Box C Resilience
As the multitude of services on IP-based networks increases, a differentiation of
dependability (e.g. reliability and availability) requirements is asked for. This is
sometimes referred to as resilience requirements. A resilience differentiated net-
work would only protect traffic flows that require higher availability that would
allow for more effective network resource utilisation. Basically, this could be
done on other layers than IP. In principle realising resilience mechanisms in
lower layers would allow for faster response, e.g. from a link is broken until an
alternative link is found. On the other hand, lower layers also operate on more
coarse granularity of traffic aggregates. Thus, higher layers give finer granularity
but commonly also longer response times.
Traffic flows requiring high availability may belong to so-called mission-critical
applications. Such applications may include all types of applications, like real-
time, elastic, etc. Therefore, applications such as multimedia as well as data-
base transactions could be asking for high availability. This means that the
resilience requirement is orthogonal to other performance variables, like delay
and loss.
A possible set of resilience classes is described in [ID_resreq]:
- High resilience requirements: Resources should be exclusively reserved on
an alternative path. For a 1+1 protection, packets are forwarded on the work-
ing path and the backup path. In case the working path fails, the receiver just
selects packet on the other path. In case of 1:1 protection, the packets are
forwarded on the alternative path in case of failure on the working path. This
asks for recovery signalling to handle unidirectional failures. - Medium resilience requirements: Spare resources may be shared between
multiple flows. The bandwidth management has to assure that enough spare
resources are available for a given set of expected failures. In case of a fail-
ure, packets are forwarded after rerouting and reservation of spare resources. - Low resilience requirements: No resources are reserved in advance. In case
of failure, packets may be forwarded after a rerouting and reservation phase
if enough resources are available. - No resilience requirements: In case of a network failure in the administrated
domain, packets may be discarded/dropped. This may happen even if the
traffic is not directly affected by the failure but rather by a rerouting of other
traffic flows having high resilience requirements.
In order to implement differentiated resilience, updates of the service implemen-
tation could be needed. For IntServ/RSVP corresponding attributes have to be
carried in the signalling messages and filters/conditioners have to be available.
For DiffServ activating a resilience marking may be used, like a bit in the ToS
octet (bit 5 of the DSCP field). Resilience attributes may also be used for MPLS.
These mechanisms are described for the different mechanisms in this article.
Box D Information Distribution
In current IP-based networks several means are used to make the distribution
of information more efficient, including:
- mirroringthe information meaning that the information is replicated in several
places/servers. This would increase the dependability and allow for faster
response; - cachingthe information, basically meaning that previously accessed informa-
tion is stored in a place closer to the user (limiting the traffic and allowing
faster response); - load balancinghaving the objective to distribute the traffic/requests among
the servers.
more interest. This addresses the selection
of paths for packets and may work well with
path-oriented solutions, that is LSPs.- Traffic mapping. This refers to the assignment
of traffic flows onto paths to meet certain
requirements considering the set of policies
relevant. A central issue arises when several
paths are present between the same pair of
originating and destination router. Appropriate
means should be taken to distribute the traffic
according to some defined ratios, still keeping
the ordering of packets belonging to the same
application (or micro-flow). - Measurement. Mechanisms for monitoring,
collecting and analysing statistical data have
to be in place. Such data may relate to perfor-
mance and traffic. In particular, being able to
construct traffic matrices per service class is
an essential part of a TE system. - Network survivability. Survivability refers to
the capability to maintain service continuity
in presence of faults. This can be realised by
rapid recovering or by redundancy. Co-ordi-
nating protection and restoration capabilities
across multiple layers is a challenging task. At
the different layers protection and restoration
would typically occur at different temporal
and bandwidth granularity. At the bottom
layer, an optical network layer may be pre-
sent, e.g. utilising WDM. Then, SDH and/or
ATM could be present below the IP layer.
Restoration at the IP layer is commonly done
by the routing protocols, which may require
some minutes to complete. Some means being
proposed relate to MPLS allowing for faster
recovery (ref. Box B). A common suit of con-
trol plane protocols has been proposed for the
MPLS and optical transport networks. This
may support more sophisticated restoration
capabilities. When multiple service classes are
present, their requirements on restoration may
differ introducing further challenges on the
mechanisms to be used. Resilience attributes
can be attached to an LSP telling how traffic
on that LSP can be restored in case of failure.
A basic attribute may indicate if all traffic
trunks in the LSP are transferred on a backup
LSP or some of the traffic is to be routed out-
side, e.g. following the routing protocols (see
Box C). Extended attributes may be intro-
duced giving indications like backup LSP is to
be pre-established, constraints for routing the
backup LSP, priorities when routing backup
LSP, and so forth. - Servers and content distribution. Location and
allocation of content on servers have signifi-
cant impact on the traffic distribution, in par-
ticular as long as much of the traffic is similar