Internet Research Task Force D. Siracusa Internet Draft A. Francescon Intended status: Informational E. Salvadori Expires: May 2014 CREATE-NET R.J. Duran I. de Miguel R.M. Lorenzo Universidad de Valladolid November 4, 2013 Framework for Cognitive Capable Optical Networks draft-siracusa-nmrg-ccon-fwk-00.txt Status of this Memo This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html This Internet-Draft will expire on May 4, 2014. Copyright Notice Copyright (c) 2013 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents Siracusa, et al. Expires May 4, 2014 [Page 1] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License. Abstract The increased complexity in the management of highly heterogeneous optical networks is recently forcing vendors and providers to look for novel mechanisms which diminish the manual intervention by favoring the autonomous execution of several operational tasks, especially when dealing with network congestion or failure events. The adoption of cognitive techniques in networking envisions a network which is able to adapt itself to current or forecasted conditions by taking into account previous history, and which is able to act proactively, rather than reactively, in order to avoid problems before they arise. In this document, a novel architectural framework that introduces cognitive techniques in the optical networking domain is described, and several use cases provided to emphasize its effectiveness. Table of Contents 1. Introduction ................................................. 3 2. Background ................................................... 5 2.1. Software-defined adaptable elements ..................... 5 2.2. Monitoring elements...................................... 6 2.3. Control system running cognitive processes .............. 8 3. Framework .................................................... 9 3.1. The Cognitive Decision System .......................... 11 3.2. Processes and knowledge bases .......................... 13 4. CCON Use Cases .............................................. 14 4.1. Quality of Transmission assessment ..................... 14 4.2. Path Computation ....................................... 15 4.3. Virtual Topology Design and Reconfiguration ............ 16 5. Implications on the Control Plane ........................... 16 5.1. Disseminate network configuration information .......... 16 5.2. Feed the cognitive processes with network data and statistics .................................................. 17 5.3. Implement the decisions of the cognitive processes on the device ...................................................... 17 6. Contributing Authors ........................................ 18 7. Security Considerations ..................................... 19 8. IANA Considerations ......................................... 19 Siracusa, et al. Expires May 4, 2014 [Page 2] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 9. References .................................................. 20 9.1. Informative References ................................. 20 10. Acknowledgments ............................................ 22 1. Introduction Optical networks are facing increased levels of heterogeneity, from types of services to transmission technologies. Hence, a key issue of highly heterogeneous networks is how to efficiently control and manage network resources while fulfilling user demands and complying with quality of service requirements. A solution for such a scenario may come from cognitive networks, also known as learning-capable communication networks [Tav2011]. +----------------+ +-------------+ | Orient | | End-to-end | | (Plan) |<---| goals |----+ | | | | | +----------------+ +-------------+ | ^ ^ | | | | | v | | | +-----------+ | | | | | | | +------------------>| Decide | | | | | | +---------+ +-----------+ | | ^ | | v | | +-------------+ +-------+ | | | | | |<---------+ | | Observe |--->| Learn | | | | | |<---------+ | +-------------+ +-------+ | | ^ | | | | | | | v +-------------+ +---------+ | | | | | Environment |<-------------------| Act | | | | | +-------------+ +---------+ Figure 1 The cognitive loop Siracusa, et al. Expires May 4, 2014 [Page 3] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 A cognitive network is defined as "A network with a process that can perceive current network conditions, and then plan, decide, and act on those conditions. The network can learn from these adaptations and use them to make future decisions, all while taking into account end-to-end goals." [Tho2006], that is, the network implements the so-called cognitive loop (Figure 1). Hence, there are three main ingredients in such a network: o Monitoring elements, which provide the network with the perception of the current condition including physical layer status, power consumption, traffic patterns, useful to enable an aware network. o Software-defined adaptable elements, which provide the network with the capacity of modifying its current configuration, thus enabling an adaptive network. o Cognitive processes, which learn or make use of past history, so that even when facing two equivalent scenarios, the network (or the entity containing those cognitive processes) may act in a different way if its previous history is different. Therefore, a cognitive network is a network which is able to adapt itself to current or forecasted conditions by taking into account previous history, and act proactively, rather than reactively, in order to avoid problems before they arise. Moreover, those tasks should be performed autonomously, with little or no intervention of the network operator. Cognitive networks are thus closely related with autonomic networks [Beh2013]. An autonomic network relies on self-configuration, self-healing, self-optimization and self- protection functionalities, so that it may make decisions without manual intervention or external help (e.g., human administrator) [Beh2013][Mov2012]. In this way, an autonomic network is not only aware and adaptive, but also automatic. Therefore, a cognitive network can be considered as a variant of an autonomic network [Mov2012], but it emphasizes the self-optimization functionality as well