CSC 791B - Sp Top: Adv Netw Dsgn

Download Report

Transcript CSC 791B - Sp Top: Adv Netw Dsgn

CSC 778 - Survivability
Anuj Dewangan
Parinda Gandhi
Outline
 Survivability in Optical Layer
 Survivability in IP
 IP over WDM
 Dynamic Provisioning/Protection Scheme
 Survivability in IP over WDM
 Restoration Scheme for IP over WDM Networks
Survivability in Optical Layer
Need for Optical Layer Survivability
 Optical fibers in WDM networks carry a large amount of data
 The optical layer can provide faster service recovery than the
higher layers
 The optical layer survivability can provide protection and
restoration with significant cost savings.
Cons:
 Because protocol transparent, it is unable to detect increased
Bit Error Rates (BER)
Failures and Optical Layer Survivability
 Link failure, node and channel failures
Link failure much more likely
 Single failure and multiple failure
Survivability against multiple failures is prohibitively
costly and unlikely.
 Our discussion is restricted to single link failures only
Optical Layer Survivability Paradigms
 Static Protection
Spare network resources are reserved during network design or at the
time of connection establishment for protection against network failures.
Faster but inefficient in resource utilization
 Dynamic Restoration
The network searches dynamically for spare network services after
network failures occurs
Better in resource utilization but slower and has no guarantees
Static Protection
Two ways to classify
 Dedicated protection and shared protection
 Dedicated Protection: 1:1 and 1+1
 Shared Protection: 1 : N
 Link protection and path protection
 Link protection: Dedicated and shared link protection
 Path protection: Dedicated and shared path protection
Static Protection – Dedicated Protection
1 + 1 Protection
 If the working link fails, the receiver only needs to simply
switch over to the protection link
 The advantage of 1+1 protection is that it can provide very
fast service recovery.
Static Protection – Dedicated Protection
1 :1 Protection
 If the working link fails, the receiver and sender both need
to switch over to the protection link
 In absence of failure, the protection link can be used to
transmit a signal that carries low priority traffic
 Because only the receiving end can detect the failure, the
receiving end must use a signaling protocol to notify the
transmitting end of the failed link so that the transmitter can
switch over to the protection link
Static Protection – Shared Protection
1:N Protection
 It can handle the failure of any single working link only.
 It increases link utilization of the protection link.
Static Protection – Link Protection
Line
SpanProtection
Protection
 A backup path or protection path is reserved for each
link of the primary path during the establishment of the
connection
 The recovery is handled at the end nodes of the failed
link
 The source and destination nodes are unconscious of
the link failure
 The wavelength used for the backup path should be
the same as that used for the primary path
Static Protection – Path Protection
 A back up path is reserved for the primary path on an end-toend basis during the establishment of a connection
 The backup path should be link disjoint
 This requires the end nodes of the failed link to inform the
source and destination nodes of the link failure
 Longer service recovery time
 Backup path wavelength need not be the same
Static Protection – Path Protection
Dedicated
Path
Protection
Shared Path
Protection
Dynamic Restoration
Two ways to classify
 Link Restoration
 End nodes of a failed link dynamically search for a backup path for each
connection that traverses the failed link
 If no backup path found, the connection is blocked
 Path Restoration
 Source and Destination nodes of each connection dynamically search for a
backup path on an end-to-end basis in event of a link failure
 If no backup path found, the connection is blocked
Outline
 Survivability in Optical Layer
 Survivability in IP
 IP over WDM
 Dynamic Provisioning/Protection Scheme
 Survivability in IP over WDM
 Restoration Scheme for IP over WDM Networks
Survivability in IP
Survivability in IP layer
 Achieved by rerouting through the convergence of routing
information after the detection of a failure
 Best effort in nature
Advantages:
 Ability to find optimal routes through the network
 Ability to provide a finer granularity of protection
 Slow in nature
Outline
 Survivability in Optical Layer
 Survivability in IP
 IP over WDM
 Dynamic Provisioning/Protection Scheme
 Survivability in IP over WDM
 Restoration Scheme for IP over WDM Networks
IP over WDM Networks
IP over WDM Network Architecture
 IP packets are directly carried over WDM networks
 Hence flexible bandwidth allocation has to be handled at the
WDM layer: Dynamic provisioning
IP over WDM Network Interconnection Models
 Defines interconnection between the IP and the optical network
 Overlay Model:
Two separate control planes
The IP layer acts as a client to the Optical layer
The internal topology of the optical network is not visible
to the IP networks
Communication between the layers is through UNIs
IP router register their IP addresses with the optical
network and request for lightpath creation or deletion
IP over WDM Network Interconnection Models
 Peer model:
There is only a single control plane
Hence the optical domain is transparent to the IP routers
Each OXC also need to be an IP router and be IP
addressable
The routers in the IP network and Optical network can run
routing protocols like OSPF or IS-IS with appropriate extensions
Now, an edge router can create an end-to-end connection
using MPLS based signaling
Outline
 Survivability in Optical Layer
 Survivability in IP
 IP over WDM
 Dynamic Provisioning/Protection Scheme
