PM8621-BI PMC-Sierra, Inc., PM8621-BI Datasheet - Page 195

• Manufacturer Part Number: PM8621-BI
• Description: Driver, NSE/SBS Narrowband Chipset Driver
• Manufacturer: PMC-Sierra, Inc.

Proprietary and Confidential to PMC-Sierra, Inc., and for its Customers’ Internal Use
Document ID: PMC-2021248, Issue 1
A
S
Overview
In a regular system, there should only be one CSD configured to run the OPA which keeps track
of all the connections in the fabric. If a more fault tolerant system is to be designed, the hardware
that runs the CSD/OPA may become a single point of failure. One possible approach to achieve a
fault tolerant system is to maintain two independent copies of the CSD/OPA running in a working
and protect hardware mechanism (Commands may be broadcast to both for concurrent
processing). In the event of a card failure, the system can be switched over to the protect
hardware (if the working hardware fails). The failed hardware can then be replaced without any
service interruption. Once boot up and initialized, the new hardware can then be synchronized
with the currently active hardware. The fault tolerant system is then fully restored.
The CSD/OPA keeps track of all the system-wide connections by maintaining internal states. This
state information is updated whenever there are call connection/disconnection or switchover
requests. A fault tolerant system will not be completely restored unless this state information can
be fully duplicated in the new hardware. Such process is being defined as the CSD
synchronization. The CSD provides API function to retrieve and restore the internal state of the
software. The saved state of the software is sometimes referred to as a checkpoint.
The CSD/OPA includes an example of a log-based with checkpointing recovery scheme. Prior to
restoring the checkpoint of the CSD/OPA, the CSD should first be initialized. The system relies
on an external repository to keep a log of all device and fabric initialization commands to the
CSD (therefore it is called a log-based with checkpointing recovery scheme). During
synchronization, the initialization command sequence is first “played back” to and then the
checkpoint is restored in the new hardware. These information will then be the exact same state
as the protect card, thus achieving synchronization.
The following outlines a typical event sequence before and after a failure recovery (assuming a
failure in working fabric). The protection switchover procedure is described in (c) and the
recovery procedure starts from step (d).
(a) Both working and protect card are brought up with the same device and fabric initialization
command sequence initially. Subsequent call commands (such as setting up or tearing down
tributaries, and protection switchovers) are always broadcast to both cards simultaneously.
Assuming a reliable communication channel, the states of both cards are in synchronization.
(b) The working CSD is normally in control and incremental change in SBS settings are sent to
remote line cards. Any SBS or NSE settings local to the switch card will be updated by the CSD.
(c) When the working fabric fails, a protection switchover occurs and all SBSs send and receive
traffic via the protect LVDS links. The protect fabric becomes the master (and the lone card) in
the system.
YNCHRONIZATION
PPENDIX
F: N
ARROWBAND
Appendix F: Narrowband Chipset Driver Synchronization
C
HIPSET
Narrowband Chipset Device Driver Manual
D
RIVER
195