Applied AI Research Papers

Conceptual Awareness for Human Navigation in A.I

2025, Conceptual Awareness for Human Navigation in A.I

Throughout human history, many significant advancements have emerged thanks to the efforts of great scholars such as Pythagoras, Da Vinci, Newton, Leibniz, Edison, Bell, and numerous others who laid the groundwork for scholarly... more

descriptionView Paper arrow_downwardDownload

The Cognitive Nexus Magazine(IEEE Silicon Valley CIS Chapter) - Volume 1, Issue 1

by Rahul Raja

2025, The IEEE Cognitive Nexus Magazine

This inaugural issue of The Cognitive Nexus, the quarterly flagship magazine of the IEEE Computational Intelligence Society, Santa Clara Valley Chapter, introduces a new scholarly medium dedicated to advancing the frontiers of... more

descriptionView Paper arrow_downwardDownload

Imago Ex Machina: An International Analysis of Copyright in the Era of Generative Artificial Intelligence

by Ivan Paduano

2025, Imago Ex Machina: An International Analysis of Copyright in the Era of Generative Artificial Intelligence

This study provides a comprehensive analysis of the profound challenges posed by generative artificial intelligence (AI) to the international copyright framework. The two central legal dilemmas are examined: the protectability of... more

descriptionView Paper arrow_downwardDownload

The Convergence of Artificial Intelligence and Visual Art: Tools, Ethics, and the Future of Creativity

by IJRASET Publication

2025, International Journal for Research in Applied Science & Engineering Technology (IJRASET)

Artificial Intelligence (AI) has rapidly emerged as a transformative force in the visual arts. From neural networks capable of generating images to real-time video synthesis, AI-powered tools such as DALL•E, Midjourney, and Sora are... more

descriptionView Paper arrow_downwardDownload

AI in the Trenches: Revolutionizing Utility Maintenance

by Hew Carroll

2025

Underground utilities such as stormwater and sanitary sewer systems form the hidden foundation of modern infrastructure. Yet across the United States and globally, these systems are aging, overburdened, and under-inspected. This white... more

descriptionView Paper arrow_downwardDownload

SoulThreads.ai: An AI-Driven Platform for Digital Legacy and Posthumous Interactions

by Nida Hussain

2025

The advent of artificial intelligence (AI) has revolutionized numerous aspects of human life, including the ways individuals remember and interact with deceased loved ones. SoulThreads.ai emerges as an innovative platform designed to... more

descriptionView Paper arrow_downwardDownload

AI-Generated Content and the Question of Copyright

by Yotam Werzansky-Orland

2024, The Market: International Journal of Business

The rise of AI-generated content has prompted a heated debate over whether such works are eligible for copyright protection. Supporters argue that since AI-generated works require human input, they should be eligible for copyright... more

Legally responsible person: Dr (hc) Yangos Hadjiyannis

Fig 2: Example of Process (Author’s Prompt)

ie legal debate arose when the photograph was published by Slater in a book and on various websites. Th: lestion at the heart of the issue was: who owned the copyright to the selfie? David Slater claimed the copy sht, arguing that he had set up the camera equipment and enabled the conditions for the photo to be taker wever, animal rights activists and organisations argued that Naruto, as the ‘creator’ and first author of th jotograph, should hold the copyright. The ‘monkey selfie’ case involves a series of events that revolved around a photograph taken by a Celebes crested macaque using a camera owned by British nature photographer David Slater in 2011. The monkey, named Naruto, took selfies by triggering the camera while playing with it. The most famous selfie, often re- ferred to as the ‘monkey selfie’, captured Naruto’s grinning face and became widely circulated on the internet:

A similar case was brought before the UK Supreme Court by the same appellant and was decided recently.’ The UK Paten the Paten given by a natural that a pa property t Act ‘per the appel person, a tent that mits only one interpretation: an inventor ... must be a natural person, and ant to the Al system; YWO] is not a person at all’. The Al system ‘is not nd it did not devise any relevant invention’, the court added. Therefore, t t Act of 1977 states that only a ‘natural person’ could be considered an inventor. The court ruled that DABUS [the name a person, let alone he court concluded ists a non-human entity as an inventor would violate fundamental principles of intellectual aw in the United Kingdom. It is interesting to note though, that the court mad question on appea was admissibility of the patents the appellant sought for on behalf of whether artificial intelligences are capable of inventing. e it clear that the the Al system, not

1. Whether it falls under the realm of literature, art or science:

The author believes that in order to answer the question of whether Al-generated content should be eligible for copyright protection and become a proprietary subject matter, the main focus should be on just a few factors, namely the actual creator of the work (who or what made the output) and if the output work was made arbi- trarily or consciously and willfully (i.e. whether the entity who produced the work made it on purpose or by way of chance). Another factor to be considered is whether the creating entity copied the output work, or did it cre- ate a completely new and original work. Having said that, receiving inspiration from previously existing works cannot and should not constitute a violation or infringement of copyright. Inspiration is something that cannot be ignored in the realm of art and creativity. It is the ‘sine qua non’ of copyright law'?, whether the new works are created by a human, another animal or a machine. Works which are produced by an Al system according to a prompt, cannot be granted copyright protection, as they were not produced by a conscious and willing entity. Copyright is fundamentally a human-centric concept, grounded in the exercise of human creativity and expre: sion. As long as Al systems lack consciousness, intentionality and the ability to experience emotions, Al-ger erated works do not possess the essential qualities that copyright law seeks to protect. However, the authc does hold the opinion that if it could be proved to a certain extent that willful, conscious thought was put int the process of creating a new work by a non-human entity (animal or machine), and that it was not made t random act of chance, then we have to consider the allocation of copyright to said entity. For example, if | the case of Naruto the monkey, it could be proved that it understood the action of pressing the button in orde to capture its own image, the author believes that it would be much harder to dismiss the notion that on because it was not human was it not eligible to hold the copyright to that photograph.

descriptionView Paper arrow_downwardDownload

UNDERSTANDING THE EFFECTS OF ARTIFICIAL INTELLIGENCE TECHNOLOGIES ON EMPLOYEE BEHAVIOURS

by Dr. Lambert Ofoegbu

2024, UNDERSTANDING THE EFFECTS OF ARTIFICIAL INTELLIGENCE TECHNOLOGIES ON EMPLOYEE BEHAVIOURS.

Abstract: Advancements in AI technologies and robotics could replace 57% of jobs by technology within the OECD. There has been a gap in research on the organisational behavioural tendencies while coping with the perceived outcomes. Given... more

descriptionView Paper arrow_downwardDownload

Unveiling the Synergy: Exploring Advances in Applied Artificial Intelligence and Narrow AI

by Smith Miller

2023

This paper presents a comprehensive assessment of recent advancements in practical artificial intelligence, focusing specifically on Narrow AI. This review scrutinizes the evolving landscape of AI applications across various sectors,... more

descriptionView Paper arrow_downwardDownload

Futurism AI Research and Discussion bot test - AI plays Hangman and tries to guess a movie

by Shifu Careaga

2023

ChatGPT-Futurism AI Research and Discussion https://chat.openai.com/g/g-4djy958q1-futurism-ai-research-and-discussion/c/0fa17edf-8801-45e2-b597-166ffff3dc51 1/29 You tell me what your knowledge base teaches you about Transformers in AI... more

"The Machine" delves into the realm of Al ethics, consciousness, and what defines being human.

