arc_ngothanhhai
New Member
Download miễn phí Đề tài Shared Protection for Multi-Domain Optical Mesh Networks
Contents
CHAPTER 1 INTRODUCTION . 6
1.1 MOTIVATION AND PROBLEM DESCRIPTION . 6
1.2 OBJECTIVE . 10
1.2.1 Network Model . 10
1.2.2 Goals to Achieve . 10
1.3 REPORT ORGANIZATION . 11
CHAPTER 2 STATE OF THE ART . 12
2.1 BACKGROUND. 12
2.1.1 Node & Link . 12
2.1.2 Protection . 12
2.1.3 Shared Risk Group. 13
2.1.4 Protection Categories . 14
2.1.4.1 Node vs. Link Protection.15
2.1.4.2 Link-based, Path-based and Segment-based Protection .15
2.1.4.3 Static vs. Dynamic Protection .18
2.1.4.4 Dedicated Protection vs. Shared Protection.18
Concluding Remarks . 20
2.2 PATH COMPUTATION FOR SHARED PROTECTION . 21
2.2.1 Path Computation for Shared Path-based Protection . 22
2.2.1.1 Accessible Bandwidth for Backup Paths .22
2.2.1.2 Link Costs and Cost Function .23
2.2.1.3 TSA for shared protection .26
2.2.1.4 Joint Derivation of working and backup paths .26
2.2.1.5 APF-BPC .27
Synthesis .28
2.2.2 Path Computation in Overlap Segment Shared Protection . 29
2.2.2.1 Method of SLSP .29
2.2.2.2 Graph Transformation Method.30
2.2.2.3 Method of PROMISE.31
Synthesis .32
2.3 INTER-DOMAIN PROTECTION . 33
2.3.1 Path-Based Protection Approach . 33
2.3.1.1 Protection Model .34
2.3.1.2 Path Computation and Path Request Signaling .35
2.3.1.3 Failure Notification .36
2.3.1.4 Backup path setup signaling.36
2.3.2 Multiple Single-Domain Protections Approach. 36
2.3.2.1 Protection Model .36
2.3.2.2 Path Computation.37
2.3.2.3 Notification and Backup Path Setup Signaling.38
2.3.3 SRLG-Disjoint Routing for Multi-Segment Protection Approach . 38
2.3.3.1 Protection Model .38
2.3.3.2 Path computation.40
2.3.3.3 Notification and Backup Path Setup Signaling.41
Synthesis . 42
2
CHAPTER 3 PROPOSAL. 43
3.1 PROBLEM STATEMENT AND OBJECTIVES . 43
3.2 PROPOSED APPROACH. 44
3.2.1 Multi-Domain Network Model . 44
3.2.2 General Approach . 45
3.2.2.1 The choice of the protection model .45
3.2.2.2 Path computation problem.46
3.2.2.3 Optimization issue.47
3.2.2.4 Communication between network nodes .49
3.2.3 Sub-problems to resolve . 49
3.3 SEGMENTATION STRATEGIES . 50
3.4 COST FUNCTIONS OF THE TWO COMPUTATIONS. 52
3.5 PATH REQUEST SIGNALING . 56
3.6 PATH COMPUTATION AT INTER-DOMAIN LEVEL . 58
3.7 PATH COMPUTATION AT INTRA-DOMAIN LEVEL . 61
3.7.1 General Approach . 61
3.7.2 Resource Cost and Notations. 63
3.7.3 Working Virtual Link Realization . 65
3.7.4 Backup Virtual Link Realization. 66
3.7.5 Information Distribution. 68
3.8 TE INFORMATION AND EXTENDED COSTS OF VIRTUAL LINKS. 70
3.8.1 Extended Working Resource Cost . 70
3.8.2 Extended Backup Resource Cost . 72
3.8.2.1 Virtual Link Disjointness .72
3.8.2.2 Virtual Link Backup Cost.74
3.8.3 Information Distribution. 77
3.9 NOTIFICATION AND BACKUP PATH SETUP SIGNALING . 77
3.10 VALIDATION OF THE PROPOSED SOLUTION . 78
CHAPTER 4 ACHIEVED RESULTS AND REMAINING WORK . 80
4.1 ACHIEVED RESULTS . 80
4.1.1 Theoretical Results . 80
4.1.2 Pragmatic Result . 80
4.2 REMAINING WORK. 80
4.3 PROSPECTIVE EXTENSION . 81
CHAPTER 5 WORKING PLAN . 82
5.1 RESEARCH PLAN . 82
5.2 PUBLICATION PLAN. 83
CHAPTER 6 CONCLUSION . 84
REFERENCES 85
Để tải bản Đầy Đủ của tài liệu, xin Trả lời bài viết này, Mods sẽ gửi Link download cho bạn sớm nhất qua hòm tin nhắn.
