Searcharxiv⌕ Search

arXiv subjects

Admela Jukan

Publications and source records attributed to Admela Jukan.

At least 73 records · Page 4Linked to original sources
arXiv2020

Modeling of Deep Neural Network (DNN) Placement and Inference in Edge Computing

With the edge computing becoming an increasingly adopted concept in system architectures, it is expected its utilization will be additionally heightened when combined with deep learning (DL) techniques. The idea behind integrating demanding processing algorithms in Internet of Things (IoT) and edge devices, such as Deep Neural Network (DNN), has in large measure benefited from the development of edge computing hardware, as well as from adapting the algorithms for use in resource constrained IoT devices. Surprisingly, there are no models yet to optimally place and use machine learning in edge computing. In this paper, we propose the first model of optimal placement of Deep Neural Network (DNN) Placement and Inference in edge computing. We present a mathematical formulation to the DNN Model Variant Selection and Placement (MVSP) problem considering the inference latency of different model-variants, communication latency between nodes, and utilization cost of edge computing nodes. We evaluate our model numerically, and show that for low load increasing model co-location decreases the average latency by 33% of millisecond-scale per request, and for high load, by 21%.

cs.NI↗
arXiv2019

Mitigating Censorship with Multi-Circuit Tor and Linear Network Coding

Anonymity networks are providing practical mechanisms to protect its users against censorship by hiding their identity and information content. The best-known anonymity network, The Onion Routing (Tor) network, is however subject to censorship attacks by blocking the public Tor entry routers and a few secret Tor entry points (bridges), thus preventing users to access the Tor. To further advance the evolution of anonymity networks, while addressing censorship attacks, we propose to enhance the well-known multi-circuit Tor technique with linear network coding (LNC) and analyze the resulting censorship success. The results show that LNC can improve the robustness of Tor against censorship.

cs.CR↗
arXiv2019

Enhancing Block-Wise Transfer with Network Coding in CoAP

CoAP (Constrained Application Protocol) with block-wise transfer (BWT) option is a known protocol choice for large data transfer in general lossy IoT network environments. Lossy transmission environments on the other hand lead to CoAP resending multiple blocks, which creates overheads. To tackle this problem, we design a BWT with network coding (NC), with the goal to reducing the number of unnecessary retransmissions. The results show the reduction in the number of block retransmissions for different values of blocksize, implying the reduced transfer time. For the maximum blocksize of 1024 bytes and total probability loss of 0.5, CoAP with NC can resend up to 5 times less blocks.

cs.NI↗
arXiv2019

Engineering a QoS Provider Mechanism for Edge Computing with Deep Reinforcement Learning

With the development of new system solutions that integrate traditional cloud computing with the edge/fog computing paradigm, dynamic optimization of service execution has become a challenge due to the edge computing resources being more distributed and dynamic. How to optimize the execution to provide Quality of Service (QoS) in edge computing depends on both the system architecture and the resource allocation algorithms in place. We design and develop a QoS provider mechanism, as an integral component of a fog-to-cloud system, to work in dynamic scenarios by using deep reinforcement learning. We choose reinforcement learning since it is particularly well suited for solving problems in dynamic and adaptive environments where the decision process needs to be frequently updated. We specifically use a Deep Q-learning algorithm that optimizes QoS by identifying and blocking devices that potentially cause service disruption due to dynamicity. We compare the reinforcement learning based solution with state-of-the-art heuristics that use telemetry data, and analyze pros and cons.

cs.NI↗
arXiv2019

Survey of Communication Protocols for Internet-of-Things and Related Challenges of Fog and Cloud Computing Integration

The fast increment in the number of IoT (Internet of Things) devices is accelerating the research on new solutions to make cloud services scalable. In this context, the novel concept of fog computing as well as the combined fog-to-cloud computing paradigm is becoming essential to decentralize the cloud, while bringing the services closer to the end-system. This paper surveys on the application layer communication protocols to fulfil the IoT communication requirements, and their potential for implementation in fog- and cloud-based IoT systems. To this end, the paper first presents a comparative analysis of the main characteristics of IoT communication protocols, including request-reply and publish-subscribe protocols. After that, the paper surveys the protocols that are widely adopted and implemented in each segment of the system (IoT, fog, cloud), and thus opens up the discussion on their interoperability and wider system integration. Finally, the paper reviews the main performance issues, including latency, energy consumption and network throughput. The survey is expected to be useful to system architects and protocol designers when choosing the communication protocols in an integrated IoT-to-fog-to-cloud system architecture.