“The Machine" raises questions about Al consciousness and humanity, pondering if an Al can be "RoboCop" is a commentary on free will, the nature of humanity, and the struggle to retain

descriptionView Paper arrow_downwardDownload

Multi-objective scheduling of micro-services for optimal service function chains

by Lav Gupta

2023, 2017 IEEE International Conference on Communications (ICC)

Lately application service providers (ASPs) and Internet service providers (ISPs) are being confronted with the unprecedented challenge of accommodating increasing service and traffic demands from their geographically distributed users.... more

descriptionView Paper arrow_downwardDownload

Fault and performance management in multi-cloud based NFV using shallow and deep predictive structures

by Lav Gupta

2023, Journal of Reliable Intelligent Environments

Deployment of Network Function Virtualization (NFV) over multiple clouds accentuates its advantages like flexibility of virtualization, proximity to customers and lower total cost of operation. However, NFV over multiple clouds has not... more

descriptionView Paper arrow_downwardDownload

The P-ART framework for placement of virtual network services in a multi-cloud environment

by Lav Gupta

2023, Computer Communications

Carriers' network services are distributed, dynamic, and investment intensive. Deploying them as virtual network services (VNS) brings the promise of low-cost agile deployments, which reduce time to market new services. If these virtual... more

Carriers' network services are distributed, dynamic, and investment intensive. Deploying them as virtual network services (VNS) brings the promise of low-cost agile deployments, which reduce time to market new services. If these virtual services are hosted dynamically over multiple clouds, greater flexibility in optimizing performance and cost can be achieved. On the flip side, when orchestrated over multiple clouds, the stringent performance norms for carrier services become difficult to meet, necessitating novel and innovative placement strategies. In selecting the appropriate combination of clouds for placement, it is important to look ahead and visualize the environment that will exist at the time a virtual network service is actually activated. This serves multiple purposes-clouds can be selected to optimize the cost, the chosen performance parameters can be kept within the defined limits, and the speed of placement can be increased. In this paper, we propose the PART (Predictive-Adaptive Real Time) framework that relies on predictive-deductive features to achieve these objectives. With so much riding on predictions, we include in our framework a novel concept-drift compensation technique to make the predictions closer to reality by taking care of long-term traffic variations. At the same time, near real-time update of the prediction models takes care of sudden short-term variations. These predictions are then used by a new randomized placement heuristic that carries out a fast cloud selection using a least-cost latency-constrained policy. An empirical analysis carried out using datasets from a queuing-theoretic model and also through implementation on CloudLab, proves the effectiveness of the PART framework. The placement system works fast, placing thousands of functions in a sub-minute time frame with a high acceptance ratio, making it suitable for dynamic placement. We expect the framework to be an important step in making the deployment of carrier-grade VNS on multi-cloud systems, using network function virtualization (NFV), a reality.

descriptionView Paper arrow_downwardDownload

COLAP: A predictive framework for service function chain placement in a multi-cloud environment

by Lav Gupta

2023, 2017 IEEE 7th Annual Computing and Communication Workshop and Conference (CCWC)

Network function virtualization (NFV) over multicloud promises network service providers amazing flexibility in service deployment and optimizing cost. Telecommunications applications are, however, sensitive to performance indicators,... more

descriptionView Paper arrow_downwardDownload

ChatGPT: A Case Study on Copyright Challenges for Generative Artificial Intelligence Systems, European Journal of Risk Regulation (2023), 1-23.

by Nicola Lucchi

2023, European Journal of Risk Regulation

This article focuses on copyright issues pertaining to generative artificial intelligence (AI) systems, with particular emphasis on the ChatGPT case study as a primary exemplar. In order to generate high-quality outcomes, generative AI... more

descriptionView Paper arrow_downwardDownload

Using Machine Learning to Detect Noisy Neighbors in 5G Networks

by Udi Margolin

2023, arXiv (Cornell University)

5G networks are expected to be more dynamic and chaotic in their structure than current networks. With the advent of Network Function Virtualization (NFV), Network Functions (NF) will no longer be tightly coupled with the hardware they... more

descriptionView Paper arrow_downwardDownload

Fault and performance management in multi-cloud based NFV using shallow and deep predictive structures

by H Anthony Chan

2023, Journal of Reliable Intelligent Environments

Deployment of Network Function Virtualization (NFV) over multiple clouds accentuates its advantages like flexibility of virtualization, proximity to customers and lower total cost of operation. However, NFV over multiple clouds has not... more

descriptionView Paper arrow_downwardDownload

Case Study: Deployment of AI Assistant for Commercial Work

by Chew Kean Ho

2023, Zenodo (CERN European Organization for Nuclear Research)

The development of AI has reached a level of sophistication where it can generate new content based on internalized datasets, including both text and images, across different AI products. As of 2023, AI assistant is being used for testing... more

Figure 6.1.2: DALL-E 2 generates large varieties of its target using diffusion model™.

Figure 6.1.1: OpenAl's DALL-E 2 CLIP Latent Algorithm®™

Figure 6.2.2: Chat GPT-3 safety was tested and yielded consistent positive result.

Figure 10.1.1: StableDiffusion failed to generate handshake accurately (more than 6 fingers and out of size). Figure 10.1.1: StableDiffusion failed to generate handshake accurately (more than 6 fingers and out of size). where those are the decision of the publication. should not be limited by certain rules and norms

Table 9.1.2: Total Saving Done using AI Assistant

Table 9.1.1: Total Financial Cost of all 3 Tasks Similarly, Table 9.1.4 tabulated the total cost ir terms of time, Table 9.1.5 tabulated the successful cost saving in terms of time and Table 9.1.6 tabulated the overall performance for all = tasks where a total of 22.88% days saved.

Table 9.1.4: Time Resources Cost of all 3 tasks Table 9.1.5: Time Resources Cost of all 3 tasks using

Table 9.1.3: Financial Performance for Applying AI

descriptionView Paper arrow_downwardDownload

Anomaly detection framework for SFC integrity in NFV environments

by Lisandro Granville

2023, 2017 IEEE Conference on Network Softwarization (NetSoft)

With the increasing deployments of Network Functions Virtualization (NFV) in both industry and academia, it becomes necessary to design mechanisms for keeping the integrity of Service Function Chains (SFC) responsible for NFV services... more

descriptionView Paper arrow_downwardDownload

Case Study: Deployment of AI Assistant for Commercial Work

by Chew Kean Ho

2023, Zenodo

The development of AI has reached a level of sophistication where it can generate new content based on internalized datasets, including both text and images, across different AI products. As of 2023, AI assistant is being used for testing... more

descriptionView Paper arrow_downwardDownload

Multi-objective scheduling of micro-services for optimal service function chains

by Mohammed Samaka

2022, 2017 IEEE International Conference on Communications (ICC)

