UK VoIP Interconnect - White Paper
In SCTP, there is a multi-homing option which allows an SCTP source to define two locations from where it can transmit and receive data. These can be used to provide very rapid circuit protection at the IP transport layer. In the diagram above, both ends of the connection, (the Call Agents) have chosen to present primary and secondary addresses.
The SCTP layer in each Call Agent decides to send packets from and to either interface as it pleases. While normally there is the concept of primary and secondary, some SCTP implementations send alternate packets from primary and secondary as shown above.
This introduces some difficulties for Session Border Controllers. The SBC terminates the transport layer, re-assembles the SIP messages, creates media pinholes and then re-forms the SIP messages with changed SDP and source details and sends them on.
To do this the SBC has to be able to see all multi-homed streams, and of course in the distributed SBC scenario this is not possible as shown below. The two multi-homed SCTP streams don't come together in the border devices.
So a workable architecture, if SCTP is used, is to have the SCTP terminated and then resourced as a multi-homed stream across the interconnect as illustrated below.
In this case the media path can be successfully controlled because the Signalling Border devices can reassemble at the Transport Layer and can thus understand the SIP signalling.
Separate Signalling and Media
In the general model, media and signalling boundary functions can be provided in either separate or combined devices. However, while a loss of signalling can be easily detected (if a little slowly when using TCP as the transport protocol), the loss of media due to some fault in the interconnect is less easy to determine. So when the link fails between two peer Service Providers, the callers will know before the Switching System that the media connection has gone and will clear down. However, when the caller tries to re-establish the call, because the switching system has no knowledge of the failure it will still try to use the same resources to connect the media, and the call will probably fail.
What is required is for the Call Agent to know, as rapidly as possible, that the connectivity has failed using techniques such as O&AM capabilities in layer 2 interconnect. Prior to the widespread availability of suitable standards, such as Bidirectional Forwarding Detection (BFD-which is still at the IETF draft stage - draft-ietf-bfd-base-03.txt July 2005), the best approach is, in the short term, to pass media and signalling down the same link so that when the link fails the signalling fails also and the Call Agent can take appropriate actions to avoid the failure.
As the standards evolve to support comprehensive link failure detection, the 1460 can be re-deployed to with split signalling and media architectures, such as those being defined by the ETSI TISPAN initiative.
Overload Control and Interconnect Congestion Management
The interconnect functions will be supported by a limited capacity interconnect capability, defined by the bandwidth of the interconnecting links. In order to prevent the overloading of these links and the consequential decline in call quality, a bidirectional session admission control system, such as provided by the Newport Networks 1460 should be used.
This will regulate the use of the interconnect, but it cannot effectively deal with overloads related to focus events. These events, often initiated by television programmes that involve subscriber voting, consist of very high volume, short calls. These events are very financially rewarding and calls should be handled efficiently whenever possible. At the same time other calls such as emergency calls should be preferentially allowed to pass.
This "Session Admission Control Plus" capability, which can regulate call volumes and bandwidth usage to target ranges of telephone numbers rather than simple destination IP addresses, not only caps interconnect usage but also allows the early back-off of damaging call volumes, through the selective and increasing use of SIP 503 messages (this message allows the SBC to provide a server unavailable message but also to ask the sender to retry after time x) to achieve traditional call gapping functionality.
Conclusion
The 1460 product family includes both Signalling and Media Border Functions that can be deployed in a number of ways to meet the requirements of NICC and UK service provider PSTN emulation VoIP interconnect. The H.248 capabilities of the product allow a range of evolutionary deployment scenarios as standards become available and deployed. The 1460 products enables carrier resilience to be deployed with a number of multi-node mechanisms to provide a PSTN equivalent service using VoIP, which is as good as or better than current PSTN practice, and at a much lower cost. 
|
