IP endpoint recovery
Issue 6 January 2008 277
While the LSP is not call preserving, the fail-over from primary gatekeeper to LSP is an
automatic process, and does not require human intervention. The fail-back from LSP to primary
gatekeeper, however, is not currently automatic, and requires a system reset on the LSP. During
the fail-back to the primary gatekeeper, all calls are dropped, with the exception of IP-to-IP calls.
IP Endpoint Time to Service
The Time to Service (TTS) feature improves the time required to bring an IP phone system into
service by reducing the amount of required signaling for a phone to reach the in-service state.
Once a phone is registered, TTS keeps the registration persistent for a relatively long Time to
Live (hours) regardless of TCP connection failure, network outages, or even restarts of the
endpoint. This significantly reduces the number of times that IP phones need to re-register with
Communication Manager due to outages. As a result, the TTS feature improves system
availability after a network outage.
There are two functions in TTS that improve IP endpoints availability. One function is that the IP
Endpoint Time-To-Service feature changes the way IP endpoints register with their gatekeeper,
reducing the time to come into service. In the current Avaya Communication Manager
architecture, there are two activities to bring the IP endpoints into service. The H.323 IP
endpoint must register with the communication manager and then it must establish a TCP
socket connection between the server and the endpoint for call signaling. Since all the IP
endpoints in a system strive to get into service as quickly as possible after an outage, the main
processor can be flooded with activity. In a system with a large number of IP endpoints, this
flooding leads to delays not only for phones trying to get into service, but also for endpoints
already in service trying to make calls.
With Avaya Communication Manager Release 4.0, the TTS separates the timing of the H.323
registration process from the timing of the TCP socket-connection setup process. This
decoupling of the steps considerably improves the time for phones to be in-service.
With TTS, after all the IP phones within a system register to the Communication Manager, the
TCP socket is established when the processor occupancy level returns to normal. However,
when the main processor occupancy level is high, the TCP socket is established on demand
(when users make a first call or when a call needs to be delivered to a user) or via background
maintenance. Once the TCP socket is established, the socket remains up for subsequent calls.
In addition, with TTS, Communication Manager, rather than the IP endpoint, initiates the
establishment of the TCP socket resulting in faster establishment of TCP sockets.
The second function of TTS significantly reduce the number of times that IP endpoints need to
re-register with the Communication Manager. This feature provides the capability to persist IP
endpoint registrations across many network failures and other outages. Currently, whenever
TCP sockets are dropped, the IP endpoints must re-register. With TTS, IP endpoints will not
usually need to re-register for network outages that do not cause the system to failover to an
Enterprise Survivable Server (ESS) or Local Survivable Processor (LSP). Since most issues
with registration delays in the past have been after short network outages, this capability
dramatically reduces the number of times that an IP endpoint needs to re-register with
Communication Manager.