Ai cần download tài liệu gì mà không tìm thấy ở đây, thì đăng yêu cầu down tại đây nhé:
Nhận download tài liệu miễn phí
Tóm tắt nội dung tài liệu:
r node and no more nodes are recorded because the backup path does not passthrough AS 3. B13 linked to B8 continues the same process as B6. Furthermore, this time
some nodes to exclude are added into the exclusion list because backup path goes though
AS 4 also. And so on until B is reached. After that, the working path is set up. Next, a
similar signaling process carrying the exclusion list begins from A, looking for backup
segments within every AS in the AS path of the backup path. Certainly, all nodes in the
exclusion list must be excluded in the search. The backup path is after that reserved.
This search assures that the two paths are node-disjoint but as we can notice, no
information about the possibility of sharing bandwidth with other backup paths is taken
into account. Furthermore, AS paths are calculated by using the algorithms in
[Suurballe74] [Suurballe84] [Bhandari99], which have no ideas of shared protection.
Furthermore, backup paths are reserved without considering the possibility of reusing the
bandwidth of other backup paths. Obviously, no backup resource sharing is supported and
the protection can be classified as dedicated protection.
36
2.3.1.3 Failure Notification
The original document does not mention the failure notification. Since this is path-based
protection, the failure notification process should begin from the node adjacent to the
failure; propagate along the failed lightpath in two directions until reaching to the two end
nodes. Of course, this is a long notification over multiple AS’s.
2.3.1.4 Backup path setup signaling
As for notification, backup path setup is not clearly mentioned in the cited document.
However, because the protection is proposed for MPLS, the backup path reservation
process at the end of the path computation also includes backup LSP1 setup. No signaling
is necessary to activate backup paths after failures.
If we apply this approach to the optical network, backup paths can be set up before
failures because they are dedicated for their working paths.
2.3.2 Multiple Single-Domain Protections Approach
We refer to the work of Akyamaç et al. presented in [Akyamac02a] as “Multiple single
domain protections”. This work is a further improvement of the work in [Ou01].
2.3.2.1 Protection Model
The work deals with a multi-domain optical network that is similar to ours. However, it
considers that the set of border nodes and inter-domain links of two neighboring domains
also forms a domain. One domain connects to another by at least two gateway nodes:
primary and secondary gateways (Figure 14).
A segment-based protection with link-disjoint segments is proposed. End-to-end lightpath
consists of smaller lightpath segments within the domains it crosses. Segments begin and
end at primary gateways. Each working segment has a backup segment bounded in its
domain. The two segments join at the primary gateway. When the working segment fails
at a cable or an intermediate node, the backup segment is activated by the primary
gateways. Thus, routing and restoration is strictly limited to each domain and independent
from one domain to another. The recovery will be faster than that of CCAMP’s approach.
However, when the primary gateway fails, “the secondary gateway is used to restore
against the failure of the primary gateway node (handle via lightpath re-provisioning)”
(cited from the original document). That means a new end-to-end backup path is
calculated.
1
Label Switched Path
37
Thanks to the independent recovery in different domains, this solution can protect against
multiple simultaneous failures in various domains. Since working and backup segments
are bounded in the same domain, the resource sharing between backup segments of the
domain can be enabled easily by using single-domain path computation algorithms for
shared protection.