Lately application service providers (ASPs) and Internet service providers (ISPs) are being confronted with the unprecedented challenge of accommodating increasing service and traffic demands from their geographically distributed users.... more

descriptionView Paper arrow_downwardDownload

Fault and performance management in multi-cloud based NFV using shallow and deep predictive structures

by Mohammed Samaka

2022, Journal of Reliable Intelligent Environments

Deployment of Network Function Virtualization (NFV) over multiple clouds accentuates its advantages like flexibility of virtualization, proximity to customers and lower total cost of operation. However, NFV over multiple clouds has not... more

descriptionView Paper arrow_downwardDownload

Automatic fault detection and diagnosis in cellular networks using operations support systems data

by Payam Alavizadeh

2022, NOMS 2016 - 2016 IEEE/IFIP Network Operations and Management Symposium

Self-Healing is one of the key important functionalities in self-organizing mobile communication networks. Despite its importance, self-healing has yet to receive a deserving research attention in the literature. This paper considers... more

descriptionView Paper arrow_downwardDownload

Multi-objective scheduling of micro-services for optimal service function chains

by Lav Gupta

2022, 2017 IEEE International Conference on Communications (ICC)

Lately application service providers (ASPs) and Internet service providers (ISPs) are being confronted with the unprecedented challenge of accommodating increasing service and traffic demands from their geographically distributed users.... more

descriptionView Paper arrow_downwardDownload

Using Machine Learning to Detect Noisy Neighbors in 5G Networks

by Udi Margolin

2022, ArXiv

5G networks are expected to be more dynamic and chaotic in their structure than current networks. With the advent of Network Function Virtualization (NFV), Network Functions (NF) will no longer be tightly coupled with the hardware they... more

descriptionView Paper arrow_downwardDownload

Multi-objective scheduling of micro-services for optimal service function chains

by deval bhamare

2022, 2017 IEEE International Conference on Communications (ICC)

Lately application service providers (ASPs) and Internet service providers (ISPs) are being confronted with the unprecedented challenge of accommodating increasing service and traffic demands from their geographically distributed users.... more

descriptionView Paper arrow_downwardDownload

Fault and performance management in multi-cloud based NFV using shallow and deep predictive structures

by deval bhamare

2022, Journal of Reliable Intelligent Environments

Deployment of Network Function Virtualization (NFV) over multiple clouds accentuates its advantages like flexibility of virtualization, proximity to customers and lower total cost of operation. However, NFV over multiple clouds has not... more

descriptionView Paper arrow_downwardDownload

A survey on service function chaining

by deval bhamare

2022, Journal of Network and Computer Applications

Cloud computing is gaining significant attention and virtualized datacenters are becoming popular as a costeffective infrastructure. The network services are transitioning from a host-centric to a data-centric model moving the data and... more

descriptionView Paper arrow_downwardDownload

HYPER-VINES: A HYbrid Learning Fault and Performance Issues ERadicator for Virtual NEtwork Services over Multi-Cloud Systems

by Lav Gupta

2022, 2019 International Conference on Computing, Networking and Communications (ICNC)

Fault and performance management systems, in the traditional carrier networks, are based on rule-based diagnostics that correlate alarms and other markers to detect and localize faults and performance issues. As carriers move to Virtual... more

descriptionView Paper arrow_downwardDownload

Fault And Performance Management In Multi-Cloud Virtual Network Services Using AI: A Tutorial And A Case Study

by Lav Gupta

2022, Computer Networks

Carriers find Network Function Virtualization (NFV) and multi-cloud computing a potent combination for deploying their network services. The resulting virtual network services (VNS) offer great flexibility and cost advantages to them.... more

descriptionView Paper arrow_downwardDownload

Fault and performance management in multi-cloud based NFV using shallow and deep predictive structures

by Lav Gupta

2022, Journal of Reliable Intelligent Environments

Deployment of Network Function Virtualization (NFV) over multiple clouds accentuates its advantages like flexibility of virtualization, proximity to customers and lower total cost of operation. However, NFV over multiple clouds has not... more

descriptionView Paper arrow_downwardDownload

The P-ART framework for placement of virtual network services in a multi-cloud environment

by Lav Gupta

2022, Computer Communications

Carriers' network services are distributed, dynamic, and investment intensive. Deploying them as virtual network services (VNS) brings the promise of low-cost agile deployments, which reduce time to market new services. If these virtual... more

Carriers' network services are distributed, dynamic, and investment intensive. Deploying them as virtual network services (VNS) brings the promise of low-cost agile deployments, which reduce time to market new services. If these virtual services are hosted dynamically over multiple clouds, greater flexibility in optimizing performance and cost can be achieved. On the flip side, when orchestrated over multiple clouds, the stringent performance norms for carrier services become difficult to meet, necessitating novel and innovative placement strategies. In selecting the appropriate combination of clouds for placement, it is important to look ahead and visualize the environment that will exist at the time a virtual network service is actually activated. This serves multiple purposes-clouds can be selected to optimize the cost, the chosen performance parameters can be kept within the defined limits, and the speed of placement can be increased. In this paper, we propose the PART (Predictive-Adaptive Real Time) framework that relies on predictive-deductive features to achieve these objectives. With so much riding on predictions, we include in our framework a novel concept-drift compensation technique to make the predictions closer to reality by taking care of long-term traffic variations. At the same time, near real-time update of the prediction models takes care of sudden short-term variations. These predictions are then used by a new randomized placement heuristic that carries out a fast cloud selection using a least-cost latency-constrained policy. An empirical analysis carried out using datasets from a queuing-theoretic model and also through implementation on CloudLab, proves the effectiveness of the PART framework. The placement system works fast, placing thousands of functions in a sub-minute time frame with a high acceptance ratio, making it suitable for dynamic placement. We expect the framework to be an important step in making the deployment of carrier-grade VNS on multi-cloud systems, using network function virtualization (NFV), a reality.

descriptionView Paper arrow_downwardDownload

HYPER-VINES: A HYbrid Learning Fault and Performance Issues ERadicator for VIrtual NEtwork Services over Multi-cloud

by Lav Gupta

2022, ICNC

Fault and performance management systems, in the traditional carrier networks, are based on rule-based diagnostics that correlates alarms and other markers to detect and localize faults and performance issues. As carriers move to Virtual... more

descriptionView Paper arrow_downwardDownload

COLAP: A predictive framework for service function chain placement in a multi-cloud environment

by Lav Gupta

2022, 2017 IEEE 7th Annual Computing and Communication Workshop and Conference (CCWC)

