbeat, heartbeat ACK or a data message, that is sent after an initial message for an SCTP association. If the message is determined to be and initial message, connection load balancer 150 may assign the message to one of message processors 102.sub.1-102.sub.N using a load balancing algorithm, which will be described in detail below. If the message is determined to be a subsequent message associated with a Sigtran connection, connection load balancer 150 will route the message to the message processor 102.sub.1-102.sub.N previously assigned to the Sigtran connection. If the message is a subsequent message and no message processor has been assigned to the Sigtran connection, connection load balancer 150 may drop the message.

[0049] FIG. 4 is a diagram illustrating the flow of egress message traffic from STP 100 to remote SS7 peer 104. Referring to FIG. 4, egress traffic from message processors 102.sub.1-102.sub.N is not routed through connection load balancer 150 and thus does not affect the capacity of connection load balancer 150. The mechanism by which traffic is routed around connection load balancer 150 includes IP routing protocols where routing paths from STP 100 to signaling devices 108 are determined using standard routing information protocols, such as routing information protocol (RIP) a open shortest path first (OSPF). Such paths may not include connection load balancer 150. Since gratuitous ARP or ICMPv6 neighbor discovery messages are broadcast for the public IP address used by the connection load balancer, ingress messages (i.e., messages from an SS7 peer to one of the STP message processors) addressed to the public IP address are routed through the connection load balancer. Egress messages are not addressed to the public IP address of the connection load balancer. Rather, they are addressed to the IP address of the destination SS7 peer. As stated above, the route to the SS7 peer may not include connection load balancer 150, and thus egress messages (i.e., messages from an STP message processor to an SS7 peer at the other end of an SCTP association) may be routed over routes determined using RIP or OSPF, which bypass connection load balancer 150.

[0050] FIGS. 5 and 6 illustrate connection load balancer initialization. Referring to FIG. 5, in step A1 connection load balancer 150 publishes IP addresses for each network on which connection load balancer is present to an SS7 peer, which in the illustrated example is border gateway 156. The publishing of the IP addresses may be accomplished by broadcasting gratuitous ARP or ICMPv6 neighbor discovery messages associating IP address IP-STP-a with MAC address MAC-a and IP address IP-STP-b with MAC address MAC-b. In step A2, border gateway 156 updates its MAC or ARP table to associate the IP addresses with the MAC addresses. Table 160 illustrates and example of the configuration of the ARP table of border gateway 156 after receiving the gratuitous ARP message from connection load balancer 150.

[0051] FIG. 6 is a flow diagram illustrating initialization of the message processors. Referring to FIG. 6, in step B1, all of the message processors configure an alias address of the IP address IP-STP-a and IP-STP-b published by connection load balancer 150 on the loopback interface. The IP addresses are configured on the loopback interface of each message processor 102.sub.1-102.sub.N so that message processors 102.sub.1-102.sub.N do not send ARP or ICMP messages for these IP addresses. In step B2, each message processor 102.sub.1-102.sub.N configures multi-homed responder connections which start listening on IP address IP-STP-a for the primary path and IP address IP-STP-b for the secondary path.

[0052] After initialization, Sigtran connections can be distributed among processors at STP 100 using a load balancing algorithm. FIG. 7 is a message flow diagram illustrating the distribution of new Sigtran connections to SS7 message processors using connection load balancer 150. FIG. 7 Referring to FIG. 7, in step C1, a remote SS7 client 104 sends a Sigtran connection request on link L1. The destination IP addresses for the connection request include IP addresses IP-STP-a and IP-STP-b. The source IP address in this example is R-IP-x and R-IP-y. The destination port for the connection request is P1.

[0053] In step C2, border gateway 156 queries its MAC table for the MAC address corresponding to IP address IP-STP-a. In this example, border gateway 156 locates the address MAC-a, which corresponds to an interface of connection load balancer 150. Accordingly, border gateway 156 forwards the message to connection load balancer 150.

