Collaboration PoliMi – Nokia Vimercate

General considerations

In the following document there is the detailed description of the proposed projects. The number of people associated to each project depends on the different scenarios the students want to investigate (e.g., if ILP or GA have to be implemented or not, implementation of the Auxiliary Graph, …).

Any project could be modified and simplified depending on the number of students that want to work on the project.

About the simulation assumptions, students are invited to use the data provided in Section 2.

In case there is any doubts on the project aim or inputs, students are invited to contact me at the following email address: annalisa.morea@nokia.com

1. Project descriptions

Project 1: Compare different routing strategies for elastic optical networks (3/5 people)

Goal: Benchmark routing, grooming, modulation level and spectrum assignment (RGMLSA) algorithm with diverse end-to-end aggregation strategies. The comparison shall be done in terms of deployed OEO devices (transponders for add/drop and regenerators) and free spectrum per link.

In backbone optical networks, client traffic demands are set up along several periods and, in each period, a routing, grooming, modulation level and spectrum assignment (RGMLSA) algorithm is used to determine the resources required to accommodate each demand. A key part of the decision process is the Optical Channel (OCh) format do deploy, balancing factors such as reach, capacity and spectral occupation.

The traffic generated in the network usually has a sub-lambda granularity (e.g., 10s of Gb/s) and it is groomed together to fill as much as possible the OCh transported in the network, which propagates at wavelength of the order of 100s Gb/s. For the proposed study, demands are aggregated and routed end-to-end (i.e. no intermediate grooming). RSGA algorithms can enforce one of three different options described in the following. Each option impacts the OCh format selected in each execution and has an impact on the network efficiency, particularly with respect to transported capacity, spectral efficiency and spectral occupation.

RGSA Methods to compare: The three RGSA methods to compare are the following:

1) MSE-MaxC (Most Spectral Efficient – Maximum Capacity): This method selects the OCh format/path with highest spectral efficiency (SE), i.e., max{R(φ) / ∆f(φ)}, and break ties with the one with maximum capacity, i.e. max{R(φ)}.

2) MSE-MinS (Most Spectral Efficient – Minimum Spectrum): This method selects the OCh format/path with highest SE, i.e., max{R(φ) / ∆f(φ)}, and break ties with the one with minimum spectrum usage, i.e. min{∆f(φ)}.

3) JEC (Just Enough Capacity): Let r(d) denote the data rate of d and 𝑟𝑟(d) denote the cumulative data rate of all traffic demands between this node pair. This method selects the OCh format/path with best capacity fit, i.e. min{R(φ): R(φ) ≥ r(d) + 𝑟𝑟(d)} and break ties with the one with minimum spectrum usage, i.e. min{∆f(φ)}.

The possible algorithm to follow for solving the problem is described hereafter:

Input: Demand, d; candidate routing paths, Π; MCh format list, Φ; existing MChs over path π ∈ Π, M(π).

1. If MCh m ∈ M(π) has idle capacity to support d, route d over m. End.

2. For each OCh format φ ∈ Φ and path π ∈ Π, create auxiliary graph G’ with nodes representing physical nodes and links representing feasible OChs (in terms of performance, spectrum continuity and add/drop port availability). Set link cost according to the number of idle interfaces and cards at the edge nodes of the link. Compute lowest cost solutions over the graph (i.e. with minimum number of new line interfaces).

3. Let R(φ) and ∆f(φ) denote the data rate and frequency slot width of OCh format φ. Shortlist OCh formats/paths that minimize the number of new line interfaces/cards and select the optimum solution according to one of the before described strategies:

Output: Routing path π*; created OCh with format φ*, provide the overall spectral occupation and number of deployed OEO devices.

An ILP design could be used for bench-marching the solutions provided by the previously described RSGA strategies. Please consider that the cost function will be driven by the number of deployed OEO interfaces.

It is also possible to obtain the comparison results with the use of GA.

Results obtained with the different strategies (native Net2Plan, ILA and Net2Plan with GA) could be then compared in terms of output quality (number of OEO resources and occupied spectrum) and performance (computation time).