Network function virtualization (NFV) over multicloud promises network service providers amazing flexibility in service deployment and optimizing cost. Telecommunications applications are, however, sensitive to performance indicators,... more

descriptionView Paper arrow_downwardDownload

COLAP: A Predictive Framework for Service Function Chain Placement in a Multi-cloud Environment

by Lav Gupta

2022

Network function virtualization (NFV) over multi-cloud promises network service providers amazing flexibility in service deployment and optimizing cost. Telecommunications applications are, however, sensitive to performance indicators,... more

descriptionView Paper arrow_downwardDownload

Fault and Performance Management in Multi-Cloud Based NFV using Shallow and Deep Predictive Structures

by Lav Gupta

2022, Journal of Reliable Intelligent Environments

Deployment of Network Function Virtualization (NFV) over multiple clouds accentuates its advantages like flexibility of virtualization, proximity to customers and lower total cost of operation. However, NFV over multiple clouds has not... more

descriptionView Paper arrow_downwardDownload

Fault And Performance Management In Multi-Cloud Virtual Network Services Using AI: A Tutorial And A Case Study

by Deval Bhamare

2022, Computer Networks

Carriers find Network Function Virtualization (NFV) and multi-cloud computing a potent combination for deploying their network services. The resulting virtual network services (VNS) offer great flexibility and cost advantages to them.... more

descriptionView Paper arrow_downwardDownload

COLAP: A Predictive Framework for Service Function Chain Placement in a Multi-cloud Environment

by Deval Bhamare

2022

Network function virtualization (NFV) over multi-cloud promises network service providers amazing flexibility in service deployment and optimizing cost. Telecommunications applications are, however, sensitive to performance indicators,... more

descriptionView Paper arrow_downwardDownload

Fault prediction and reliability analysis in a real cellular network

by Yash Kumar

2022, 2017 13th International Wireless Communications and Mobile Computing Conference (IWCMC)

Today, the importance of cellular networks is evergrowing. The increasing complexity of networks is expected to decrease reliability. In order to continue reliable operation in a cost-efficient manner, previous literature has explored... more

Fig. 1. Percentage of Faults in Given Time Interval Hence, the results of these faults are a drop in quality of service provided to the customer and a dramatically increased cost for the provider as maintenance times remain high, and the network remains in the sub-optimal state. The problem lies in the current reactive approach utilized by most carriers. This way only responds to the problem and can only hope to limit the magnitude of the problems caused, not prevent it from happening. This method is not experiencing the best success as when the people need to make calls the most, the network experiences the greatest number of faults. From the data set, which was an array of timestamped faults in a month from a national US mobile operator, Fig. 1 was generated.

Using these two matrices, various transient analyses can be performed. The behavior of CTMC is described by the Kolmogorov differential equation and can be found using the generator matrix Q. Then, the probability vector can be obtained by P(t) = P(0) * e2 with P(0) being the initial probability vector. This paper utilizes the uniformization method because it leads to more efficient computation and results in higher accuracy. | Using this, a probability vector is calculated by: This series utilizes the Probability Transition Matrix, P = I + Q/B, and the summation can be truncated with the error formula:

Furthermore, the results of the CTMC analysis provide crucial information about the reliability of the entire network. These analyses can be adapted each time for new values in order to compute new expected times for occupancy times, the first occurrence of a fault, and steady-state distribution. The transient analysis using the probability matrix in Fig. 4 shows that the probability that the network switches from healthy to unhealthy is very high. In fact, after 12 hours, there is a 95% chance that the system is in an unhealthy state.

TABLE 1 RESULTS OF PREDICTION TECHNIQUES Furthermore, a model was generated from the transient probability vector. The model was developed by calculating the probability of a fault after a certain amount of time and considering it a detection of a fault if the value met a designated threshold level. For this paper, the model was checked every 4 hours with a threshold level of 75%. The model was awarded a 1 for correct predictions, 0 for incorrect predictions, and a resulting percentage was calculated. This model can also be updated by adjusting values with new prediction values, with techniques such as exponential smoothing. However, this is irrelevant for this model because the probability transition matrix values are too high to be effected since maintenance time is much higher than fault inter-arrival time.

descriptionView Paper arrow_downwardDownload

Management And Security Of Multi-Cloud Applications

by Lav Gupta

2022

Single cloud computing platforms, like Amazon’s EC2, Google Cloud and Microsoft Azure, are common and popular today. Obtaining resources from a multiple cloud system gives clients competitive pricing, flexibility of resource provisioning,... more

Figure 5.5. (a) Traffic variation on Chicago-Seattle link

Figure 5.5 (b) Traffic variation on Los Angeles-San Jose Link

Figure 5.6. Comparison of generalization error with an integrated model and FPTV model

relationship with the algorithms discussed. The placement strategy described above has been implemented in a placement framework called

Table 4.1 Platform level execution data intensive.

Train-loss = 2.1575, Test-loss = 4.1718, Train-accuracy = 0.992, Test-accuracy = 0.993 b) Public cloud merged layers (cross-trained) Figure 7.11 Public cloud merged model training and testing

other works we examined the work done in [148]. The two methods have been performed under

The total time spent by voice and data packets in the network can follow any distribution. queueing delays on the link. Intercloud simulation was carried out covering all significant delays

1etworks right to every host. The CloudLab set up created for this study is shown in Figure 5.9. international infrastructure from Internet2, so it has been possible to extend, software-definec

Figure 5.16. Latencies achieved by randomized and M-Sequential in the number of iterations shown in Figure 5.15 Figure 5.15. Number of iterations required by randomized and M-Sequential to achieve latency below the threshold

Figure 4.1 OpenADN Multi-Cloud Management System

It chooses an initial set of clouds and acquires a pre-configured minimum number of virtual

Figure 4.3 Functions of OpenADN relevant to the bootstrap process backoff and repeats request checks for resources and launches workflow instances.

guest OS is Ubuntu 12.04 LT on all the VMs.

Figure 4.6 The Layered Profiling Model measure the absolute timing of events for carrying out the experimental study.

Figure 4.7 User, System and overall CPU time for OpenADN

Legend: nealls: the total number of calls, tottime: total time spent in the given function (excluding sub functions) seconds, percall: tottime divided by ncalls, cumtime: total time in this and all sub-functions seconds, Percall: cumtime divided by primitive calls, filename: data for each function between modules on different virtual hosts.

effects related to functions and workloads (Table 4.6). The ‘t’ value used for the calculation of Table 4.5 The estimated y and the residuals

VNFs and SFC through the EMSs. Operation Support System/Business Support System (OSS/BSS) of the carrier manages the

Figure 5.2. Mapping service function chain to the multi- reconfigured by scaling up functions or even moving the VNFs to different clouds.

5.5.1 Information Available from Carriers and CSPs

State at the time of planning deployment clouds, which give cost and latency below the stated thresholds, can be determined. eing able to provide service. When the states of the target clouds are known, the set of least-cost

5.10.8 Refinement of Latency Prediction by Compensating Concept Drift The FTVP method for handling the concept drift in telecommunication traffic was presented in

Figure 5.11 Extract of FTVP windows divided into four windows. A sample from each of these is given in Figure 5.11.

In another trial of 5000 experiments, 50 each with the number of clouds increasing from 10 to

Figure 5.18 Placement time reported in [148] network functions per request or a total of 3000 functions) take about 1000 seconds.

Figure 5.19. Average placement time reported in [149] functions/instances JoraNFV will take 4285 seconds (Figure 5.19). while SA takes about 2000 sec. Even if we assume a linear increase in time taken, for 3000

Figure 5.21. Acceptance % reported in [153] end-to-end delay and resources allocation cost. A comparison of successful requests is given in

Figure 6.1 A carrier’s broadband service network

a) Virtual Network Service (VNS) functions in a top-down manner. From the illustration of a complex network service in Figure 6.1, we abstract a subset to

b) Service Function Chains (SFCs) A carrier may obtain resources from multiple cloud service providers for the reasons discussed

resources to form one or more SFCs [165] [166].

