28-01-2013, 04:09 PM
Resilience technologies in Ethernet
Resilience technologies.pdf (Size: 1.56 MB / Downloads: 67)
ABSTRACT
In choosing a network service technology, a subscriber considers many features such as
latency, jitter, packet loss, security, and availability. The most important feature, and usually
the one that determines the final selection, is the service availability. In this article, a
full spectrum of applications are studied, ranging from the minimal constraints of home
networks to the rigorous demands of Industrial Ethernet Networks. This is followed by a
thorough examination of Ethernet layer resilience technologies.
Introduction
Quintessentially, Ethernet is a simple networking technology
to connect two endpoints at the data link layer.
Using Ethernet, a local area network (LAN) can be built
and configured in a short amount of time. Its success is
in part due to standardization that enables the interoperability
among equipment vendors. Techniques for plug-nplay
and auto-negotiation means that an Ethernet LAN
does not require additional equipment, such as a rate converter
because a 10 Mbps interface can communicate directly
with a 100 Mbps interface. In addition, Ethernet
has become the aggregation protocol, allowing other network
protocols to run it, such as MPLS over Ethernet and
SONET over Ethernet.
Topology
A network topology comprises the following fundamental
topologies: linear, tree, ring, star, or mesh. The linear
topology and tree topology are configured without any
redundancy; whereas ring and mesh topologies have
redundant links built-in to protect the network. Redundancies
within a network include network elements such as
switches and links that exceed the minimum number for
the network to operate. The redundancies create more
than one path between the source and destination to reroute
the traffic at the time of failure. Fig. 3 shows a typical
linear topology where the switches are connected in a line.
Each switch has at most two links connecting the immediate
adjacent switches. The latency for any communication
is proportional to the distance of the ingress and egress
switch. Therefore, due to the latency requirement of some
applications, when a linear topology reaches a certain size
it must be branched out, as shown in Fig. 4.
Fig. 5 shows an example of a tree topology with a root
switch. Within a tree topology, there is only one path between
any nodes, leaf nodes or switches. Essentially, traffic
tends to be forwarded toward the root on its path to the
destination. In effect, this topology has the bottleneck at
links around the root. A subset of the tree topology is a star
topology, as shown in Fig. 6.
Protection mechanism
To achieve a high level of service availability, a network
architecture can provide a system of physical redundancy
in parallel with software for efficient management. The
physical redundancy is needed to eliminate the single
point of failure syndrome on the routing path. There are reserved
resources in a system, such as redundant links and
redundant nodes, which after a failure occurs these standby
resources are used to reroute the traffic. There are different
levels of protection ranging from 1+1, offering
100% protection, to m:n where the protection resources
are shared offering only partial protection of the traffic.
The protection type 1+1 is the most expensive mechanism,
but it guarantees 100% protection. At the ingress
node, the traffic is replicated and is sent to the destination
via two disjoint paths. The egress node is responsible for
forwarding one frame and discarding the duplicate. The
decision is performed on a per frame basis and is triggered
by an event such as missing frames from the primary flow.
As there are always two flows carrying identical traffic, the
bandwidth utilization is very inefficient.
Metropolitan area networks
In Metro Area Networks (MEN), the following protection
approaches that can coexist are under consideration
by the Metro Ethernet Forum (MEF) [8]:
1. Aggregated Line and Node Protection (ALNP).
2. End-to-End Path Protection (EEPP).
3. MP2MP Protection.
Aggregated Line and Node Protection (ALNP) provides
protection for local link and local nodes via a detour mechanism.
The detour path temporarily traverses the point of
failure and merges back onto the primary path. ALNP supports
1:1 and 1:n protection on the detour routes. Since the
protection is local, it has a quicker recovery rate than the
End-to-End Path Protection (EEPP). ALNP can also be invoked
before EEPP.
Unlike ALNP, EEPP provides disjoint backup paths from
the source to the destination than the primary path supporting
1+1, 1:1, and 1:n protection. The number of the
backup paths depends on the policy requirement of the
network in question.
DRP
Proposed by Feng et al. as an addendum to IEC 62439
under clause 8, Distributed Redundancy Protocol [30] is a
high availability network solution for a ring topology to
detect a single failure and to recover in a deterministic
time period. DRP synchronizes all nodes in the ring so
that the scheduling can be divided into intermittent periods,
called a macrocycle. Within each macrocycle, only
one node is allowed to send a Ring Check frame that is
used to detect a fault on the ring. Each node in the DRP
ring will take a turn to send out the Ring Check frame
on two of its active ring ports as shown in Fig. 26. In addition,
each node sends Link Check frames to its immediate
neighbor to detect any fault on the adjacent fault per
macrocycle. If enough diagnostic frames are missing, the
node changes the faulty link to BLOCKING mode.