Network and traffic assumptions

The network comparison shall be estimated for backbone and national wide networks. The students can select any network available in Net2Plan inputs, but the network characteristics have to satisfy the features described in Section 2.a of this document.

About traffic assumptions, the traffic shall be generated as described in Section 2.b of this document.

About the modulation level, students must consider values for long reach OEO devices presented in Section 2.c.

Project 2: Compare different regeneration strategies in elastic optical networks (2/3 people)

Goal: In elastic optical networks, evaluate whether the regeneration with a change of the modulation format/symbol rate between the input and output optical channel allows some gains in terms of spectral resources, overall OEO devices and service blocking.

In backbone optical networks routing, grooming, modulation level and spectrum assignment (RGMLSA) algorithm is used to determine the resources required to accommodate each demand. A key part of the decision process is the Optical Channel (OCh) format do deploy, balancing factors such as reach, capacity and spectral occupation.

An Optical Channel is the connection carrying data to be transported between the source and destination nodes along a given path and with a specified spectral occupancy.

Elastic optoelectronic devices can adapt their modulation format/symbol rate as a function of the transported capacity requirements: capacity to be transported and distance to cover. If this distance is longer than the optical reach of the selected modulation level, a regeneration (3R) has to be placed along the path. Legacy routing algorithms after regenerating keep the same modulation level on the OCh. A possible advantage of Elastic devices relies on the capability of tuning the modulation format / symbol rate to the distance to be covered, in this way it would be possible to change the modulation format / symbol rate of the two tunnels composing the OCh.

The figure hereafter given an example of the proposed concept:

Moreover, because of grooming capability of some cards, it is possible to carry a signal on a single carrier or two carriers. So for very long paths, a 200 Gb/s channel could be transported by 2 different carriers at 100 Gb/s and after regeneration be transported on a single carrier.

The grooming capability of the OEO elastic device depends on the number of client cards and line cards it has. The figure below gives an idea of this type of cards:

The students are invited to study the advantages associated to the possibility or not of changing such pair so that in a network with different traffic load and compare:

1. The total number of Client / Line interfaces

2. The total number of blocked demands 3. The free spectrum and network fragmentation. About the network fragmentation the adopted

metric will be the External fragmentation. The external fragmentation on a link is computed as

follows: EF= 1 − 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 𝐵𝐵𝑜𝑜 𝐹𝐹𝐿𝐿𝐿𝐿𝐿𝐿 𝐵𝐵𝐵𝐵𝑐𝑐𝐿𝐿𝐿𝐿𝐵𝐵𝑐𝑐𝐿𝐿𝑐𝑐𝑐𝑐𝐿𝐿 𝐿𝐿𝐵𝐵𝐵𝐵𝐿𝐿𝐿𝐿Total number of free slots

This study could be done by adopting the auxiliary graph and provide a modification of that graph while doing the routing (students can ask to me about such graph modification) or with other methods available in Net2Plan.

An ILP design could be used for bench marching the solutions provided by the previously described RSGA strategies. Please consider that the cost function will be driven by the number of deployed OEO interfaces.

It is also possible to obtain the comparison results with the use of GA.

Results obtained with the different strategies (native Net2Plan, ILA and Net2Plan with GA) could be then compared in terms of output quality (number of results) and performance (computation time).

Network and traffic assumptions

The network comparison shall be estimated for backbone and national wide networks. The students can select any network available in Net2Plan inputs, but the network characteristics have to satisfy the features described in Section 2.a of this document.

About traffic assumptions, the traffic shall be generated as described in Section 2.b of this document

About the modulation level students must consider values for long reach OEO devices presented in Section 2.c.

Project 3: Restoration for just-enough capacity (1/2 people)

Goal: In an elastic network, it is possible to exploits the rate adaptation of optoelectronic devices when restoration occurs and drop from the optical channel that fails services that do not require recovery. A comparison between restoration with and without traffic drop shall be done, the evaluation shall be done by considering the overall amount of blocked traffic (nominal / restoration) and regenerators.

In backbone optical networks, client traffic demands are set up along several periods and, in each period, a routing, grooming, modulation level and spectrum assignment (RGMLSA) algorithm is used to determine the resources required to accommodate each demand.