Table 6.1 Examples of commercially available VNFs shows a VNF with its |] Element Management System (1 EMS) and interfaces to the rest of the

Figure 6.6. Network function virtualization infrastructure e) Routing Protocols for SFC connectivity may be possible through the designated switch and routing devices [165].

6.4.1 Virtual Infrastructure Manager (VIM)

6.8.3 Design and Implementation of the HY PERVINES Framework

In the training mode, the data is curated and partitioned into training and test datasets. Shallow

responsibilities and ill-defined interfaces of the management platforms. the Detection and Localization functions with an optional inclusion of the data pre-processing TYPER-VINES obtains markers from multiple sources, it mitigates the problem of overlapping module. We discuss the important details of the framework below.

particularly important, as the presence of alarms does not always indicate a fault.

stacking a number of autoencoders and carefully designing the hidden layers.

Figure 6.14 Stacked sparse autoencoders minimize the mean square error in the labeled test datasets [29]. tuned using back-propagation and simultaneous adjustment of weights of all the layers |

Table 6.14 Detection Level | performance comparative baseline and the accepted model (SVM in this case) results are shown in Table 6.14

Figure 6.16. Detection Level 2 effectiveness compared to baseline

Figure 6.18 Localization Layer 2 effectiveness compared to baseline 6.9.4 Localization of Impending FP Issues

sparsity proportion (SP). SP gives the proportion of training examples a neuron reacts to. A low

Figure 6.21 Localization of impending fault stacked autoencoder and SVM (baseline) severity levels of the impending faults

Figure 7.1 Overall layout of the loT-Cloud based healthcare the next generation healthcare service in Figure 7.1. 7.3.1 The Healthcare Network Domains

the network. A brief discussion on the important constituent of various zones is given below: gateway in the layout diagram of Figure 7.1. This zone is composed of devices like sensors,

Figure 7.3 The security architecture of the next generation healthcare Figure 7.3 in a simplified form.

Figure 7.4 Trust boundaries between zones Figure 7.4. 7.4.1 Attack Surfaces

Figure 7.5 Hierarchical autoencoder based system The dataflow coming from the IoT domain is passed through the autoencoder of the edge clouds intrusion system is given in Figure 7.5.

the edge models are merged to create public cloud model. way the training of layers at the edge can be used in the public cloud. Figure 7.6 illustrates how

Figure 7.7 Healthcare testbed for dataset generation The IoT domain: Consists of an Arduino Mega based microcontroller and gateway for health IoT sensors. Because of the absence of any Wi-Fi or Ethernet post on this microcontroller, an

autoencoder layers. The other edge cloud models are created similarly.

Train-loss = 4.697, Test-loss = 17.595, Train-accuracy = 0.993, Test-accuracy = 0.995

Train-loss = 8.526, Test-loss = 8.661, Train-accuracy = 0.986, Test-accuracy = 0.985

cloud neural network models created ab-initio with the models created by reusing trained layers from edge cloud neural networks.

Train-loss = 4.517, Test-loss = 6.151 Train-accuracy = 0.989, Test-accuracy = 0.990 a) Public cloud merged layers cloud while Figure 11 b) shows results for the merged model.

Figure 7.14 Learning Performance and Explainability of current and future systems [DARPA]

Table 4.2 Time used for user and system activities on physical machines This simple profiling indicates the possibility of higher load on the CPU, because of potentially (unless otherwise indicated, all times are in seconds).

‘workloads’ and the ‘experimental’ errors. The second column of Table 4.4 shows the sum of 7.66% and 31.35% more.

important for optimization regardless of the workloads.

Table 4.6 Calculation of CI of Effects It can be seen that Cls of the functions are all significant. Also the means of a do not lie in the CI

Table 4.7 Performance of client connection for The Port initialization delays: This activity is required for connecting any node or service with he Port initialization delays: This activity is required for connecting any node or service with affected i.e., connection of the client and the http service to OpenADN are not hindered.

Figure 4.16 Setting up the pPort and overall reduction in the initialization time by 77.1% (Lines 206 and 211).

Total time for initialization 0.001938 seconds Tee A EO) Ghattlae ena thie Total time for initialization 0.002394 seconds Figure 4.18 Setting up the cPor

4.20 and 4.21 line number 125 For sPort the reduction was about 87% with the overall reduction by 40.7% as seen from Figures

Table 4.8 gives a summary of the time gained by optimization of initialization of various ports Total time for initialization 0.002041 seconds Figure 4.21 Setting up the sPort (after optimization udp tunnel port init. The concerned process initialized a communication port with the GC and

Table 5.1 Symbols used Some of the important constraints subject to which the cost optimization is carried out are:

subset of least-cost clouds meeting the latency constraint. The complete algorithm is given ir

short-term changes in traffic as already discussed above. not been separately elaborated as it is based on the SVR model(s) refined for concept drift anc Algorithm 5.3 gives the details of random selection. The procedure PRI ENCY has EDICT LAT] set of clouds. The total cost of the selected clouds is checked against the current minimum cost, gnore any of the clouds in the space N. These conditions are met in our implementation

Table 5.2 Categorization of server resources 5.10.2 Selection of Features for Training the Prediction Models Considering the importance of the selection of predictor variables, due attention was given to this storage. In our experiments, the categories defined are as shown in Table 5.2.

Table 5.3 Predictor variables and the output label 5.10.3 Obtaining Training Datasets We were cognizant of the fact that if a model has been trained with the adequate, realistic obtained good compensation of concept drift with four windows as indicated by the results.

Table 5.4 An extract of the integrated training dataset 5.10.6 Selection of the Machine Learning Model training sets is shown in Table 5.4. There are quite a few AI techniques, involving machine learning, that are potentially applicable

unlimited. The number of iterations or number of trees is set as 100. when new examples are introduced takes time. The maximum depth of each tree has been set at which is not linearly separable. Parameter values that obtained for good results are C=200,

A comparative plot of training and test error ratios (defined as prediction_error/acutal_latency) dataset. These can be overcome by curating the training dataset. is given in Figure 5.10. It can be seen that the model training errors are low and it generalizes be slightly higher than the training errors. A lower test error may indicate overfitting or biases in the

Table 5.8 Probability distribution parameters in different windows SVR with separate window models gives much better predictions on new data-points falling in those

errors show an improved model. Table 5.11 Performance after online updation

1) Virtual Machines (VMs) and Network Function Virtualization Infrastructure (NFVD sed to deploy VNFs. This is also referred to as a NFVI Point of Presence (NFVI-PoP). The

6.5 Comparison of Competitive Network Service Orchestration Offerings

information is not available to adjudge the product on the corresponding criterion.