[0054] In step C3, connection load balancer 150 applies a connection distribution or load balancing algorithm to determine which STP message processor should receive the connection request. In this example, it is assumed that message is sent, as a result of the load balancing algorithm, to message processor 102.sub.i. In one example, connection load balancer 150 may perform load balancing based on the following parameters in the algorithm. For each message processor, the connection load balancer 150 may maintain a group count. A group count represents the number of links from a remote peer that land on a given message processor. Initially, the group count will be set to 0 for all message processors. Each time a new Sigtran connection is assigned to a message processor, the group count is incremented. Another parameter that may be used in the load balancing algorithm is the link distribution threshold, which is valid across a linkset. The link distribution threshold is configurable value set by the user to define a threshold difference between group counts above which a message processor will not be considered for load balancing. For example, the link distribution threshold may be set to 4. When connection load balancer 150 receives a new Sigtran connection request, connection load balancer 150 determines the group count of each message processor 102.sub.1, 102.sub.i, and 102.sub.N. Connection load balancer 150 then identifies the message processor with the lowest group count and adds the link distribution threshold to the lowest group count. Any message processor having a group count that is greater than the sum of the lowest group count and link distribution threshold will not be considered for load balancing. Any message processor having a group count that is less than the sum of the lowest group count and the link distribution threshold will be considered as a connection distribution candidate. For example, if the group counts of message processors 102.sub.1, 102.sub.i, and 102.sub.N are 1, 3, and 4, respectively, and the link distribution threshold value is 2, then only the message processors 102.sub.1 and 102.sub.i will be considered as candidates for connection distribution.

[0055] Once the group of candidates for connection distribution is identified, connection load balancer 150 may select from one of the candidates, e.g., using a random selection algorithm, and forward the connection request to the assigned message processor (see step C4). Connection load balancer 150 may also create an entry in its association database that associates the Sigtran connection with the assigned message processor. An example of parameters that may be stored in the association database is as follows: [0056] {R-IP-x, R-IP-y, R-Port-->STP-MP-i}
In this example, R-IP-x and R-IP-y are IP addresses of the SS7 peer that sends an SCTP INIT message to connection load balancer 150. More than one IP address may be used for a single peer because the peer implements SCTP multi-homing. R-Port is the SCTP port number used by the sending node in the SCTP INIT message. STP-MP-i is an identifier used by connection load balancer 150 for the STP message processor assigned to the connection.

[0057] After a record such as that illustrated above is created by connection load balancer 150, subsequent messages, such as SCTP data or control messages sent over the same SCTP association, will be forwarded to the same STP message processor. For example, when an SCTP data or control message (also referred to as chunks) arrives at connection load balancer 150, connection load balancer 150 may perform a lookup in its association database using the source IP addresses and SCTP port number in the message. In this example, it is assumed that the source IP addresses are R-IP-x and R-IP-y and the source port is R-Port. Accordingly, connection load balancer 150 will locate the matching entry from the example above and forward the SCTP data message to the assigned STP message processor STP-MP-i.

[0058] It should also be noted that if an SCTP INIT message is received a connection load balancer 150 and a connection has already been established with one of the STP message processors, connection load balancer 150 will forward the message to the assigned message STP message processor. The SCTP association state machine on the assigned message processor will handle out of state messages, such as INIT messages for which a connection has already been established. Continuing with the above example, if an SCTP INIT message is received with source IP addresses R-IP-x and R-IP-y and a source SCTP port of R-Port, and connection load balancer locates a matching entry in its association database, connection load balancer 150 will forward the INIT message to the assigned message processor. The assigned message processor may receive the INIT message, and depending on the current state of the SCTP association state machine, process or drop the INIT message.

[0059] Returning to FIG. 7, in step C5, the assigned message processor accepts the connection and sends the response directly to the remote peer via border gateway 156. Connection load balancer 150 is not in the path for the egress messages.

[0060] FIG. 8 is a message flow diagram illustrating the processing of subsequent messages by connection load balancer 150. As stated above, the term subsequent messages refers to messages other than connection establishment messages. Referring to FIG. 8, in step D1, a Sigtran message arrives at border gateway 156 from remote client 104. The Sigtran message arrives on link L1 with destination IP addresses IP-STP-a and IP-STP-b, source IP addresses R-IP-x and R-IP-y, and destination port P1.