The traffic transported in backbone networks is composed of different types of services, some with high quality and other with low quality of service. To enhance the efficiency of the transported capacity (i.e., amount of carried traffic per GHz), in an OCh different QoS services may be groomed together. The OCh results with a QoS associated to the more constrained service (i.e., the one with the highest QoS). This means that in case of network failure, the whole OCh is recovered and low QoS services will be recovered even if they have not paid for (this behavior is called recovery upgrade of service).

When recovery paths are computed, it results longer than the nominal path (the path associated to the OCh when no failure is present in the network), so it is likely the need for a regeneration along the recovery path. In case of restoration, the recovery path has to be computed by considering un-failed links. For this project, only one link fails per Failure Scenario.

Elastic optoelectronic devices can adapt their modulation format/symbol rate as a function of the transported capacity requirements: capacity to be transported and distance to cover. So in elastic networks it could be possible to adjust the capacity of the OCh to recover so as to only carry the services that only payed for recovery and prune others. In this way, by reducing the capacity of the OCh, modulation formats / symbol rates with longer reaches and / or narrower bandwidth could be adopted. This strategy would firstly reduce the amount of required 3Rs and enhance the amount of free spectrum in the network.

The students are invited to evaluate the possible advantages of dropping low QoS services with respect to legacy methods (that do not drop services) in the context of elastic optical networks. Comparison of the network performance under these two assumptions has to be done with different traffic loads and in terms of:

a) The total number of OEOs b) The total number of blocked demands and unfeasible restoration paths.

c) The free spectrum and network fragmentation. About the network fragmentation the adopted metric will be the External fragmentation. The external fragmentation on a link is computed as

follows: EF= 1 − 𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 𝐵𝐵𝑜𝑜 𝐹𝐹𝐿𝐿𝐿𝐿𝐿𝐿 𝐵𝐵𝐵𝐵𝑐𝑐𝐿𝐿𝐿𝐿𝐵𝐵𝑐𝑐𝐿𝐿𝑐𝑐𝑐𝑐𝐿𝐿 𝐿𝐿𝐵𝐵𝐵𝐵𝐿𝐿𝐿𝐿Total number of free slots

For this study, when OCh are created, demands could be aggregated and routed end-to-end (that is, only services having the same source and destination nodes are aggregated, there is no intermediate grooming).

Network and traffic assumptions

The network comparison shall be estimated for backbone and national wide networks. The students can select any network available in Net2Plan inputs, but the network characteristics shall satisfy the features described in Section 2.a of this document.

About traffic assumptions, the traffic must be generated as described in Section 2.b of this document

About the modulation level students have to take into account values for long reach OEO devices presented in Section 2.c.

Project 4: Aging of the network devices (2 people)

Goal: The network infrastructure degrades with time (aging effects), so some optical channels risk to become unfeasible. When a network operator plans a network, the time period for deploying regenerators is important because their cost decreases along the network lifecycle. This project will compare two different methods for computing the resources required by the network due to aging and define which is the more cost-effective.

The cost of an optical network depends on the overall traffic to be routed and the physical impairments acting on the signal. The network infrastructure degrades with time (aging effects), so some optical channels risk to become unfeasible. For a network operator it is important to understand when and where to deploy regenerators for minimizing their overall number and at the same time the total cost of the network. If regenerators are added in an already deployed optical channel, the traffic that is carried on is interrupted until the optical channel maintenance terminates. When a network operator plans a network, the time period for deploying regenerators is important because their cost decreases along the network lifecycle.

So, it is important to do a cost analysis for estimating the best strategy of cost savings as a function of the traffic outage.

For estimating the impact of aging on optical network and defining the number of required regenerators, two possible approaches are possible:

1. Consider the impairments acting at each period and for each period

a) Check if already routed services (routed at N-1, with N the actual time period) are feasible, otherwise place regenerators where needed

b) Route the new services associated to the time period N.

c) The cost of deployed OEO devices (transponders and regenerators) follows the cost decrease provided in the table presented in Section 2.d).