6.6.3 The FP Problem Statement The FP problem of the carrier networks can be defined as follows:

Table 6.5 List of markers for different carrier services related to mobile, fixed and broadband networks are given in Table 6.5. able to extract relevant features automatically, without human intervention. Some of the markers

6.8.2 AI Techniques used in the Framework values (where applicable), from an actual network [174] are given in Table 6.6. A number of AI techniques, involving machine learning and deep learning, were tested and the

Table 6.7 summarizes the machine learning and deep learning techniques useful for NFV-Cloud

Table 6.8 Coarse (Layer!) and fine (Layer 2) categorization of faults the maximum margin can be classified with f(x) = arg max; fi(x), i.e., choose the class that classifies the example with

parameters. Let us discuss a little more about the manifested and impending faults. maintenance is relied on to catch issues early. In our framework, localization of impending faults procedure impending localization (line 17) calls the deep learning model with the required

Table 6.9 (a) Training dataset The test dataset has “id” and the /ocation for which severity has to be predicted

Table 6.10 List of features from network fault dataset complete set of features from featurel to feature386) is shown for compactness. As part o indicating the intensity of the warning and fault_severity ranging from 0-2 as explained before.

this study, a correlation threshold of 23% was found to improve accuracy.

With this dataset, SVM, on the whole, performs better than ADT and RF giving > 95.49% comparatively better than RF. In each case, 10% cross-validation was used.

For Level-2 detection, we have chosen One-R as the baseline algorithm. The accuracy of ow framework is 13.03% better for ‘impending’ faults and 5.97% better for ‘manifested’ faults, Kernel works well as can be seen from Table 6.15.

At Layer-1, we chose the sequential minimal optimization (SMO) multi-class support vector Table 6.16 Localization Performance the accuracy of Layer-1 localization is 97%, which is a substantial improvement over the baseline

Table 7.1 Suitability of deep learning for IoT healthcare networks as the data is passed through the layers of the deep structure of the networks, the shown in Table 7.1.

Table 7.2 The feature set of the BoT-IoT dataset

Figure 7.8 Extract of the Training Dataset The normal and attack datasets have the features given in Table 7.3. An extract of the

To see the performance of the model in filtering out the attack cases, we used the attack data

The random runs generally give low false positives and the accuracy ranges from about 95% to for several runs is given in Table 7.4. value that decides whether the data has been affected by malicious activity. The confusion matrix Summary and Future Directions

descriptionView Paper arrow_downwardDownload

HYPER-VINES : A HY brid Learning Fault and P erformance

by Lav Gupta

2022

Fault and performance management systems, in the traditional carrier networks, are based on rule-based diagnostics that correlate alarms and other markers to detect and localize faults and performance issues. As carriers move to Virtual... more

descriptionView Paper arrow_downwardDownload

Multi-objective scheduling of micro-services for optimal service function chains

by Lav Gupta

2022, 2017 IEEE International Conference on Communications (ICC)

Lately application service providers (ASPs) and Internet service providers (ISPs) are being confronted with the unprecedented challenge of accommodating increasing service and traffic demands from their geographically distributed users.... more

descriptionView Paper arrow_downwardDownload

COLAP: A Predictive Framework for Service Function Chain Placement in a Multi-cloud Environment

by Lav Gupta

2022

Network function virtualization (NFV) over multi-cloud promises network service providers amazing flexibility in service deployment and optimizing cost. Telecommunications applications are, however, sensitive to performance indicators,... more

descriptionView Paper arrow_downwardDownload

A Deep Learning Approach Towards Prediction of Faults in Wind Turbines

by Joyjit Chatterjee

2022, ArXiv

With the rising costs of conventional sources of energy, the world is moving towards sustainable energy sources including wind energy. Wind turbines consist of several electrical and mechanical components and experience an enormous amount... more

descriptionView Paper arrow_downwardDownload

Automated service delivery platform for C-RANs

by Deval Bhamare

2022, 2017 Second International Conference on Fog and Mobile Edge Computing (FMEC)

Cloud computing is gaining significant attention and virtualized data-centers are becoming popular as cost-effective infrastructure. Recently, there has been a trend to collocate the baseband unit (BBU) functionalities and services from... more

descriptionView Paper arrow_downwardDownload

Multi-objective scheduling of micro-services for optimal service function chains

by Raj Jain

2021, 2017 IEEE International Conference on Communications (ICC)

Lately application service providers (ASPs) and Internet service providers (ISPs) are being confronted with the unprecedented challenge of accommodating increasing service and traffic demands from their geographically distributed users.... more

descriptionView Paper arrow_downwardDownload

Fault And Performance Management In Multi-Cloud Virtual Network Services Using AI: A Tutorial And A Case Study

by Raj Jain

2021, Computer Networks

Carriers find Network Function Virtualization (NFV) and multi-cloud computing a potent combination for deploying their network services. The resulting virtual network services (VNS) offer great flexibility and cost advantages to them.... more

Fig. 1. A carrier’s broadband service network

The broadband services, like other carrier services, are currently provided through networks constructed from physical appliances like routers, aggregation switches and Digital Subscriber Line Access Multiplexers (DSLAMs) from various Original Equipment Manufacturers (OEMs). In discussing the

Fig. 3. Virtual broadband service The figure also shows that the carrier has retained DSLAMs as Physical Network Function (PNF) from their legacy network, as these functions might not have reached their end-of-life. VNFs of a service may belong to different vendors, owned by different operators, managed by different platforms and even unaware of each other. In such a case, the service is a multi- domain VNS [16].

Fig. 4. SFC created on multiple clouds Fig. 4 shows an S VNF2-VNF3-VNF4-V VNF1-VNF2-VNF3-V FC with two paths, i.e., PNFI-VNF1- NF6 for the Internet access and PNF1- NF5-VNF8 for content services. As can be seen in this figure, there are multiple paths through meshed core routers through which traffic can be routed if the selected link fails or if there is congestion on the selected link. VNFs of an SFC are connected in the same manner as the physical

A VNF is the virtual counterpart of a network appliance or a middlebox im off-the-shelf plemented by running software over commercial general purpose servers. The software to implement a network function may run over a general-purpose physical mac physical mac also be instan hine or over the virtual machine(s) created on hines using virtualization software. VNFs can tiated on virtual resources obtained from one or more cloud service providers. Each VNF has a well-defined functional be havior and interfaces for interconnection with other VNFs or PNFs. Fig. 5 shows a VNF with its Element Management System (EMS) and interfaces to the rest of the network [19] 20). Some examples of pre-programmed VNFs are given in Table 1.

Fig. 6. Network function virtualization infrastructure Protocols for routing of traffic, through an SFC, are being worked upon by the standards organizations. For example, two protocols - Segment Routing (SR) and the Network Service Header (NSH) - are under development in the IETF [84]. SR is a modified version of source routing. In SR the IPv6 header is extended to include the Segment Routing Header (SRH), which decides the path of the traffic packets. A segment is a path through a carrier network. The internals of the segment may not be exposed to the users. For example, it may be a Multi-Protocol Label Switching (MPLS) tunnel or may be a sequence of IP routers. Each segment has an ID and may contain information about the treatment of the traffic on that segment. A Software Defined Networking (SDN) controller may utilize the Path Computation Element Protocol (PCEP) to find the most appropriate segments and instruct the classifier to direct the traffic flow accordingly. NSH, on the other hand, can work with IPv4, IPv6, and Ethernet. NSH is an 8-byte header followed by a number of optional variable length context headers containing some metadata to be used by NSH-aware devices. Implementation of service function chains with NSH capabilities requires NSH-aware virtual switches and a central controller.

