5 8 0 NETWORK SURVIVABILITY These protection routing tables are similar to the routing tables maintained in IP networks, which work well even in very large IP networks with thousands of
Trang 15 8 0 NETWORK SURVIVABILITY
These protection routing tables are similar to the routing tables maintained
in IP networks, which work well even in very large IP networks with thousands
of nodes However, we need to realize that routing tables in IP networks are not always consistent If the tables are inconsistent, routing pathologies, such as looping, can be present in the network with fairly high probabilities For example,
at the end of 1995, the likelihood of encountering a major routing pathology
in the Internet was 3.3% [Pax97] These pathologies can cause packets to be forwarded incorrectly in the network, but these packets eventually find their way to their destination or are dropped by the network In the latter event, the packets are retransmitted by a higher-layer protocol (TCP) While this approach works well in IP networks, we cannot afford to have routing pathologies in transport networks because they could prevent restoration of service after a failure Therefore, fast and reliable topology update mechanisms need to be in place to maintain the protection routing tables
We now look at the different variations of mesh protection One aspect of this
is whether the entire network is protected as a single domain, or whether it is broken down into multiple domains, with each domain protected independently, and the different domains then tied together In a degenerate scenario, each domain could be a single ring, in which case we get back to the usual mode of ring-based protection
Another important aspect that differentiates protection schemes is whether the protection routes are precomputed ahead of time (offline), or whether they are com- puted after a failure has occurred (online) In both cases, another dimension to consider is the degree of distributed implementation This affects the complexity of the signaling protocols required and has a direct impact on the speed of restoration Let us first consider the case where the protection routes are precomputed In this case, the protection route for a lightpath is computed at the time it is set up and stored in the network Sufficient bandwidth is allocated on all the links so as to ensure the lightpath can be restored in the event of any possible failure (Note that this protection bandwidth is still shared among many lightpaths and is not dedicated
to a single lightpath This is the distinction between 1 + 1 dedicated protection and shared protection.) Depending on the sophistication of the scheme used, there may
be one or many possible alternate routes for a given lightpath, based on the actual failure scenario For example, the simplest scenario is to compute a single disjoint path through the network as the protection route Alternatively, we may use multiple protection routes, based on which link fails in the network Clearly the amount of information needed to be stored in the network depends on the number of protection routes per lightpath
Trang 210.5 Optical Layer Protection Schemes 581
In a centralized implementation of this scheme, a central controller in the network
is notified if a failure occurs The central controller then sets up all the alternate routes for the lightpaths by signaling to all the affected network elements to reconfigure their switches as needed The problem with this approach is that the central controller is
a single point of failure and is likely to be a significant bottleneck, both in terms of communication and processing speed
Several variants of a distributed implementation are possible In one variant, the failure information is flooded to all the network nodes Each node then looks up its routing table and reconfigures its switch, based on the exact failure that occurred Another possibility is to signal the failure to the sources/destinations of all the affected lightpaths Each source-destination pair then sets up the alternate routing path by signaling to the nodes along the new path
Next let us consider computing routes on the fly In this case, new routes are computed after the failure has been discovered One major issue that comes up in this context is whether sufficient bandwidth is available in the network to handle all the lightpaths that need to be restored Without essentially precomputing the routes,
it is not possible to determine the amount of protection bandwidth needed a priori
In this case, it is possible that some lightpaths are restored and others aren't Again this scheme can be implemented in a centralized or distributed manner The distributed implementation is more complex than for the case where routes are precomputed Here it is possible that multiple nodes acting independently may con- tend for the same link or wavelength resource to restore two independent lightpaths These contentions will have to be dealt with, making the signaling scheme more com- plex and the recovery possibly slower A centralized implementation would avoid such conflicts, but would suffer even worse communication and processing bottle- necks, compared to the centralized implementation for the case where the routes are precomputed
Based on our discussions so far, we see that mesh protection requires the fol- lowing functions: route computation, topology maintenance, and signaling to set
up the protection routes These functions have been implemented in IP and ATM networks For example, in IP networks, route computation is done using a Dijkstra shortest-path-first algorithm, and the topology is maintained using a routing pro- tocol such as OSPF (open shortest path first) Signaling has been used to establish paths in MPLS networks and ATM networks Several signaling protocols are avail- able for this purpose, including the resource reservation protocol (RSVP) [BZB+97], private network-network interface (PNNI) signaling protocol [ATM96], and Signal- ing System 7 (SS7) [ITU93] Today, there is a significant amount of work under way
to expand MPLS (called GMPLS, for generalized MPLS) [AR01] to provide similar capabilities in optical networks
Trang 3582 N E T W O R K SURVIVABILITY
10.5.9 Choice of Protection Technique
We have explored a number of different optical layer protection options It is still too early to determine which ones will be deployed widely An operator wanting to offer the different types of protection on the lightpaths as discussed in Section 10.4.1 must use an OCh layer protection method On the other hand, an operator who
is satisfied with protecting all lightpaths together will likely prefer an OMS layer scheme Many of the protection schemes discussed above are being implemented in commercial products
We have seen that protection functions can be done in the optical layer, SONET/SDH layers, or in the service layer (IP/ATM) How should protection in the network be coordinated between all these layers?