as the use of learning mechanisms, in contrast with other types of autonomic networks, which generally rely on policy-based methods rather than on learning techniques to support the adaptations [Mov2012, table VIII]. This document partially leverages on [Tav2011], an informative document which describes the opportunities and challenges for a technology-independent Learning Capable Communication Network (LCCN). Of course, given the focus on transport networks, the document will apply the concepts envisioned in that document to the specific optical networking domain. Siracusa, et al. Expires May 4, 2014 [Page 4] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 The structure of the document is the following. In Section 2 a review of the recent enhancements of the optical and control technologies that are enabling Cognitive Capable Optical Networks (CCON) is provided. Section 3 describes the framework and the related building blocks of a CCON. Section 4 focuses on a set of applications of cognition proposed in optical networks while Section 5 describes the implications on the Control Plane. 2. Background A novel paradigm, like the one proposed in this document, has emerged thanks to the recent availability of novel optical and control system technologies. The scope of this section is to review these enabling technologies. 2.1. Software-defined adaptable elements Software-defined adaptable elements are essential for the realization of the cognition concept in optical networks since they allow the optimum and on-demand use of resources, according to the intelligent (i.e., cognitive) processing of connection demands. Although a cognitive network could rely on a set of fixed transceivers in the nodes, the higher degree of flexibility provided by software-defined transmitter and receiver subsystems is turning them into key network elements to perform the adaptable allocation of traffic demands. In practice, the transmitted bandwidth adaptability in optical transceivers is realized by: a) altering the modulation level or format (i.e., the bits per symbol) per optical carrier and b) varying the number of electronic or optical carriers in multi- carrier formats [Ger2012]. The general purpose of these adaptable schemes is to apply the optimum format over the minimum number of carriers, thus maximizing the spectral efficiency (i.e., the number of bits/s/Hz) for a certain traffic demand over an optical path with certain end-to-end performance requirements. Format adaptability can be performed either in the optical domain, by simply enabling or disabling the different arms of nested Mach Zehnder Modulator (MZM) structures at the transmitter and the related output port of 90 deg hybrid at the receiver, or directly in the electronic domain by appropriately defining the signal levels of the modulation signals. Moreover, for multi-carrier schemes based on electronic generation of subcarriers, the subcarrier number is defined in the electronic domain by the length of the digital signal processing function prior to the optical modulation, while for optically generated subcarriers, their number is defined either by Siracusa, et al. Expires May 4, 2014 [Page 5] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 filtering the appropriate number of carriers or by gating the appropriate number of subcarrier transmitter outputs directly in the optical domain. The bandwidth adaptable data transmission schemes mentioned above can realize the optimum use of network resources according to the traffic demands, but they result in added complexity in terms of control. This is attributed to the fact that any decision mechanism must account for a large number of possible combinations (i.e., central wavelength allocation, format and number of subcarriers) to optimally serve a demand for a given optical path. The role of cognitive optical networking is particularly beneficial for the practical implementation of these schemes, since it can significantly relax the decision mechanism, by exploiting past history. It is noted that cognition can apply in combination with any adaptable (flexible) transmission technique, since all of them are intrinsically software-defined schemes. 2.2. Monitoring elements Both traffic and optical performance monitoring techniques are required to know the current state of the network. That information can be used not only for making immediate decisions but also as an input for forecasting procedures facilitating the execution of proactive actions. While existing techniques for traffic monitoring - leveraging on protocols like SNMP [RFC1157], RMON [RFC4502] - can be also exploited in cognitive optical networks, the introduction of new optical transmission systems, and their coexistence, triggers the need of the development of novel Optical Performance Monitoring (OPM) techniques. Thus, to guarantee that the Quality of Service (QoS) and resiliency are achieved along the lightpaths, monitoring of the physical properties of the signal is required. OPM analyzes the accumulation of the so called "non-catastrophic" transmission impairments such as Chromatic Dispersion (CD), Polarization Mode Dispersion (PMD) and non-linear effects [Cha2010]. These effects, combined with the accumulation of network physical impairments, like crosstalk, Amplified Spontaneous Emission (ASE) noise, Polarization-Dependent Loss (PDL) and filter/ROADM (Reconfigurable Optical Add and Drop Multiplexer) concatenation make the information data unrecoverable even though the received optical signal power is at an acceptable level. Furthermore, the so called "catastrophic" impairments such as accidental fiber cuts and damaged or improperly installed network elements can cause critical network performance degradations. Meanwhile, other channels co-propagating in the same link can be affected as well due to transients in the amplifiers caused by the Siracusa, et al. Expires May 4, 2014 [Page 6] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 rapid change of the total optical power (several dBs). Despite the nature of the failure, it becomes clear that an accurate and fast parameter monitoring would allow an early fault analysis with a fast switching to