Fig. 7. Orchestration and management of VNS Fig. /. Orchestration and management of VNS In view of what has been said above, the fault management function becomes a shared responsibility. The relative distribution of responsibilities among various platforms and

Fig. 6. [he FP management framework input with good accuracy. Good reconstruction performance helps in achieving good prediction. For prediction of fault classes or severity of impending faults, a layer of Softmax classifier replaces the decoder layers with h” forming the

Fig. 9 The detection subsystem We propose learning models that have predictive and deductive properties to meet the FP requirements of virtua network services. All the markers available from the management platforms, i.e., the runtime monitoring and measurements, alarms, notifications and warnings configuration changes, and environmental factors are used along with machine learning models trained with historica data to draw inferences about the manifested performance and fault issues. Additionally, the capability of deep learning to map the intricate relationship among the features has been used to predict the impending faults. The framework shown in Fig. 8 consists of three main sub-systems: Data pre-processing, Detection and Localization. Data pre-processing involves collation and normalization of the dataset to remove biases. The pre-processing policy may also involve the reduction of features based on some criterion like correlation with the labels. In the training mode, the available dataset is split into training and test datasets, which are used to train and test all the models. During operation, the marker data is run through the framework to detect and localize problems.

Thus, while manifested faults are manageable with shallow models, impending faults have been tackled with deep

networks can be tuned well for sparse data by using parameters like sparsity regularization and sparsity proportion as discussed in the evaluation section.

Fig. 12. The stacked encoder used for prediction 7. Evaluation of the model

For Level-2 detection, we have chosen One-R as the baseline algorithm. One-R is a simple but accurate classification algorithm, which generates one rule for each predictor and then selects the one with the smallest error. The accuracy of our framework is 13.03% better for ‘impending’ faults and 5.97% better for ‘manifested’ faults, which is a significant improvement (Fig. 13.) Fig. 13. Detection Level 2 effectiveness compared to baseline

Fig. 15. Localization Layer 2 effectiveness compared to baseline amples. Fig. 15 gives the graphical comparison of Level 2 performance of the implemented model (MCML: Multi-class, Multi-layer, in our case SMO) and the baseline. It is seen that the implemented model gives a higher percentage of correctly classified and lower percentage of wrongly classified

To make the deep learning model predict with high accuracy, the first step is to train the stacked autoencoders such that the output is as close a replica of the input as possible. To achieve optimum performance, the stacked autoencoder parameters like the number of hidden layers, the size of the layers, sparsity regulation (SR) and sparsity proportion need to be judiciously arrived at. An example of comparative reconstruction performance is given in Figures 16(a) through 16(d). It is seen that the model with 3 hidden layers of 200/150/100 neurons, respectively, converges quite fast to a low mean-square error. Reconstruction accuracy is important as it affects the prediction based on the trained encoders, which the model is eventually used for [22]. 8. Summary Fig. 16. Mean square error for reconstruction of the input

Fig. 18. Localization of impending fault stacked autoencoder and SVM (baseline) We baselined the above results with those obtained with a shallow model, viz., SVM with RBF kernel which worked very well for detection and could only obtain 73.1% accuracy in localizing impending faults. A comparison between SSAE and SVM models is shown in Fig. 18. The deep structure thus provided a substantial improvement in terms of accuracy of prediction of the severity level of the impending faults.

Fine tuning of the model was done using backpropagation. The accuracy ranges between 72 and 85% with the abridged dataset (~1000 examples) and ~92% with the enhanced dataset (~5000 examples). Experiments were carried out for SR = | and SP=04.

COMPARISON OF SOME COMPETITIVE NETWORK SERVICE ORCHESTRATION SOLUTIONS FROM FP PERSPECTIVE TABLE 3

SHARED FP RESPONSIBILITIES OF DIFFERENT MANAGEMENT ENTITIES TABLE 4

TABLE 5 We have introduced markers by an operational network and before as indicators produced measurements taken by the operations staff. There are a large number of markers that are directly or indirectly related to t he occurrence of an FP issue. These markers become important features in our datasets. Events, that produce these markers, relate to communication, QoS, processing, equipment, and environment. Of course, not only each FP issue would usual also many of the markers would of issue. Also, at any given time y have multiple markers, but appear in more than one type the markers produced may be a result of more than one FP issue. Thus, there is a complex relationship between the markers and the FP issues. This would usually mean that when using machine learning for fault detection and localization, feature engineering, i.e., selection of appropriate markers would be required to get better results. However, deep-learning models , are able to extract relevant features automatically, without human intervention. Some of the markers related to mobile, fixed and broadband networks are given in Table 5.

The metrics used by carriers to measure the health of the network provide important information about the FP problems at the macro level. Use of these as features in the dataset would help learning algorithms to narrow down the scope of localization effort. According to ITU Recommendation regarding the QoS criteria and parameters, a number of basic aspects have to be considered while identifying measurable metrics of service availability [59]. ETSI documents on service availability [56] and on service quality [60] mention metrics that need to be collected and analyzed. The ETSI group specification on service quality metrics [61] recognizes that it is important to have an objective and quantitative metrics to assist in identifying problems when they arise and provide good service to the consumers. Examples of metrics, and their realistic values (where applicable), from an actual network [62] are given in Table 6. We will see how these markers and metrics are used in our framework in later sections.

redundancies, as a good amount of similar information may be available from OSS and MANO. However, making use of redundant data makes up for the gaps in communication among various management platforms.

The Telstra log_feature sub-dataset contains 58,672 examples, with events displaying the presence of different features. The event_type sub-dataset has 31,170 examples, the resource_type sub-dataset has 21,076 and severity_type sub- dataset has 18,552 examples. The test and the training files have 11,171 and 7381 records respectively. They have not been split from a common dataset so the standard 80:20 or a similar ratio is not maintained. A dataset prepared by the consolidation of all sub-datasets has more than 800 features as shown in Table 10. Each fault (with a unique id) is associated with a location, up to 6 features and corresponding volumes, up to three affected resources, up to 5 events, and up to 5 severity types indicating the intensity of the warning and Tr fault’ and 1 indicating ‘fault.’ The detection classification of ‘fault’/‘no-fault’? was implemented with a number of supervised learning techniques of which SVM, ADT, and RF have been

LIST OF FEATURES FROM NETWORK FAULT DATASET A part of the consolidated Telstra dataset is shown in Table 11. Only feature] (out of the complete set of features from feature! to feature386) is s hown for compactness. As part of preprocessing of the dataset, selection of features was carried out based on the degree of correlation of each feature with the labels using the Weka tool study, a correlation thresho accuracy. [78]. With the dataset used in this d of 23% was found to improve

With this dataset, SVM, on the whole, performs better ADT and RF giving = 95.4% accuracy. Considering definitions in Table 13, the true positive (TP) rate for ‘fault’ cases were the highest for SVM showing that these were correctly classified as ‘fault’ cases. Considering the nature of the dataset, this result indicates a good result. There were no fau fau pos hig fau per ts and system said no faults in 2.4% cases. The formance and fault issues. SVM and RF also gave classified by them. than the ts and system said fault in 5.7% cases, while there were false itive and negative rates were the lowest in SVM and the hest in Random Forest. A desirable outcome is that besides classifying faults and faults and no faults as no faults with accuracy, it classifies a very low percentage of faults as ts, thus, helping to do what is intended to do — detect high no- high precision indicating that ‘no-fault’ cases were correctly shown in Table 12. On the basis of accuracy, SVM and ADT perform comparatively better than RF. In each case, 10% cross- validation was used.