By default, the protection mechanisms in different layers will work independently
In fact, a single failure might trigger multiple protection mechanisms, all trying to restore service simultaneously, which would result in a large number of unneces- sary alarms flooding the management center This results in allocating protection bandwidth at each of the layers, which is inefficient
An area of significant concern is that protection mechanisms in different layers could potentially contend with each other, preventing or delaying service restoration, although careful design can eliminate such occurrences The following argument shows that multilayer protection schemes will eventually converge and restore traffic under the right assumptions:
Consider two network layers, a client layer operating over a server layer, each with its own protection mechanisms If the following conditions are met, the network will always restore traffic in the event of a failure:
1 A viable protection path exists for each layer
2 The server layer does not depend on the client layer to detect failures and invoke its protection-switching functions
3 The client layer protection is revertive in the sense that it will repeatedly try switching to the other path if its current path fails
Observe that since the server layer is independent of the client layer and does not depend on client layer indicators, in the event of a failure, the server layer will detect the failure and restore the traffic After the failure occurs, there may be a period of time when the client layer is unable to restore service because the server
Trang 4Summary 583
layer is invoking its protection scheme Ultimately, since the server layer converges, the client layer will see either a working path or a protection path available for it, and will therefore eventually converge
If any of the conditions above are not met, then the protection scheme may not converge For example, if the client layer protection is ~nonrevertive, it may switch over once to the protection path, discover that path is not available, and not switch back to its primary path
While it is desirable to have some sort of coordination between protection mecha- nisms in different layers, this may not always be possible For example, the protection mechanisms in different layers may actually be activated by different nodes In some cases, it may be possible to add a priority mechanism where one layer attempts to restore service first, and only afterwards does the second layer try One automatic way to ensure this is to have the restoration in one layer happen so quickly that the other layer doesn't even sense that a failure has occurred For example, consider a
W D M network carrying IP traffic As we saw in Section 10.3, it can take several seconds for the IP layer to detect a failure It is entirely feasible for the optical layer
to have completed its restoration within this time scale so that the IP layer doesn't detect the failure This may not, however, be feasible when we have SONET rings operating over a W D M network The SONET rings detect failures very quickly and can initiate protection switching as early as 2.3 #s after a failure occurs
Another way to implement orderly restoration would be to impose an additional
hold-off time in the higher layer before it attempts restoration so as to provide sufficient time for the lower layer to do its restoration However, a large hold-off time would increase the overall restoration time and is therefore not highly desirable either In general, it would make sense to have the priorities arranged such that the layer that can provide the fastest restoration tries first
Summary
Engineering the network for survivability plays an increasingly important role in transport networks Protection techniques are well established in SONET and SDH and include point-to-point, dedicated protection rings, and shared protection rings Point-to-point protection schemes work for simple systems with diverse fiber routes between node locations Dedicated protection rings are primarily used to aggregate traffic from remote locations to one or two hub locations Shared protection rings are used in the core parts of the network where the traffic is more distributed Protection in the optical layer is emerging, with several commercial products now implementing optical layer protection Optical layer protection is needed to protect the data services that are increasingly being transported directly on the optical layer
