Specifications

Page 122 /148
4.1.7 Network Level Reliability
While these design practices help to make an individual Mxxx more reliable, they
do not guarantee it will never fail. What is likely, however, is that with the common
hardware failures taken care of the most likely remaining failures will either be
due to operator error or to software bugs. The best way to provide network-level
reliability is to use loosely coupled redundant Mxxx units in the network and route
around failures when they do occur. The definition of loosely coupled specifically
implies that paired Mxxx routers are not exposed to the same traffic and therefore
do not have the same internal software state.
4.2 Hardware Redundancy
4.2.1 Redundancy of Central Control and Processing
The design goal of all the Juniper products was to make the architecture simple,
modular and therefore reliable. The physical separation of Routing and Packet
Forwarding Engines generally results in a more reliable system because there
are fewer common-mode faults.
The Mxxx Internet backbone router is unique in its ability to provide the rock-solid
stable performance that other systems lack during periods of extreme stress:
The Mxxx Internet backbone router is fully sized with respect to both route
processing and packet forwarding. During exceptional conditions, the Routing
Engine continues to receive and transmit routing updates, perform route
calculations, maintain peer relationships, react to interfaces going down, and so
forth. Similarly, the Packet Forwarding Engine continues to switch packets at a
rate of 40 Mpps regardless of packet size or load on the system.
Complementing the architectural separation of the routing and packet-forwarding
processes, atomic updates permit the state of the Packet Forwarding Engine to
concur with the state of the Routing Engine without impacting forwarding
performance. During exceptional conditions, atomic updates allow the Mxxx
system to avoid destabilizing the links that still remain up, thus eliminating the
primary reason for cascading failures.
The M20 and M160 have support for redundant routing engine (RE) and packet
forwarding engine.
4.2.2 Redundant Routine Engine
The M20 and M160 support a redundant routing engine with the goal of
minimizing mean time to repair (MTTR) in the case of an RE failure. The
redundancy support that has been implemented is a warm-backup “spare-in-the-
box, manual switch-over” model in order to provide protection against hardware
failure. The model allows M20/160 customers to fall back to a backup RE in the
case of failure, without having to be physically present to swap out components.
Based on customer input, additional functionality will be added to the model in
future JUNOS releases.
RE Election Priorities: Master, Backup, and Disabled
An RE can be configured to be in one of three election priorities: Master, Backup
or Disabled. The running state of an RE is determined as a result of a mastership
election upon system boot. Changing the running state of an RE is also
accomplished manually by using a switchover command.
§ Master—If an RE is configured to be the master then it has full functionality
as a routing engine. Specifically, it receives and transmits routing
information, builds and maintains routing tables, communicates with
interfaces and packet forwarding engine components, and has full control
over the chassis. Once an RE becomes master, it resets the switch plane
and downloads its current version of the microkernel to the PFE
components. This guarantees software compatibility.
§ Backup—If an RE is configured to be the backup then it does not maintain
routing tables, nor does it communicate with PFE or chassis components.
However it has run through its memory check and boot sequence to the
point of displaying a login prompt. A backup RE supports full management
access through the Ethernet, console, and auxiliary ports and can
communicate with the master RE. Additionally, a backup RE responds to
the RE switchover command. The backup RE maintains a connection with
the master RE and monitors the heartbeat of the master RE. If the