a protection path. The efficiency and reactivity to different problematic events also depends on the critical interaction between OPM and higher-level control and management plane systems. Therefore, monitoring devices must be placed in strategic places during the planning stage of an optical network. In 10 Gb/s and 40 Gb/s optical networks, various OPM techniques have been developed relying on external devices such as Optical Spectrum Analyzers (OSAs), RF devices and frequency-selective polarimeters. On the other hand, modern transmission technologies for 100 Gb/s, 400 Gb/s, 1 Tb/s and beyond are based on coherent technologies by taking advantages of powerful and cost-effective Digital Signal- Processing (DSP) capabilities. OPM techniques based on DSP, where expensive external devices are not required, are adaptable to varying data rates and modulation formats, and are capable of realizing joint monitoring of key physical layer parameters like CD, PMD, PDL, OSNR, Bit-Error-Rate (BER), etc. The DSP has already been integrated in the receiver side, so it will provide network information at the end points. Furthermore, in the future, DSP could also be integrated in optical amplifiers or ROADMs, thus allowing the derivation of relevant information at these mid-points. In DSP-based OPM techniques, Frequency-Domain (FD) equalization combined with Data-Aided (DA) channel estimation can be considered as a promising technology. Compared with Non-Data-Aided (NDA) methods based on gradient algorithms for Time-Domain (TD) filters, which are strongly dependent on the modulation format and suffer from a relatively slow convergence with potential sub-optimum acquisition and even failures, the DA channel estimation, based on a periodically transmitted Training Sequences (TS), allows instantaneous filter acquisition, immediate OPM, and the modulation format can be altered arbitrarily in between the fixed training patterns. All these benefits come at the cost of slight bandwidth efficiency degradation due to the insertion of TS, and the required overhead can be below 5%. Moreover, in a coherent burst-mode receiver, each burst must be instantaneously equalized and only DA channel estimation is suitable. These DSP-based OPM techniques can be implemented in hardware, and therefore real time physical impairment information will be available for the control plane. However, if off-line DSP processing is used instead, then the control plane database can be periodically updated by the OPM with the physical impairment information, and thus the control plane does not need to wait for the DSP processing. Siracusa, et al. Expires May 4, 2014 [Page 7] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 2.3. Control system running cognitive processes In a cognition-capable optical network architecture, the coordination among the "brain" that makes decisions and establishes network operations and the data plane (photonic layer) is provided by a control system, which implements the mechanisms supporting the cognitive intelligence in an automated and reconfigurable manner. Two different approaches can be envisioned to implement a cognitive architecture: (i) centralized, in which the network and all components are under the control of a single cognitive entity, which receives all the information related to network configuration, availability, monitored parameters, etc.; and (ii) distributed, in which there is not a specific node with a prominent role, and where the cognition is distributed among all the network nodes (or a large part of them), which exchange the information mentioned above. Both the centralized and the distributed cognitive architectures need for a system delivering updates related to network status, reserving the resources, and configuring the optical devices. These tasks are carried out by the Control Plane (CP). A cognitive optical network is expected to make effective decisions by leveraging on a knowledge base, built with the support of the CP. Decisions are made for different activities, such as lightpath activation in response to a user request or re-arrangement of active network connections. In such a context, and in particular for the latter activity, knowledge about the status of currently active lightpaths is required. While it is evident that this information can be disseminated by adapting already existing protocols, it is also clear that it would demand for the exchange of a non-negligible amount of data between distributed control nodes (including the path of each active connection, physical layer impairments, etc.). Hence, from the operational point of view, a solution with distributed control entities may not be cost-effective. Furthermore, the lack of a global view about the network status in a distributed architecture may lead to conflicting decisions. Finally, cognitive decisions also rely on the values collected by the monitoring system of the network. In this case, a distributed solution is hard to be kept updated, since the information collected by the monitors flows through the network and is hard be processed reliably at a single instance (i.e. as in the case of a centralized approach). On the other side, a centralized approach may suffer of scalability issues, and the cognitive entity is potentially a single point of failure of the network. While the latter issue may be lessened by enhancing the protection/robustness of the cognitive entity and by introducing backup entities, the former is a matter of network scenarios. In the context of optical networks and with a limited amount of managed Siracusa, et al. Expires May 4, 2014 [Page 8] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 nodes, a centralized approach could still scale sufficiently, while ensuring a high level of reliability and providing more effective path computations. This is true for the case of core networks and in particular national backbone networks that require the employment of bandwidth flexible resource allocation mechanisms over a well- defined and limited number of nodes. 