Trang 55 84 N E T W O R K SURVIVABILITY
without the SONET/SDH layer being present It can also be more efficient with respect to reducing the protection bandwidth required (by sharing the bandwidth across multiple clients) and therefore more cost-effective
Optical channel layer protection is needed if some channels are to be protected while others are not Optical multiplex section layer protection is more cost-effective for those cases where all the traffic needs to be protected There is a growing trend toward the use of shared mesh protection in the optical layer, which is viewed as being more bandwidth-efficient and flexible, compared to the traditional ring-based approaches
Further Reading
There is a vast literature on protection in SONET and SDH networks SONET rings and protection schemes are described in ANSI T1.105.1 and Telcordia GR-253 and GR-1230 ITU G.841 describes the equivalent SDH architectures We also refer the reader to the books by Sexton and Reid [SR97] and Wu [Wu92]
Providing reliable service in IP and MPLS networks is a topic of great interest today Several protection schemes are being developed See, for example, [DR00, Section 7.4], [CO99], and several Internet drafts available at www.ietf.org
There is a lot of activity under way on optical layer protection schemes, with several being implemented in products today These have not yet been standard- ized [DWY99, RM99a, RM99b, Ram01, MM00, Bar00, GR00a, GR00b, Dos99, MBN99, Wu95, WO95, Tel98, GR96, GRS97] provide good coverage of the major issues Interworking of protection schemes between different layers is covered in [Dem99, MB96]
10.1
10.2
10.3
Problems
Consider a shared protection ring with two types of restoration possible In the first scheme, the connection is rerouted by the source and destination around the ring in the event of a failure In the second, the connection is rerouted around the ring by the nodes adjacent to the failed link (as in a BLSR) Give an example of a traffic pattern where the first scheme uses less ring bandwidth than the second Give another example where the two require the same amount of bandwidth
Show that in a ring architecture if the protection capacity is less than the working capacity, then service cannot be restored under certain single failure conditions Compare the performance of UPSRs and BLSR/2s in cases where all the traffic is between a hub node and the other nodes Assume the same ring speed in both
Trang 6Problems 585
Figure 10.24 Network topology for Problem 10.6
10.4
10.5
10.6
cases Is a BLSR/2 any more efficient than a UPSR in traffic-carrying capacity in this scenario?
Construct a traffic distribution for which the traffic-carrying capacity of a BLSR/4 is maximized What is this capacity as a multiple of the bit rate on the working fibers? Assuming a uniform traffic distribution, compute the traffic-carrying capacity of a BLSR/4 as a multiple of the bit rate on the working fibers
Consider the topology shown in Figure 10.24 over which STS-ls are to be transported
as dictated by the bandwidth demands specified in the table below for each node pair Assume all the bandwidth requirements are bidirectional
Given the fiber topology and the STS-l-based bandwidth requirements, we will utilize a two-fiber OC-N SONET ring architecture, but we need to determine which SONET ring architecture is the most suitable for the given n e t w o r k ~ t h e UPSR or the BLSR/2
(a) Provide a detailed illustration of how the six STS-ls between nodes A and
C would be transported by a UPSR and a BLSR/2 Redraw Figure 10.24 to begin each illustration
(b) Suppose that a backhoe cuts the fiber pair between nodes B and C Again, redrawing Figure 10.24 and referencing your illustrations above, provide a detailed illustration of how the six STS-ls between nodes A and C would be transported just after this failure for the UPSR and the BLSR/2 Use dashed lines to highlight any differences in the routing from normal operation
Trang 75 8 6 NETWORK SURVIVABILITY
10.7
10.8
10.9
10.10
10.11
10.12
(c) Using the bandwidth demands given in the table above, design best-case ring routing plans for the UPSR and the BLSR/2 Illustrate the routing on the network topology of Figure 10.24 In addition, specify the quantity of STS-ls being transported over each fiber link for both cases
(d) Assuming that we want to use a single OC-N ring, what would be the minimum standard value of N in each case for the designed UPSR and BLSR/2?