[0061] In step D2, border gateway 156 queries its MAC table using the IP addresses IP-STP-a and IP-STP-b from message and identifies the MAC addresses from connection load balancer 150. Border gateway 156, in step D2, forwards the Sigtran message to connection load balancer 150.

[0062] In step D3, connection load balancer 150 queries its association database to determine whether the received message is associated with Sigtran connection that has already been assigned to a message processor of STP 100. As stated above, the association database is to a database maintained by connection load balancer 150 to keep track of assignments of Sigtran connections (SCTP associations) to STP message processors. Parameters used in the association database to associate a Sigtran connection with a message processor may include source IP addresses and source ports. In this example, it is determined that messages with remote IP address R-IP-x on port P1 are associated with STP message processor STP-MP-i. In step D4, connection load balancer 150 forwards the message to message processor STP-MP-i. In step D5, message processor STP-MP-i accepts the data (or other subsequent) message and sends responses directly to the remote peer over the same SCTP association via the border gateway without passing through connection load balancer 150. Subsequent messages received by connection load balancer 150 for which no association entry is found in the association database are dropped.

[0063] FIG. 9 is a network diagram illustrating operations performed by a connection load balancer after reconnection of an SS7 message processor. Referring to FIG. 9, in step 1, a message processor of STP 100 goes down. In step 2, SS7 links connected to the failed message processor also go down. In step 3, remote SS7 peer 104 attempts a reconnection of link 1. The reconnection is performed using the same IP addresses used by the failed message processor. Connection load balancer 150 reroutes the connection request to one of the available message processors based on the connection distribution algorithm described above.

[0064] FIG. 10 is a block diagram of connection load balancer 150 and STP 100. Referring to FIG. 10, connection load balancer 150 includes Ethernet interfaces 152A and 152B for networks A and B, respectively. Connection load balancer also includes at least one processor 154. A Sigtran connection distributor 155 may execute on processor 154 to perform the steps described herein for distributing Sigtran connections among message processors 102.sub.1-102.sub.N of STP 100.

[0065] STP 100 includes message processors 102.sub.1-102.sub.N. Each message processor includes Ethernet interfaces 106A and 106B respectively connecting to networks a and b. Each message processor 102.sub.1-102.sub.N may also include a processor 158 and an SS7/Sigtran function 159. SS7/Sigtran function 159 may perform SS7 functions, such as routing messages based on point codes and Sigtran functions, such as setting up multi-homed SCTP associations with remote peers. Each message processor 102.sub.1-102.sub.N may be a physical message processor or a virtual message processor. As physical message processors, each message processor 102.sub.1-102.sub.N may be a printed circuit board that is connected to other printed circuit boards in STP 100 through a communications medium 162, such as a bus or backplane. As virtual entities, each message processor 102.sub.1-102.sub.N may represent a virtual machine that runs on a hypervisor on cloud computing hardware. Thus, the subject matter described herein includes distributing connections among real or virtual message processors in a real or virtual STP.

[0066] FIG. 11 is a flow chart illustrating an exemplary process for distributing Sigtran connections among STP message processors. Referring to FIG. 11, in step 200, the connection load balancer publishes its IP address that it uses for connection load balancing to remote peers. This step may be performed by broadcasting gratuitous ARP messages to associate the IP addresses for load balancing with MAC addresses of the connection load balancer. If IPv6 is used, step 200 may be performed by broadcasting ICMPv6 neighbor discovery messages to adjacent nodes.

[0067] In step 202, message processors of the STP are configured with the IP addresses as an alias on the loopback interface. The message processors are also configured to listen on the IP addresses. The loopback interface is used so that the message processors will not broadcast ICMPv6 neighbor discovery or ARP messages concerning the IP addresses.

[0068] In step 204, a Sigtran message is received at the connection load balancer. The Sigtran message may be an SCTP message, such as an SCTP control or data chunk. The SCTP message may carry an SS7 over M3UA payload or other SS7 over Sigtran payload.