3. Framework Several architectures leveraging on cognitive mechanisms have been recently proposed in literature to determine how (and where) the three aforementioned key ingredients - software adaptable elements, monitoring elements, and control systems - are implemented, how they are glued together, as well as which tasks are going to be solved with the help of cognition. Generic cognitive architectures have been proposed in [Tho2006], [Kli2010] and [Tav2011], while cognitive architecture for optical networks have been proposed in [Zer2010], [Wei2012] and [deM2013]. These architectures show that cognition can be implemented in different dimensions, in terms of devices and protocol layers. For instance, in a cognitive network implementation, software-defined transceivers may include monitoring functionalities together with internal intelligence to modify their configuration autonomously, i.e., being truly cognitive transceivers. However, another implementation may opt for shifting the intelligence in charge of configuring those transceivers to the upper layers of the nodes, where the transceivers are located, thus being the network nodes the cognitive elements rather than the transceivers themselves. That example may find its way in a network with distributed cognition, where all network nodes are equipped with cognitive capabilities and collaborate in sharing the acquired knowledge. Nevertheless, another possibility is a network with centralized cognition, where a single node (the control node) contains the intelligence and makes decisions which are then communicated to the remaining network nodes by means of control plane protocols with suitable extensions. On the other hand, the level and type of cognition to be added to a network depends not only on the adopted approach but also on the capabilities of network monitors and software-adaptable elements employed; the higher the flexibility of the available elements, the higher the potential of cognition. However, although the utilization of software-defined transceivers and flexible networks, as well as software-defined networking techniques [Das2012], is usually associated with cognitive optical networks, it should be noted that these technologies are not strictly necessary for adopting a cognitive networking approach. Siracusa, et al. Expires May 4, 2014 [Page 9] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 +-+ +------------------------------------------------------------+ | |<=>| User interface module | | | +------------------------------------------------------------+ | | ^ | | +--------------------|-----------------------------------+ | | | | Knowledge bases | | | +----------------------|---------------------------------+ | | | | v Learning modules | | | | +--------------------------------------------------------+ | | |C| | Cognitive Decision System (Cognitive processes) | | | |o| | +----------------------------+ | | | |n| | | | | | | |t| | v v | | | |r| |+---------+ +----------+ +--------+ +-----------+ | | | |o| || Traffic | | Virtual | | RWA/ | | QoT | | | | |l| ||Grooming |<->| Topology |<->| RMLSA |<->| Estimator | | | | | |<=>|| Module | | Design | | Module | | Module | | | | |P| || | | Module | | | | | | | | |l| |+---------+ +----------+ +--------+ +-----------+ | | | |a| | ^ ^ ^ ^ | | | |n| | | | | | | | | |e| | v v v v | |-+ | | |+--------------------------------------------------+ | | |p| || Network Planner & Decision Maker Module | |-+ |r| |+--------------------------------------------------+ | |o| +--------------------------------------------------------+ |t| ^ ^ |o| | | |c| v | |o| +-----------------------------------+ | |l| | Optical signal monitoring & | | |s| | transponder interface | v | | |+-----------++--------++----------+| +------------------+ | | || Software ||Optical || Digital || |Traffic monitoring| | |<=>|| Defined || Signal || Signal || | interface | | | ||Transponder||Monitors||Processing|| | +--------------+ | | | |+-----------++--------++----------+| | | Traffic | | | | | Optical network (PHY interface) | | | Monitoring | | | | +-----------------------------------+ | | System | | | | | +--------------+ | | |<=======================================>| NMS | +-+ +------------------+ Figure 2 CCON schematic architecture The proposed framework focuses on a key building block called Cognitive Decision System (CDS). The CDS determines how to handle Siracusa, et al. Expires May 4, 2014 [Page 10] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 traffic demands and network events, and optimizes network usage and performance by taking into account both the current status of the network and past history. The CDS also instructs the control plane to configure network elements accordingly. Cognition can be implemented in a either centralized or distributed ways, depending on whether the CDS is a single instance running on a single control node in the whole network or it is implemented in different network nodes. Figure 2 shows the major building blocks of a centralized architecture, where the CDS is based on a single running instance. The CDS is assisted by a Control Plane (CP), which feeds the CDS with updates regarding network status and resource availability, grants the delivery of the decisions made by the CDS to all the interested nodes, and watches over the device configuration process, notifying any malfunctioning or anomaly. The proposed architecture also includes software-defined adaptable elements, which implement the decisions made by the CDS (that are in turn communicated by the CP) and a Network Monitoring System (NMonS), which provides traffic status and optical performance measurements to the CDS (again, by means of the CP). The functionalities of adaptability and monitoring are handled in each node through a physical layer manager, which works as a common interface toward the CP, and also through the Network Management System (NMS). In the following, an overview of each component of the proposed framework is provided. 