Figure 14: Multiple single domain protections approach (from [Akyamac02a])
However, the protection scheme strictly requires that working and backup segments cross
in parallel over the same domains. This limits the choice of backup segments implying
increasing blocking probability due to resource insufficiency for the two segments, which
must also be node-disjoint. Meanwhile, there are still other backup segments crossing
other domains than those of the working segment and they can protect the working
segment.
Further more, the lightpath re-provisioning on the fly, in the case the primary gateway
node fails, transforms the proposed “protection” into “restoration”. This implies a QoS
declining in terms of recovery guarantee as well as restoration time. Supposing that the
alternative path using the secondary gateway of the failed primary gateway does not exist,
the lightpath cannot be recovered and the affected traffic must be dropped. In fact, border
nodes are intensive-traffic points. Multi-domain lightpaths transit more frequently
through border nodes than internal nodes. Low quality of recovery at border nodes may
cause a severe impact. According to us, restoration for border nodes is inadequate.
2.3.2.2 Path Computation
The original document does not mention the end-to-end path computation. We know only
that working and backup segments are computed within each domain as a single-domain
38
lightpath. Therefore, all path computation methods for single-domain shared protection
can be applied to identify working and backup segments between gateway nodes. It
remains always to identify the path of the gateways of the end-to-end path.
2.3.2.3 Notification and Backup Path Setup Signaling
As for path computation, the notification and backup path signaling are limited to the
scale of one domain. Obviously, this approach offers a much faster notification and
signaling than CCAMP.
2.3.3 SRLG-Disjoint Routing for Multi-Segment Protection Approach
S1 1 2 D
43S2
a) Physical layer topology
SRLG #1 SRLG #3
SRLG #5
SRLG #4
SRLG #2
S1 1 4 D
23S2
b) Optical layer topology
SRLG #1 SRLG #4
SRLG #3
SRLG #5
SRLG #6
SRLG #4SRLG #2
common
conduit
SRLG #6
Figure 15: Example of two node-disjoint paths that belong to the same SRLG
In this section, we review the approach proposed by T. Miyamura et al. in [Miyamura04].
The work is reported providing an SRLG-disjoint routing algorithm for multi-segment
protection in inter-area networks. There are two main contributions. First, routing for
multi-segment protection is defined. Second, the routing guarantees that two working and
backup paths are SRLG-disjoint. The condition SRLG-disjoint is more general than the
link-disjoint constraint at the optical layer because the physical and optical topologies
may be different. Two fibers that are node disjoint at the optical layer are not necessarily
SRLG disjoint physically if they are bundled in the same conduit. A failure at the
common conduit causes the failures of both fibers even though they are node-disjoint.
Figure 15 shows an example of such paths: S1-1-4-D, S2-3-2-D. The links that have the
same risk are assigned the same SRLG identification.
2.3.3.1 Protection Model
Before talking about the protection model, let us introduce the inter-area network model
referred to by this work. Initially, the multi-area MPLS network, which is seen from a
39
routing viewpoint as a multi-area OSPF network [Moy98], is referred to. The multi-area
OSPF network contains one backbone and various non-backbone areas as depicted in
Figure 16. Border routers connect directly to different areas. All border routers belong to
the backbone area. The connectivity of the backbone area is maintained by configuring
virtual links between the border routers of the same area. The concrete routes of a virtual
link cross internal routers of one area, thus they are invisible from the border routers of
the other areas. For example, in Figure 16, B4 does not know the concrete routes of the
virtual link B2-B3. The topology of the backbone dictates the backbone paths used
between areas [Moy98]. Those paths are in fact the virtual link paths.
backbone area
border router
virtual link
internal router
B1
B2
B4
B5
B3
S D
area 4
area 1
area 2
area 3non-backbone
area
a physical route of
the virtual link B3-B4
Figure 16: Multi-area network
However, the authors based their work implicitly on a simpler network model to build
their path computation algorithm. In that model, two non-backbone areas reach each other
without passing through other intermediate non-backbone areas (Figure 17). This model
may be suitable for modeling the metro/core networks relationship but does not reflex
general multi-area networks.
T. Miyamura et al. propose a special Segment-based Protection scheme (Fi...