(e) Given all of this information, which ring architecture is better suited for this application? Briefly explain your reasoning
The UPSR, BLSR/4, and BLSR/2 are designed primarily to handle single failures However, they can handle some cases of simultaneous multiple failures as well Carefully characterize the types of multiple link/node failure combinations that these different architectures can handle
The 1 + 1 protection in a SONET UPSR is not implemented at a fiber level but at
an individual SONET connection level: for each connection, the receiver picks the better of the two paths An alternative and simpler approach would be to have the receiver simply pick the better of the two fiber lines coming in, say, based on the bit error rate In this case, the receiver would not have to look at the individual connections in order to make its decision, but rather would look at the error rate of the composite signal on the fiber Why doesn't this work?
Suppose you had only two fibers but could use two wavelengths, say, 1.3 # m and 1.55/~m, over each fiber This can be used to deploy a BLSR/4 ring in three different ways: (1) the two working fibers could be multiplexed over one fiber and the two protection fibers over the other, (2) a working fiber and a protection fiber in the same direction could be multiplexed over one fiber, or (3) a working fiber and a protection fiber in the opposite direction could be multiplexed over one fiber Which option would you choose?
Consider a four-fiber BLSR that uses both span and ring switching What are the functions required in network management to (a) coordinate span and ring switching mechanisms, and (b) allow multiple failures to be restored?
Consider the example shown in Figure 10.14 Carefully characterize the set of simul- taneous multiple fiber cuts that can be handled by this arrangement
Consider a five-node optical ring with one hub node and four access nodes The traf- fic to be supported is one lightpath between each access node and the hub node You can deploy either a two-fiber OCh-DPRing or a two-fiber OCh-SPRing in this appli- cation No wavelength conversion is allowed inside the network, so each lightpath must use the same wavelength on every link along its path Compare the amount of
Trang 8References 587
10.13
protection and working capacity needed for each case Using a wavelength on a link counts as one unit of capacity Would your answer change if wavelength conversion was allowed in both types of rings at any node in the ring?
Develop computer software that performs the following functions:
(a) Allows you to input a network topology graph and a set of lightpaths (source- destinations)
(b) Routes the lightpaths using a shortest-path algorithm
(c) Computes protection bandwidth in the network for two cases: 1 + 1 0 C h protection and OCh shared mesh protection
For 1 + 1 0 C h protection, use an algorithm to provide two disjoint shortest paths for each lightpath, such as the one in [ST84] For shared mesh protection, use the following algorithm: for each failure i, determine the amount of protection capacity, Ci(l), that would be required on each link 1
in the network Prove that the total protection capacity needed on link 1 is then simply maxi Ci (l)
(d) Experiment with a variety of topologies, traffic patterns, and different routing/protection computation algorithms Summarize your conclusions
References
[AJY00] C Alaettinoglu, V Jacobson, and H Yu Towards millisecond IGP convergence In
North American Network Operators Group Fall Meeting, 2000 See also IETF drafts draft-alaettinoglu-isis-convergence-OO.txt and draft-ietf-ospf-
scalability-OO.txt
JAR01] D Awduche and Y Rekhter Multiprotocol lambda switching: Combining MPLS
traffic engineering control with optical crossconnects IEEE Communications Magazine, 39(4):111-116, Mar 2001
[ATM96] ATM Forum Private Network-Network Interface Specification: Version 1.0, 1996 [Bar00] S Baroni et al Analysis and design of backbone architecture alternatives for IP