[0069] In step 206, it is determined whether the message is an initial message or a subsequent message. Determining whether the message is an initial message or a subsequent message may include examining the SCTP chunk type field of the message. The chunk type field carries a code that identifies the type of SCTP message. The value 0x01 indicates that the chunk type is INIT or initiation. The chunk type 0x00 indicates that the chunk type is data, or that the message carries data for the applications using the SCTP connection. Other chunk types of interest may include 0x0a, identifying a cookie echo message, which is part of the handshake procedure for establishing an SCTP association.

[0070] If the message is determined to an initial message in step 206, control proceeds to step 206A, where it is determined whether a message processor has been assigned to the connection. Determining whether a message processor has been assigned to the connection may include querying the association database maintained by connection load balancer 150 using the source IP addresses and source SCTP port in the message. If a matching record is located in the association database, then it is determined that a message processor has been assigned to the connection. If a record is not located in the association database, then it is determined that a message processor has not been assigned to the connection.

[0071] If the message is an initial message and a message processor has been assigned to the connection, control proceeds to step 208 where the message is forwarded to the message processor assigned to the connection. The SCTP association state machine on the assigned message processor will handle out of state messages, such as an INIT message for which a connection has already been established. If the message is an initial message and a message processor has not been assigned to the connection, control proceeds to step 210 where the Sigtran connection for which the initial message is intended to initiate establishment is assigned to one of the message processors and results in SCTP/Sigtran connection establishment between the assigned message processor and the remote SS7 peer. SCTP connection establishment involves the sending of an INIT acknowledgment message from the assigned message processor to the SS7 peer. The SS7 peer then sends a cookie-echo message to the assigned message processor via the connection load balancer. The connection load balancer treats the cookie-echo, data, and other non-INIT messages received from the SS7 peer over the SCTP association as subsequent messages for the SCTP association.

[0072] Assigning the Sigtran connection to one of the message processors may include applying the load balancing algorithm described above to identify connection distribution candidates based on group counts which indicate relative loading of the message processors with Sigtran connections, and then selecting one of the connection distribution candidates using a selection algorithm. Because the load balancing algorithm identifies connection distribution candidates with similar Sigtran connection loadings, the selection algorithm may be any suitable selection algorithm, such as random selection, round robin selection, etc. Once a message processor is selected, the final step associated with assigning the Sigtran connection to the message processor includes creating a record in the association database maintained by connection load balancer 150 that associates the source IP addresses and source SCTP port number from the message with the assigned STP message processor. After the connection is assigned to one of the STP message processors, control proceeds to step 210 where the message is forwarded to the assigned message processor.

[0073] Returning to step 206, if the message is determined to be a subsequent message, control proceeds to step 206B where it is determined whether a message processor has been assigned to the Sigtran connection with which the subsequent message is associated. As stated above, examples of subsequent messages include SCTP data and control messages (other than INIT messages). Determining whether a message processor has been assigned to the subsequent message may includes performing a lookup in the association database maintained by connection load balancer 150 using the source IP addresses and source SCTP port number in the message. If a match is located in the database, the connection may be determined to be assigned to one of the message processors. If a match is not located in the association database, the connection may be determined not to be assigned to one of the message processors.

[0074] If the message is a subsequent message for a Sigtran connection for which no STP message processor has been assigned, control proceeds to step 207B where the message is discarded. If the message is a subsequent message for a Sigtran connection for which a message processor has been assigned, control proceeds to step 210 where the message is forwarded to the message processor previously assigned to the Sigtran connection.

[0075] Thus, using the steps described herein, real and virtual STP message processors can be seamlessly added to or removed from a network without requiring remote nodes to be reconfigured with new IP addresses. The connection load balancer also balances loading of message processors in the STP. The subject matter described herein can be used to distribute messages among real or virtual STP message processors.

[0076] It will be understood that various details of the presently disclosed subject matter may be changed without departing from the scope of the presently disclosed subject matter. Furthermore, the foregoing description is for the purpose of illustration only, and not for the purpose of limitation.