3.1. The Cognitive Decision System The Cognitive Decision System (CDS) is involved in very diverse tasks related to network control and optimization. Hence, rather than implementing the whole CDS as a monolithic module, this is divided in different modules, each offering a functionality (or a set of related functionalities), and all of them exploiting cognition. Each module implements a feedback loop where interactions with the "environment" guide current and future interactions. However, the feedback loop should not only observe and provide decisions, but a learning module must also be implemented so that it prevents mistakes from previous iterations from being made on future iterations. Each module implements the cognitive loop shown in Figure 1. The CDS modules consist of two main parts: Siracusa, et al. Expires May 4, 2014 [Page 11] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 1. The Cognitive Process, which implements the algorithms to make decisions. It takes into account the current network status and previous history. 2. The Knowledge Engineering Subsystem, which handles the information used by the cognitive process. This element consists of a knowledge base and a learning module, which links the cognitive process with its associated knowledge base and executes methods to update the knowledge base with acquired experience. +--------------------------------------------+ | Cognitive Decision System | | | |+------------------------------------------+| || Knowledge Engineering Subsystem || +----------+ Network || +-----------+ +----------+|| | Network | status || | Generic | | Specific ||| |Monitoring|---------|->| Knowledge |<---+ +--------| Knowledge||| | System | || | Base | | | | Base ||| +----------+ || | (Network | | | | ||| OBSERVE || | status) | | | +----------+|| || +-----------+ | | ^ || || | | LEARN | || |+----------------------------+ v || | | | | +----------+|| | v v | | Specific ||| +---------+ Request | +-------------+ | | Learning ||| | Control |----------|------------>| Specific | | | Module ||| | Plane | Decision | | Cognitive | | +----------+|| |protocols|<---------|-------------| Process | +-------------+| +---------+ | | Module | | ACT | +-------------+ | | ORIENT & DECIDE | | | +--------------------------------------------+ Figure 3 Relationship between a cognitive process and its associated knowledge base Figure 3 presents the building elements of a module of the CDS, as well as their relationship to the network monitoring system and the control and management system. The network monitoring system gathers the network status to a generic knowledge base. Separately, there are specific knowledge bases containing all the information associated with each of the cognitive processes. Therefore, there are as many specific knowledge bases as cognitive processes in the CDS. These databases are updated through a specific learning module, Siracusa, et al. Expires May 4, 2014 [Page 12] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 which is associated to a single cognitive process. Consequently, a cognitive process module can access these two databases (generic and specific) to retrieve information and to update them. Finally, when a decision is made to handle a request or network event, the decision will be communicated to the control plane for its execution. 3.2. Processes and knowledge bases The proposed CDS consists of five cognitive processes running in parallel, as shown in Figure 2. These processes are implemented in the following modules: 1. Traffic Grooming (TG) Module: It is in charge of routing non- optical traffic demands, as for example time division multiplexed (TDM) label-switched paths (LSPs) through existing lightpaths composing the current virtual topology. 2. Virtual Topology Design (VTD) Module: It is in charge of (re)designing the virtual topology and hence the set of lightpaths to be established in the network. This module is used for optimizing network performance by rearranging existing connections. 3. RWA/RMLSA Module: In networks following the ITU-T grid, it solves the routing and wavelength assignment (RWA) problem as well as determines the modulation level. In elastic networks, it solves the routing, modulation level and spectrum allocation (RMLSA) problem. 4. QoT Estimator Module: It provides estimation (i.e., a theoretical prediction) about the quality of transmission (QoT) of new lightpaths to be established in the network as well as the impact on existing connections when undertaking a new one. Thus, the establishment of impairment-aware optical connections relies on this module. Once a new lightpath is established, it verifies the real QoT (which is provided by network monitors) and uses this information to improve the performance of the module for future estimations. Siracusa, et al. Expires May 4, 2014 [Page 13] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 5. Network Planner & Decision Maker (NPDM) Module: This module receives user requests and handles them. It is in charge of deciding whether a traffic demand should be directly routed on the virtual topology or a new lightpath should be established and which parameters should employ. It also determines if the virtual topology or the spectrum allocated to connections should be optimized. In order to solve those tasks it coordinates the operation of the other modules relying on their results. The network planner communicates the actions to be performed to the network nodes through control plane protocols and handles the information received from the network monitoring system. Each module has an associated Knowledge Base (KB) in the knowledge engineering subsystem, which is linked to the cognitive process by means of a learning module. Some of these databases can be read by all modules, since they contain services requirements as well as current network status. These generic databases are: 1. Global Traffic Engineering Database (GTED): contains the information about traffic status in the network. 2. Global Physical Parameters Database (GPPD): contains the information about the physical topology of the network, and the physical monitoring data. 3. SLAs/QoS/QoT requirements: This database contains the service level agreements (SLAs) QoS and QoT parameters associated with different services. Hence, when the cognitive system receives a request associated to a class of service, it can obtain the values of quality of transmission that should be guaranteed when handling that request. 4. CCON Use Cases In this section, a number of applications of cognition proposed in optical networks are discussed. 