optical networking IEEE Journal of Selected Areas in Communications,
18(10):1980-1994, Oct 2000
[Bat98] R Batchellor Optical layer protection: Benefits and implementation In
Proceedings of National Fiber Optic Engineers Conference, 1998
[BZB+97] R Bradon, L Zhang, S Berson, S Herzog, and S Jamin Resource Reservation
ProtocolmVersion 1 Functional Specification Internet Engineering Task Force, Sept 1997
Trang 95 8 8 NETWORK SURVIVABILITY
[CO99] T.M Chen and T H Oh Reliable services in MPLS IEEE Communications
Magazine, 37(12):58-62, Dec 1999
[Dem99] P Demeester et al Resilience in multilayer networks IEEE Communications
Magazine, 37(8):70-77, Aug 1999
[Dos99] B.T Doshi et al Optical network design and restoration Bell Labs Technical
Journal, 4(1):58-84, Jan.-March 1999
[DR00] B.S Davie and Y Rekhter MPLS Technology and Applications Morgan
Kaufmann, San Francisco, 2000
[DWY99] P Demeester, T.-H Wu, and N Yoshikai, editors IEEE Communications
Magazine: Special Issue on Survivable Communication Networks, volume 37, Aug
1999
[GR96] O Gerstel and R Ramaswami Multiwavelength optical network architectures and
protection schemes In Proceedings of Tirrenia Workshop on Optical Networks,
pages 42-51, 1996
[GR00a] O Gerstel and R Ramaswami Optical layer survivabilityma services perspective
IEEE Communications Magazine, 38(3):104-113, March 2000
[GR00b] O Gerstel and R Ramaswami Optical layer survivability: An implementation
perspective IEEE JSAC Special Issue on Optical Networks, 18(10):1885-1899,
Oct 2000
[GRS97] O Gerstel, R Ramaswami, and G H Sasaki Fault tolerant WDM rings with
limited wavelength conversion In Proceedings of IEEE Infocom, pages 508-516,
1997
[HYCG00] G Hjalmtysson, J Yates, S Chaudhuri, and A Greenberg Smart routers simple
optics: An architecture for the optical Internet IEEE/OSA Journal on Lightwave Technology, 18(12):1880-1891, 2000
[ITU93] ITU-T Recommendation Q.700" Introduction to CCITT Signaling System No 7,
1993
[Kha97] S Khanna A polynomial time approximation scheme for the SONET ring loading
problem Bell Labs Technical Journal, 2(2):36-41, Spring 1997
[LC97] C.Y Lee and S G Chang Balancing loads on SONET rings with integer demand
splitting Computer Operations Research, 24(3 ):221-229, 1997
[MB96] J Manchester and P Bonenfant Fiber optic network survivability: SONET/optical
protection layer interworking In Proceedings of National Fiber Optic Engineers Conference, pages 907-918, 1996
[MBN99] J Manchester, P Bonenfant, and C Newton The evolution of transport network
survivability IEEE Communications Magazine, 37(8):44-51, Aug 1999
Trang 10References 589
[MM00] G Mohan and C S R Murthy Lightpath restoration in WDM optical networks
IEEE Network Magazine, 14(6):24-32, Nov.-Dec 2000
[Pax97] V Paxson End-to-end routing behavior in the Internet IEEE/ACM Transactions
on Networking, 5(5):601-615, Oct 1997
[Ram01] R Ramamurthy et al Capacity performance of dynamic provisioning in optical
networks IEEE/OSA Journal on Lightwave Technology, 19(1):40-48, 2001 [RM99a] B Ramamurthy and B Mukherjee Survivable WDM mesh networks, Part
Improtection In Proceedings of IEEE Infocom, pages 744-751, 1999
[RM99b] B Ramamurthy and B Mukherjee Survivable WDM mesh networks, Part
II restoration In Proceedings of IEEE International Conference on
Communication, pages 2023-2030, 1999
[SR97] M Sexton and A Reid Broadband Networking: ATM, SDH and SONET Artech
House, Boston, 1997
[ST84] J.W Suurballe and R E Tarjan A quick method for finding shortest pairs of
disjoint paths Networks, 14:325-336, 1984
[Tel98] Telcordia Technologies Common Generic Requirements for Optical Add-Drop
Multiplexers (OADMs) and Optical Terminal Multiplexers (OTMs), Dec 1998 GR-2979-CORE, Issue 2
[WO95] L Wuttisittikulkij and M J O'Mahony Multiwavelength self-healing ring
transparent networks In Proceedings of IEEE Globecom, pages 45-49, 1995 [Wu92] T.H Wu Fiber Network Service Survivability Artech House, Boston, 1992 [Wu95] T.H Wu Emerging techniques for fiber network survivability IEEE
Communications Magazine, 33(2):58-74, Feb 1995