 Survivability in IP over WDM
 Restoration Scheme for IP over WDM Networks
Dynamic Provisioning/Protection
Scheme
Dynamic Provisioning/Protection Scheme
 Focuses on service reliability for Border LSRs
 It is a protection scheme
 It dynamically computes a primary lightpath and a backup path,
when LSRs request a lightpath
 The flow is blocked if either the primary or the backup path fails
to establish
 Initially, based on aggregated traffic demands, a virtual
topology is designed by using optimization approach. A backup
path is also reserved
Dynamic Provisioning/Protection Scheme
Dynamic Provisioning/Protection Scheme
Notlightpath
enough
bandwidth
available
from
the
New
from
A’ toconnected
F’ needs to
 F’ not
directly
tobe
A’setup
logical topology
Dynamic Provisioning/Protection Scheme
 Since
bandwidth
backup path is α f(k) <
Need for
a backupofpath
f(k), logical topology A’-E’ and E’-F’ can be
selected
Dynamic Provisioning/Protection Scheme
Dynamic Provisioning/Protection Scheme
Outline
 Survivability in Optical Layer
 Survivability in IP
 IP over WDM
 Dynamic Provisioning/Protection Scheme
 Survivability in IP over WDM
 Restoration Scheme for IP over WDM Networks
Survivability in IP over WDM
IP over WDM
 WDM technologies with high bandwidth capacity are going to
play a dominant role in future networks
 In WDM, lightpaths are set up to provide end-to-end
connections between Optical cross connects
 IP/WDM is a simple example of multilayer network where IP
layer resides above an optical network
Multilayer Survivability in IP over WDM
 Protection/Restoration capability at IP and WDM layer
 IP layer provides a finer granularity of protection
Example: Packet level or LSP level
 WDM layer provides protection at a coarse granularity
Example: Fiber or Wavelength level
 Service Recovery at IP is slower than WDM layer
 MPLS based IP networks provide fast restoration/protection
capabilities compared to IP layer
Mechanisms for MPLS recovery
 End-to-End Path Protection (Path level)
 Local protection (link level)
 Local loopback
 Rerouting
LSP 1
LSP 1
LSR
LSR
Ingress LSR
LSP 2
LSP 1
Egress LSR
LSR
LSR
LSP 2
End-to-End Path Protection
LSP 2
Failure detection in IP over WDM
 Failure at IP/MPLS layer cannot be cannot be detected at the
WDM layer and vice versa
 Failure at IP/MPLS layer can only be recovered by itself while
failure at WDM layer can be recovered by both layers
 Important to coordinate service recovery at IP/MPLS and
WDM layer in an effective manner
Physical failure at the WDM layer
 Strategies for service recovery at IP/MPLS and WDM layers
- Parallel and Sequential Strategies
 In parallel strategy service recovery is initiated in IP/MPLS
and WDM layers simultaneously
Problems:
- Service recovery at WDM layer is faster than IP
- Difficult to coordinate resulting in insufficient resource
utilization or even failure in service recovery
Physical failure at the WDM layer
 In sequential strategy service recovery is activated at the
WDM and the IP/MPLS layers in a sequential manner
Problem:
 Escalate failure detected at the WDM layer to IP/MPLS layer
 Coordination can be implemented using hold-off timer method
or recovery token method
Physical failure at the WDM layer
 Escalation strategies for coordinating multi-layer service recovery
- Bottom-up
- Top-down
IP/MPLS
Signaling
IP/MPLS
Signaling
WDM
Signaling
WDM
Signaling
Bottom-up
Top-down
Outline
 Survivability in Optical Layer
 Survivability in IP
 IP over WDM
 Dynamic Provisioning/Protection Scheme
 Survivability in IP over WDM
 Restoration Scheme for IP over WDM Networks
Restoration Scheme in IP over WDM
Networks
Restoration scheme in IP/WDM
 In which layer should one provide network survivability ?
 Survivability at each layer has its own pros and cons
 Solution: Two layer recovery mechanism
LSP 1
LSP 2
LSP 1
LSR/OXC
LSP 2
LSR/OXC
LSR/OXC
Optical network
LSR/OXC
LSP 2
LSP 1
LSR/OXC
LSR/OXC
LSP 2
IP router
IP router
Two layer recovery mechanism
 Bottom-up approach
 Restoration used instead of protection
 Path switching used due to limited wavelength resources
around failed link
 Full wavelength conversion is assumed at every node in the
network
Two layer recovery mechanism
Algorithm for Optical layer recovery
 Construct a graph G based on physical topology where each
edge containing vertices <s , d> represents there are spare
wavelengths in the link
 On receiving failure notification for a light path, at the source
OXC Dijkstra’s algorithm is used to compute a new light path
with minimum hops from source to destination implying
minimum number of wavelengths are used to reroute the light
path
Two layer recovery mechanism
 If the light path cannot be restored at the Optical layer the IP
layer has to take care of rerouting the LSPs carried by the
affected light path.
 Existing light paths with spare bandwidths or new lightpaths
can be can be established to reroute the affected LSPs
 Various algorithms have been proposed
Simulation results
Topology
Performance of two layer restoration
with recovery token is compared with
single IP/MPLS restoration
Comparison of hold-off timer versus
Recovery Token mechanism
References
 Dongyun Zhou and Suresh Subramaniam,
“Survivability in Optical Networks”,
IEEE, 2000.
 Jun Zheng and Hussein T. Mouftah, “Optical WDM Networks- Concepts and Design
Principles”, Wiley, 2004
 Arun K. Somani, “Survivability and Traffic Grooming in WDM Optical Networks”,
Cambridge University, 2006
 Yinghua Ye, Chadi Assi, Sudhir Dixit, Mohammed A. Ali, “A Simple Dynamic
Integrated Provisioning/Protection Scheme in IP over WDM Networks”, IEEE, 2001
 Yang Qin, Lorne Mason, Ke Jia, “ Study on a Joint Multiple Layer Restoration
Scheme for IP over WDM Networks”, IEEE, 2003
http://cufagradforum.files.wordpress.com/2007/05/questions.gif
Thank You