To get a sense of the performance of our detection model, using SVM with RBF Kernel, we compared the results with baseline results obtained by Zero-R model. The Zero-R model predicts the majority class. Running on our datasets, the baseline result was about 63%, which indicates that our chosen model gives a substantial improvement over the naive baseline.

Fig. 14 gives a comparison of the performance of our Multi- Class Multi-Layer (MCML) model with the baseline. It can be seen that the accuracy of the classification of MCML is 97.03% against the baseline accuracy of 86.14%. A useful metric for comparison of classifiers is Precision-Recall Area (PRC Area), which gives the tradeoff between precision and recall. A high value indicates high precision (i.e., low false positives) and high recall (i.e., low false negatives). We can see that MCML Level 1 gives a high PRC Area of 0.967 compared to 0.806 of the baseline.

LOCALIZATION LAYER 2 BASELINE PERFORMANCE

TABLE 17 A seen from Table 18, when compared with the baseline algorithm result, we see that multi-class classification with SMO and OvO has a much superior performance, indicating the efficacy of the model. The localization accuracy of the model is 96.04% compared to 90.1% of baseline. The PRC Area of the MCML Level 2 classification is 0.955 against 0.846 of the baseline.

descriptionView Paper arrow_downwardDownload

A survey on service function chaining

by Raj Jain

2021, Journal of Network and Computer Applications

Cloud computing is gaining significant attention and virtualized datacenters are becoming popular as a costeffective infrastructure. The network services are transitioning from a host-centric to a data-centric model moving the data and... more

descriptionView Paper arrow_downwardDownload

NFVIaaS Architectural Framework for Policy Based Resource Placement and Scheduling

by Arun Yerra

2021

One of the goals of Network Functions Virtualization (NFV) is to offer the NFV infrastructure as a service to other SP customers - this is called NFVIaaS. Virtual Network Function (VNF) deployment in this paradigm will drive support for... more

Other average utilization parameters such as VM CPU utilization, VM. Memory utilization, VM disk read IOPS, Network utilization/latency etc. could be used by the policy engine to enforce other types of placement policies. Figure 1: NFVIaaS Architecture for Policy Based Resource Placement and Scheduling

policy" is just a derived subsystem configuration using an appropriate policy enforcement technology to support an actually intended policy. Figure 1: Example of Subsystem Translations of Policy Actions

Figure 2: Global versus Local Policy Engines

Figure 3: A Hierarchical Policy Framework Metro and wide-area network (WAN) used to interconnect data centers would also be independent subsystems with their own policy engines.

Figure 5: Pub/Sub Bus - Hierarchical Policy Framework policy framework requires a policy pub/sub bus between all levels to allow for conflict detection, conflict information propagation, and conflict resolution (Figure 5).

Figure 6: Simplified view of a service provider’s NFV Architecture: Multipoint Ethernet Service Example

Figure 4: Universal Node - an innovative element that integrates on the same platform both compute and networking components Figure 3: A virtualized Point of Presence (PoP) with software resources (Compute Node - CN)

Figure 5: Resource Orchestrator: information base, inputs and outpu' In Figure 5 we illustrate a resource orchestrator (RO) as a functional entity whose task is to map the service graph to the infrastructure resources under some service constraints and tak into account the NF resource descriptions.

Figure 6: The RO Control Plane view. Control plane interfaces are indicated with (line) arrows. Data plane connections are indicated with simple lines.

In turn, each single Compute Node (CN) may also have internal switching resources (see Figure 8). In a carrier environment, in order to meet data path requirements, allocation of compute node internal distributed resources (blades, CPU cores, etc.) may become equivalently important.

Figure 8: Compute Node with internal switching resource

Figure 9: The RO Control Plane view considering a recursive programmatic interface for joint compute, storage and network provisioning Based on the recursion principles shown above and the complexity implied by separate interfaces for compute and network resources, one could imagine a recursive programmatic interface for joint compute, storage and network provisioning as depicted in Figure 9.

Figure 10: PoP DC Network with Compute Nodes (CN) considering the U interface and the URC element

Note however, that traffic between the DC internal SFs (VNF1, VNF2Z VNF3) need to exit and re-enter the DC through the external SDN switch. This, certainly, is sub-optimal an results in ping-pong traffic similar to the local and remote DC case discussed in [I-D.zu-nfvrg-elasticity-vnf]. Figure 2: Black Box Data Center with ext Overlay

Figure 3: White Box Data Center with L2 Overlay

Figure 4: Recursive Domain Virtualization and Joint VNF FC programming: Overarching View Figure 4 illustrates a joint domain virtualization and programming setup. In Figure 4 "[x]" denotes ports of the virtualized data plane while "x" denotes port created dynamically as part of the VNF deployment request. Over the joint software and network virtualization VNF placement and the corresponding traffic steering could be defined in an atomic, which is orchestrated, split and handled to the next levels (see Figure 5) in the hierarchy for further orchestration. Such setup allows clear operational separation, arbitrary domain virtualization (e.g., topology details could be omitted) and constraint based optimization of domain wide resources.

Figure 5: Recursive Domain Virtualization and Joint VNF FC programming: Domain Views

Internet-Draft Figure 6: Big Switch with Big Software definition

Figure 7: Big Switch with Big Software definition with a Network Function - Forwarding Graph (NF-FG)

Figure 8: Virtualizer’s YANG data model: tree view

Figure 10: Functional decomposition of the NFVO and the VIM according to the ETSI MANO » managing the association of the virtualised resources to the physical compute, storage, networking resources.

Management as shown in Figure 11, then a more flexible virtualization programming architecture could be created as shown in Figure 12. Figure 11: Functional decomposition of the NFVO and the VIM with the Joint Software and Network control API

Figure 12: Joint Software and Network Control API: Recurring Flexible Architecture

Figure 13: Single node infrastructure report example: Virtualization view

Figure 15: 3-node infrastructure report example: Virtualization view

Figure 17 and Figure 18 show the allocation request for 3 NFs (NF1: Parental control B.4, NF2: Http Cache 1.2 and NF3: Stateful firewall C) as instrumented over a BiS-BiS node. It can be seen that the configuration request contains both the NF placement and the forwarding overlay definition as a joint request. Figure 17: Simple request of 3 NFs on a single BiS-BiS: Virtualization view

Figure 14: Single node infrastructure report example: xml view

descriptionView Paper arrow_downwardDownload

Automatic fault detection and diagnosis in cellular networks using operations support systems data

by Hamidreza Radmanesh

2021, NOMS 2016 - 2016 IEEE/IFIP Network Operations and Management Symposium

Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

descriptionView Paper arrow_downwardDownload

Applied AI

Related Topics