2. Consider impairments at the end of life of the network, so that all periods have the same impairment estimation

a) Route new services (no check on the feasibility of the already set-up routes is required because the physical status is unchanged) and apply the cost of OEO devices required for routing these services with cost provided in the table presented in Section 2.d).

The cost comparison between these two approaches must be done by considering the overall amount of required OEO resources and the device cost at each time periods. The total number of time periods is N = 4.

Network and traffic assumptions

The network comparison shall be estimated for backbone and national wide networks. The students can select any network available in Net2Plan inputs, but the network characteristics have to satisfy the features described in Section 2.a of this document.

About traffic assumptions, the traffic shall be generated as described in Section 2.b of this document

About the modulation level students must take into account values for long reach OEO devices presented in Section 2.c.

Project 5: Fast heuristics for minimizing the regenerators to use for restoration purposes (2/3 people)

Goal: By the means of simulations, compare different 3Rs placement algorithms (heuristics based) to reduce the overall amount of 3R devices when restoration is calculated.

An optical network is affected by failures and some services could be impacted by the disruption of one of several devices / links / nodes. When a network is planned, a service level agreement is specified to guarantee its robustness with respect to the failures that could occur during the network lifecycle.

The performance of a planning tool could be estimated by considering the overall cost / number of resources necessary to set-up a given amount of traffic. Some services could require restoration (shared recovery paths, with restoration resource sharing, e.g., wavelengths and optical regenerators) and the minimization of restoration resources like optical regenerators is very important.

A service is routed over a main (nominal) path, to which can be associated a secondary path. This path could be:

• Calculated independently of the failures and dedicated to the nominal one (protection)

• Calculated when the failure arises and be shared with other nominal paths that does not fail at the same time (restoration)

To each service it is specified the number of failures to be robust. For this project, only restoration must be studied and the algorithm to propose has to take into account only N link failures (N ≥ 1) at time. A Failure Scenario (FS) is the occurrence of N failures at the same time in the network. All combinations of N failures must be taken into account when a planning has to be done, so the number of Failure Scenario to emulate are 𝑐𝑐!

(𝑐𝑐−𝐵𝐵)!𝐵𝐵! .

For each FS the impacted routes have to be rerouted and optical devices allowing the correct reception have to be reused or deployed. The use of a heuristic method that only considers the number of optoelectronic resources required for optical channels impacted by a single FS without considering further FS, provides local optimum resource placement.

The goal of this project is to provide a method that allows the minimization of regenerators used for restoration purposes. Restoration has to be computed for the source based strategy (SBR), that consists in compute the restoration path by taking into account the resources (links for this project) still available in the network; please, note that a restoration path may partially use resources belonging to the nominal path if they are not failed.

The results of the proposed method must be compared with the one of a method where no minimization strategy guides the placement of regenerators. Results to compare are: number of regenerators, restoration paths that can be calculated and unfound restoration paths.

To have a local optimum result means that the overall number of regeneration devices deployed with this method could be higher than the number of devices that could be placed if the knowledge of all the devices to be set-up is known in advance.

The use of ILP, on the other side, could provide a global optimal solution.

No Aggregation or grooming is required for this study.

Network and traffic assumptions

The network comparison shall be estimated for backbone and national wide networks. The students can select any network available in Net2Plan inputs, but the network characteristics have to satisfy the features described in Section 2.a of this document.

About traffic assumptions, the traffic shall be generated as described in Section 2.b of this document

About the modulation level students shall take into account values for long reach OEO devices presented in Section 2.c.

The number of simultaneous failures occurring at each failure scenarios must be up to N=2 (i.e., the students have to study N=1 initially and then N=2).

Project 6: Upgrade planning considering aging of the network devices and 1-link failure restoration of traffic (5 people)

Goal: The network infrastructure degrades with time (aging effects), so some optical channels risk to become unfeasible. When a network operator plans a network, the time period for deploying regenerators is important because their cost decreases along the network lifecycle. The services transported in the network have different quality of service, defining the resiliency of the services with respect to device failures that can occur during the network life. This project will compare two different methods for computing the resources required by the network due to aging and guaranteeing the robustness to 1 failure per time. The student shall define which is the more cost-effective method for the studied scenarios.