4.1. Quality of Transmission assessment As described in Sect. 3.2, the establishment of impairment-aware optical connections relies on the QoT estimator module. It should be noticed that once a new lightpath is established, the QoT is verified by means of network monitors, and the result of this verification may be used to improve the behavior of the module for future QoT estimations. Siracusa, et al. Expires May 4, 2014 [Page 14] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 The cognitive operation of this module relies on the utilization of data mining techniques [Wit2011]. A cognitive QoT estimator based on Case-Based Reasoning (CBR) has been recently proposed in [Jim2013]. The key idea in CBR is to solve a new problem by relying on previous experiences (or cases), which are stored in a Knowledge Base (KB). Thus, when facing a new problem, the most similar cases stored in the KB are retrieved, and by reusing those cases, either directly or after adapting them, a solution to the new problem is provided. Moreover, the KB can be updated to include new experiences, which can lead to improving the performance of the system. 4.2. Path Computation The CDS offers the functionality of determining the routes and wavelengths/spectrum for the connections to be established in the network thanks to its RWA/RMLSA module. In optical networks a similar role may be performed by the Path Computation Element (PCE) [RFC4655]. Assuming a fixed grid network, the CDS receives requests for lightpath establishment, and then computes a route and a wavelength for that connection according to the current availabitity of resources in the network, which is stored in the Traffic Engineering Database (TED). The result of such computation (once validated by the QoT estimator module) is used to establish the connection by means of the RSVP-TE protocol [RFC3473]. Then, the CDS can either take care itself of performing the updates to the TED, or rely on the use of the OSPF-TE protocol for that aim, which implies that the TED will be updated after some delay. Therefore, in the latter case, the CDS may decide to assign to an incoming request a resource that has already been assigned to another lightpath, but for which the confirmation from OSPF-TE has not reached yet the central TED. Hence, relying on OSPF-TE to update the TED leads to increasing the blocking probability when compared to a scenario where the TED is directly updated by the CDS. In [Rod2013] a cognitive mechanism based on an elapsed times matrix (ETM) heuristic has been proposed, which aims at avoiding the selection of resources which have been recently assigned by the CDS (or a PCE) to another request, by exploiting recent past history (a situation that may arise, for instance, during a restoration process triggered to deal with a link failure). Please note that this technique can be easily introduced in stateless PCEs without requiring protocol extensions, as it only implies the modification of the underlying PCE algorithm. Siracusa, et al. Expires May 4, 2014 [Page 15] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 4.3. Virtual Topology Design and Reconfiguration A further example of the potential of cognition in optical networks is related to the virtual topology design module of CDS. A multi- objective genetic algorithm has been proposed in [Fer2012] to design impairment-aware and survivable virtual topologies, with the aim of reducing both the energy consumption and the network congestion. In a single execution, the algorithm provides several solutions with different trade-offs in terms of the two optimization objectives just mentioned (i.e., a collection of virtual topologies which constitute a good estimate of the so-called Pareto Optimal Set). This method has been further enhanced with two cognitive techniques based on the utilization of memory to remember solutions successfully used in the past: a) a Tabu list to remember connections with low QoT and b) a learning process to select the most appropriate knowledge for the current network state from that memory. The introduction of cognitive techniques in virtual topology design and reconfiguration leads to significant savings in terms of the total cost of ownership compared to conventional methods. For instance, the case study in [Fer2012] shows that capital and operational expenditures can be respectively reduced by up to 20% and 25%. 5. Implications on the Control Plane A key building block of a Cognitive Capable Optical Network is the Control Plane (CP) that complements the Cognitive Decision System (CDS). Whatever the chosen architectural approach, current CP solutions need to be enhanced to enable the full potential of the cognitive processes running in the CDS. A brief description of the tasks that the CP must perform in order to achieve such a result is provided in the following. 5.1. Disseminate network configuration information The CP should control the network configuration providing a description of the network in terms of physical components, topology, resources availability, and configuration of the used resources. This description has to be continuously kept updated by the CP, by notifying to the cognitive entities any change occurring in the network configuration. In both centralized and distributed cognitive architectures this task can be performed by the OSPF-TE protocol [RFC4203] of the GMPLS suite. The OSPF-TE protocol has to be extended to describe the status of the fixed and configurable parameters of the devices inside a node or associated with a link Siracusa, et al. Expires May 4, 2014 [Page 16] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 (e.g., amplifiers, filters). Regarding the configuration of the devices and the physical components, the CP has different ways to collect this information before disseminating it; indeed, it can be statically provided by the network operator, or it can be dynamically discovered by means of the Link Management Protocol (LMP) [RFC4204]. With respect to the disseminated information, network scalability can potentially be an issue, since OSPF-TE may have a lot of data to advertise; however, this can be mitigated by choosing an appropriate number of parameters needed by the cognitive system and encoding them accordingly. In addition, if a centralized approach is considered, it could be noted that the central cognitive entity, (this is the CDS), should be aware of resource availability, since it is such element itself that makes the decisions on the devices to configure. Nevertheless, OSPF-TE utilization remains paramount to provide the initial configuration of the devices and to update the database of the CDS when links are not available anymore. Moreover, OSPF-TE is a widely used, standardized, and stable protocol; extending it to support the cognitive features is a safer solution than implementing these new features as new in a non- standard solution. 5.2. Feed the cognitive processes with network data and statistics Cognitive processes exploit traffic status information and optical quality of transmission measurements, in order to perform effective decisions during lightpath setup and to foresee potential service disrupting situations. There are different techniques to retrieve the aforementioned information and different networking protocols are available to manage this task (i.e., SNMP [RFC1157], RMON [RFC4502]). The approach proposed in [Sir2012] leverages on a monitoring agent located on each node that collects information about monitored parameters (e.g., power, BER, OSNR, traffic) by querying the physical nodes. This information is sent to a monitoring server located in the cognitive node that collects the information and stores them in a database, which is accessible by the cognitive processes. Moreover, the cognitive entity can also receive alarms from monitoring agents when a critical (or potentially critical) situation at the physical layer faces up. 5.3. Implement the decisions of the cognitive processes on the device The CP has to reserve the resources on the basis of the decisions made by the cognitive processes running in the CDS. Also in this case, in both centralized and distributed cognitive architectures, this task can be performed by a GMPLS protocol, namely the RSVP-TE protocol [RFC3473]. On this account, the RSVP-TE protocol must be extended to carry the instructions that the cognitive entities have Siracusa, et al. Expires May 4, 2014 [Page 17] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 produced for each device throughout the path. In particular, the PATH message requires an extension to encode the configuration parameters of each device on the path (e.g., the modulation format for the transmitters, the port and connectivity parameters of the OXC switching matrices). At the end of this process, via non- standard communications, the CP may also be able to notify the CDS if the required operation has been successfully performed and, in case of failure, report the issue that caused such a failure. In the proposed framework, the CDS should be able to trigger the re- optimization of the resources, in order to achieve a better efficiency in terms of utilization, energy efficiency, etc. Complete information of network status is needed to perform this task. On this account, a distributed approach cannot be easily adopted for such a re-optimization, since the information disseminated by OSPF- TE does not allow the construction of a stateful database. For what concerns the centralized approaches, a standard Path Computation Element (PCE)-based solution [RFC4655] would not be suitable to carry out this task being the PCE's role to answer to path requests forwarded by source nodes. Although the original PCE architecture was not thought to be able to autonomously trigger lightpath activation, some recent standardization efforts are trying to address this issue by means of extensions to PCEP [Ali2013]; such mechanisms should allow a stateful PCE to remotely initiate lightpath setup. However, by the time being, the discussion within IETF is still at an early stage. A feasible centralized implementation based on GMPLS is the one proposed in [Sir2012], in which the CDS itself can initiate a lightpath setup and trigger the RSVP-TE reservation. Once the reservation has been completed, the CP sends a response to the CDS notifying if the required operation has been successfully performed and, in case of failure, reporting the issue that caused such a failure. The process of evolution of the CP may be directed to a joint control of the optical and the packet domains. In this perspective, an SDN-based controller may cooperate with the cognitive entities and the CP of an optical network [Das2012]. The cognitive entities could relieve the SDN controller from the high overhead due to the complexities at the photonic layer. In particular, they could provide to the controller already signaled and optically feasible lightpaths, whose computations are optimized on a multi-layer fashion and tailored on the basis of the needs of the packet layer. 6. Contributing Authors This document was the collective work of several authors. The text and content of this document was contributed by the editors and the Siracusa, et al. Expires May 4, 2014 [Page 18] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 co-authors listed below (the contact information for the editors appears in appropriate section and is not repeated below): Yabin Ye Huawei Technologies Duesseldorf GmbH Riesstr. 25, C-3.0G 80992 Munich - Germany Phone: +49-89-1588344052 Email: yeyabin@huawei.com Dimitrios Klonidis Athens Information Technology Center (AIT) 0.8km Markopoulou Av. 19002 Peania, Athens - Greece Phone: +30-210-6682773 Email: dikl@ait.gr Andrzej Tymecki Orange Labs Poland ul.Czere.niowa 8 21-040 Swidnik, Poland Phone: +48-81-5244467 Email: Andrzej.Tymecki@orange.com Idelfonso Tafur Monroy Technical University of Denmark (DTU) Oerstedsplads 343 DK-2800 Kgs. Lyngby, Denmark Phone: +45 45255186 Email: idtm@fotonik.dtu.dk 7. Security Considerations TBD 8. IANA Considerations This memo includes no request to IANA. Siracusa, et al. Expires May 4, 2014 [Page 19] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 9. References 9.1. Informative References [RFC1157] J. Case, M. Fedor, M. Schoffstall, and J. Devin, "A Simple Network Management Protocol," IETF RFC 1157, May 1990. [RFC3473] L. Berger, "Generalized Multi-Protocol Label Switching (GMPLS) Signaling Resource ReserVation Protocol-Traffic Engineering (RSVP-TE) Extensions," RFC 3473, January 2003. [RFC4203] K. Kompella and Y. Rekhter, "OSPF extensions in support of Generalized Multi-Protocol Label Switching (GMPLS)", IETF RFC 4203, October 2005. [RFC4204] J. Lang, "Link Management Protocol (LMP)", IETF RFC 4204, October 2005. [RFC4502] S. Waldbusser, "Remote Network Monitoring Management Information Base Version 2," IETF RFC 4502, May 2006. [RFC4655] A. Farrel, J.