About the physical degradation of the network due to aging effects, the students will refer to the description of the problem provided for Project 4.

An Optical Channel is the connection carrying data to be transported between the source and destination nodes along a given path and with a specified spectral occupancy.

Elastic optoelectronic devices can adapt their modulation format/symbol rate as a function of the transported capacity requirements: capacity to be transported and distance to cover. So in elastic networks it could be possible to adjust the capacity of the OCh to recover so as to only carry the services that only payed for recovery and prune others. In this way, by reducing the capacity of the OCh, modulation formats / symbol rates with longer reaches and / or narrower bandwidth could be adopted.

During each time period, the traffic matrix changes, and part of its services increases their capacity and other services are drop. Usually when a service increases its capacity, one possible solution could be either to fill the free capacity in already deployed Optical Channels (OCh), or to upgrade their capacity if elastic optical devices are used, or to add a further OCh.

If elastic devices are available, the increase of the OCh capacity may have an impact on the number of regenerators associated to it as the resulting optical reach associated to the new modulation level could be lower than the distance the OCh shall reach. In the latter case, if there is no free room in already deployed OCh, new ones must be set-up. When a new set-up of OCh is done, it is important to check which are the services that could be aggregated together, and the modulation level associated to such OCh has to follow the MSE-MaxC method. The MSE-MaxC (Most Spectral Efficient – Maximum Capacity) selects the OCh format/path with highest spectral efficiency (SE), i.e., max{R(φ) / ∆f(φ)}, and break ties with the one with maximum capacity, i.e. max{R(φ)}.

(If more students are interested to this study, other methods for the Modulation level selection could be investigated, e.g., as in Project 1).

About the traffic restoration, it is assumed the network shall be robust to only 1 FS (i.e., one link failing per failure scenario).

The network life-cycle is divided in N periods, with n = [0; N]. For estimating the impact of aging on optical network, two possible approaches are possible:

1. Consider the impairments acting at each period and for each period

a) Check if already routed services (routed at N-1, with N the actual time period) are feasible, otherwise place regenerators where needed

b) Route the new services associated to the time period N.

c) The cost of deployed OEO devices (transponders and regenerators) follows the cost decrease provided in table 2.d).

d) Regenerators used for restoration purposes in phase n-1 could be reused for nominal path in phase n, otherwise new regenerators must be deployed in the network. After the validation of nominal resources, restoration of services active at period n shall be calculated.

2. Consider impairments at the end of life of the network, so that all periods have the same impairment estimation

a) Route new services (no check on the feasibility of the already set-up routes is required because the physical status is unchanged)

b) Compute the restoration paths of all the OCh present in the period n

c) Apply the cost of OEO devices required for routing these services with cost provided in table presented in Section 2.d).

Please, when the cost of devices is considered, consider the cost associated to the period n at which the device has been deployed.

Network and traffic assumptions

The network comparison shall be estimated for backbone and national wide networks. The students can select any network available in Net2Plan inputs, but the network characteristics shall satisfy the features described in Section 2.a of this document.

About traffic assumptions, the traffic must be generated as described in Section 2.b of this document

About the modulation level students have to take into account values for long reach OEO devices presented in Section 2.c.

Project 7: Does the concept of opaque network is coming back? (1/3 people)

Goal: The introduction of new coherent optics enabled the router client ports to support 400 Gb/s on short distances. With these interfaces optical signals could be directly transported at Layer 0 without the use of long reach transponders (or muxponders) which collect several client interfaces to be transported in a unique wavelength at Layer 0. The introduction of these high capacity/low reach devices could change the architecture of the network and directly connect router interface to the optical layer, without the use of long reach transponders.

Traffic growth shows no signs of abating, yet the trends that have enabled service providers to accommodate that growth – increasing transmission capacity combined with smaller, lower cost, more powerful silicon – are becoming ever-harder to solve by scaling capacity alone. As ever-growing network traffic sees the use of coherent optics expand into metro and access networks, network operators need solutions that efficiently support bandwidth growth across an ever-wider set of network applications.