-P. Vasseur, and J. Ash, "A Path Computation Element (PCE)-Based Architecture", RFC 4655, August 2006. [Tav2011] W. Tavernier, D. Papadimitriou, D. Colle, "Learning Capable Communication Network (LCCN) Problem Statement", IETF draft, January 2011, draft-tavernier-irtf-lccn- problem-statement-01.txt. [Beh2013] M. Behringer, M. Pritikin, S. Bjarnason, and A. Clemm, "A Framework for Autonomic Networking", IETF draft, October 2013, draft-behringer-autonomic-network-framework-01.txt. [Tho2006] R.W. Thomas, D.H. Friend, L.A. DaSilva, and A.B. MacKenzie, "Cognitive networks: Adaptation and learning to achieve end-to-end performance objectives," IEEE Communications Magazine, pp. 51-57, Dec. 2006. [Mov2012] Z. Movahevic, M. Ayari, R. Langar, and G. Pujolle, "A survey of autonomic network architectures and evaluation criteria," IEEE Communications Surveys & Tutorials, vol. 14, no. 2, pp. 464-490, Second Quarter 2012. [Ger2012] O. Gerstel, M. Jinno, A. Lord, and S.J.B. Yoo, "Elastic optical networking: a new dawn for the optical layer?," IEEE Communications Magazine, vol. 50, no. 2, pp. s12-s20, February 2012. Siracusa, et al. Expires May 4, 2014 [Page 20] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 [deM2013] I. de Miguel, R. J. Duran, T. Jimenez, N. Fernandez, J. C. Aguado, R. M. Lorenzo, A. Caballero, I. Tafur Monroy, Y. Ye, A. Tymecki, I. Tomkos, M. Angelou, D. Klonidis, A. Francescon, D. Siracusa, E. Salvadori, "Cognitive Dynamic Optical networks", Journal of Optical Communications and Networking, vol. 5, no. 10, pp. A107-A118, Oct. 2013. [Cha2010] C.C.K. Chan, Optical Performance Monitoring - Advanced Techniques for Next-Generation Photonic Networks, Elsevier, 2010. [Kli2010] D. Kliazovich, F. Granelli, and N.L.S. Da Fonseca, "Architectures and cross-layer design for cognitive networks" in Handbook of sensor networks. World Scientific Publishing Co, 2010, Chap. 1. [Zer2010] G.S. Zervas and D. Simeonidou, "Cognitive optical networks: Need, requirements and architecture," in Proc. ICTON 2010, paper We.C1.3. [Wei2012] W. Wei, C. Wang, and J. Yu, "Cognitive optical networks: key drivers, enabling techniques, and adaptive bandwidth services," IEEE Communications Magazine, pp. 106-113, Jan. 2012. [Das2012] S. Das, G. Parulkar, and N. McKeown, "Why OpenFlow/SDN can succeed where GMPLS failed", Technical Digest ECOC 2012, paper Tu.1.D.1. [Wit2011] I.H. Witten, E. Frank, and M.A. Hall, Data Mining: Practical Machine Learning Tools and Techniques (Third Edition). Morgan Kaufmann Publishers, 2011. [Jim2013] T. Jimenez, J.C. Aguado, I. de Miguel, R.J. Duran, M. Angelou, N. Merayo, P. Fernandez, R.M. Lorenzo, I. Tomkos, and E.J. Abril, "A cognitive quality of transmission estimator for core optical networks," Journal of Lightwave Technology, vol. 31, no. 6, pp. 942-951, March 2013. [Rod2013] I. Rodriguez, R.J. Duran, D. Siracusa, I. de Miguel, A. Francescon, J.C. Aguado, E. Salvadori, and R.M. Lorenzo, "Minimization of the impact of the TED inaccuracy problem in PCE-based networks by means of cognition," in Proc. ECOC 2013, paper We.4.E.2. Siracusa, et al. Expires May 4, 2014 [Page 21] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 [Fer2012] N. Fernandez, R.J. Duran, I. de Miguel, N. Merayo, J.C. Aguado, P. Fernandez, T. Jimenez, I. Rodriguez, D. Sanchez, R.M. Lorenzo, E.J. Abril, M. Angelou, and I. Tomkos, "Survivable and impairment-aware virtual topologies for reconfigurable optical networks: a cognitive approach," in Proc. RNDM 2012, pp. 183-189. [Fer2013] N. Fernandez, R.J. Duran, E. Palkopoulou, I. de Miguel, I. Stiakogiannakis, N. Merayo, I. Tomkos, and R.M. Lorenzo, "Techno-economic advantages of cognitive virtual topology design," in Proc. ECOC 2013, paper Tu.3.E.6. [Sir2012] D. Siracusa, E. Salvadori, A. Francescon, A. Zanardi, M. Angelou, D. Klonidis, I. Tomkos, D. Sanchez, R.J. Duran, and I. de Miguel, "A control plane framework for future cognitive heterogeneous optical networks," in Proc. ICTON 2012. [Ali2013] Z. Ali, S. Sivabalan, C. Filsfils, R. Varga, and V. Lopez, "Path Computation Element Communication Protocol (PCEP) Extensions for remote-initiated GMPLS LSP Setup", IETF draft (draft-ali-pce-remote-initiated-gmpls-lsp-01.txt), July 2013 10. Acknowledgments This work is supported by the European Commission (EC) Seventh Framework Programme (FP7) CHRON project (Grant No. 258644). This document was prepared using 2-Word-v2.0.template.dot. Siracusa, et al. Expires May 4, 2014 [Page 22] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 Authors' Addresses Domenico Siracusa CREATE-NET v. alla Cascata 56D 38123 Trento - Italy Phone: +39-0461-408400 Email: domenico.siracusa@create-net.org Antonio Francescon CREATE-NET v. alla Cascata 56D 38123 Trento - Italy Phone: +39-0461-408400 Email: antonio.francescon@create-net.org Elio Salvadori CREATE-NET v. alla Cascata 56D 38123 Trento - Italy Phone: +39-0461-408400 Email: elio.salvadori@create-net.org Ramon J. Duran Universidad de Valladolid ETS Ingenieros de Telecomunicacion - Campus Miguel Delibes Paseo de Belen 15, 47011 Valladolid - Spain Email: rduran@tel.uva.es Ignacio de Miguel Universidad de Valladolid ETS Ingenieros de Telecomunicacion - Campus Miguel Delibes Paseo de Belen 15, 47011 Valladolid - Spain Email: ignacio.miguel@tel.uva.es Ruben M. Lorenzo Universidad de Valladolid Siracusa, et al. Expires May 4, 2014 [Page 23] Internet-Draft Cognitive Capable Optical Networks November 4, 2013 ETS Ingenieros de Telecomunicacion - Campus Miguel Delibes Paseo de Belen 15, 47011 Valladolid - Spain Email: rublor@tel.uva.es Siracusa, et al. Expires May 4, 2014 [Page 24]