With the introduction of the latest generation of coherent solutions router client ports can support 400 Gb/s as long distance transponders, but across a limited optical reach.

With such a capability, the network operator could rethink about its network architecture and decide to come back to opaque network solutions, where the switch of services is done directly through the IP routers, no optical by-pass is performed and there is no need for long reach transponders.

Now you can cost-effectively upgrade long-haul DCI links and national-scale packet networks to 400GE with greater spectral efficiency than prior generation solutions, without having to bond multiple, lower speed wavelengths to transport 400GE across long-distance links.

With this state of the art, an optical vendor/operator must investigate the following points:

• Does the optical by-pass is still interesting? Does the systematic regeneration in intermediate IP router is more cost effective than operating grooming and optical by-pass?

• Does the optical by-pass with optical devices having a limited reach, directly connected to the router could be a new solution for next generation of optical networks?

Figure (a) shows the complete opaque solution, where no optical by-pass is allowed. Figure (b) shows a case where 400G ZR+ interfaces are directly connected to the router, optical bypass but no regeneration is possible, the signal has to be regenerated by passing through the router. Figure (c) presents a legacy

network configuration, with transponders (here called muxponders) are connected to the router through gray optics (client cards); several client cards at 100G may be connected to a transponder. OEO regeneration is realized by a couple of transponders in a back-to-back configuration.

Network and traffic assumptions

The network comparison shall be estimated for backbone and national wide networks. The students can select any network available in Net2Plan inputs, but the network characteristics have to satisfy the features described in Section 2.a) of this document.

About traffic assumptions, the traffic shall be generated as described in Section 2.b) of this document.

About the modulation format, symbol rate and optical reach, students must take into account values presented in Section 2.c).

About the cost for comparing long reach and ZR+ OEO devices, please refer to values presented in Section 2.d)

2. Simulation assumptions

a) Network types Students can use any network that is available in Net2Plan. The networks could be divided in Continental and National wide, and must satisfy the following features:

Continental National Number of links ≥ 20 links ≤ 20 links Average link length ~ 300 km ~ 150 km Max link length ~ 1000 km ~ 400 km

b) Traffic generation

Traffic will be generated by drawing random matrices with all services at 100Gb/s.

The simulations shall be done for different traffic load. The minimum load will contain 350 services, then increase the traffic load with a step of 50 services. The maximum load will be the one for which 1% of nominal services is blocked.

c) OEO modulation levels

The OEO devices (transponders and regenerators) can adopt one of the following modulation formats with associated the distance reach indicated in the table.

The reaches provided in the following tables are associated to a system based on SSMF fibers amplified with EDFAs.

Modulation levels for long reach OEO devices

Data Rate (Gb/s) Modulation format

Bits/symbol (Gb/s) - Entropy

Channel spacing ∆f (GHz)

Reach (km)

800 PCS 64 QAM 5.67 100 150 700 PCS 64 QAM 5.00 100 400 600 16 QAM 4.00 100 700 500 PCS 16 QAM 3.60 100 1300 400 PCS 16 QAM 3.00 100 2500 300 PCS 16 QAM 2.39 100 4700 300 64 QAM 6.00 50 100 200 16 QAM 4.00 50 900 100 QPSK 2.00 50 3000

Modulation levels for ZR+ OEO devices

Data Rate (Gb/s)

Modulation format

Bits/symbol (Gb/s)

Channel spacing ∆f (GHz)

Reach (km)

400 16 QAM 4 75 600 300 8 QAM 3 75 1800 200 QPSK 2 75 3000 100 QPSK 2 50 3000

d) Cost of devices

The following table considers the cost of OEO transponders. Please consider that 3R regenerators are realized with transponders in a back-to-back configuration, so their cost twice a transponder.

Time Period T0 T1 T2 T3 Cost of long reach OEO 1 0.85 0.8 0.77 Cost of ZR+ 0.5 0.43 0.4 0.38 Cost of client cards plugged on long reach OEO (@100G)

0.07 0.057 0.05 0.045

Cost for OCh traffic interruption: 1 transponder cost at T0, whatever the considered