cs.NI↗
arXiv2018

On Coding for Reliable VNF Chaining in DCNs

We study how erasure coding can improve service reliability in Data Center Networks (DCN). To this end, we find that coding can be best deployed in systems, where i) traffic is split into multiple parallel sub-flows, ii) each sub-flow is encoded; iii) SFC along with their corresponding Virtual Network Functions (VNF) concatenated are replicated into at least as many VNF instances as there are sub-flows, resulting in parallel sub- SFCs; and iv) all coded sub-flows are distributed over parallel paths and processed in parallel. We study service reliability as function of the level of parallelization within DCN and the resulting amount of redundancy. Based on the probability theory and by considering failures of path segments, VNF and server failures, we analytically derive the probability that parallel subflows are successfully processed by the parallelized SFC and that the original serial traffic can be successfully recovered without service interruptions.We compare the proposed failure protection with coding and the standard backup protection and evaluate the related overhead of both methods, including decoding, traffic redirection and VNF migration. The results not only show the benefit of our scheme for reliability, but also a reduced overhead required in comparison to backup protection.

cs.PF↗
arXiv2018

Computing Exact and Approximate Blocking Probabilities in Elastic Optical Networks

In this paper, we propose the first exact Markov model for connection blocking analysis in elastic optical networks, based on the occupancy status of spectrum slices on all links due to arrivals and departures of various classes of connections in a network. Since the complexity of the exact Markov model grows exponentially with the link capacity, number of links, routes, and classes of demands, we further advance the state-of-the-art in computing approximate blocking probability in elastic optical networks and propose two novel approximations, i.e., load-independent and load-dependent. These approximations are used to compute state-dependent per-class connection setup rates in multi-class elastic optical networks with or without spectrum converters by taking into account the spectrum fragmentation factor in each state. We validate approximation analysis by exact and/or simulation results, and show that load-independent and load-dependent approximations can be more accurately used than previously proposed approximations, under a random-fit (RF) and a first-fit (FF) spectrum allocation policies. The approximate results match closely with the exact model, for smaller networks, and with the simulations under a variety of network scenarios.

cs.NI↗
arXiv2018

Artificial Intelligence (AI) Methods in Optical Networks: A Comprehensive Survey

Artificial intelligence (AI) is an extensive scientific discipline which enables computer systems to solve problems by emulating complex biological processes such as learning, reasoning and self-correction. This paper presents a comprehensive review of the application of AI techniques for improving performance of optical communication systems and networks. The use of AI-based techniques is first studied in applications related to optical transmission, ranging from the characterization and operation of network components to performance monitoring, mitigation of nonlinearities, and quality of transmission estimation. Then, applications related to optical network control and management are also reviewed, including topics like optical network planning and operation in both transport and access networks. Finally, the paper also presents a summary of opportunities and challenges in optical networking where AI is expected to play a key role in the near future.

cs.AI↗
arXiv2017

Improving Reliability of Service Function Chains with Combined VNF Migrations and Replications

The Network Function Virtualization (NFV) paradigm is enabling flexibility, programmability and implementation of traditional network functions into generic hardware, in form of Virtual Network Functions (VNFs). To provide services, the VNFs are commonly concatenated in a certain ordered sequence, known as Service Function Chains (SFCs). SFCs are usually required to meeting a certain level of reliability. This creates the need to place the VNFs while optimizing reliability jointly with other objectives, such as network and server load balancing. Traditional migration and replication mechanisms, commonly used for Virtual Machines (VM) in data centers, can be used to improve SFC reliability. We study how to improve service reliability using jointly replications and migrations, considering the chaining problem inherent in NFV. While replications provide reliability, performing migrations to more reliable servers decreases the resource overhead. A Linear Programming (LP) model is presented to study the impact of active-active configurations on the network and server resources. Additionally, to provide a fast recovery from server failures, we consider N-to-N configurations in NFV networks and study its impact on server resources. The results show that replications do not only improve reliability, but can also be used to achieving a better server and network load balancing, and when used jointly with migrations can improve resource utilization without degrading reliability.

cs.NI↗
arXiv2017

A Reliability Study of Parallelized VNF Chaining

In this paper, we study end-to-end service reliability in Data Center Networks (DCN) with flow and Service Function Chains (SFCs) parallelism. In our approach, we consider large flows to i) be split into multiple parallel smaller sub-flows; ii) SFC along with their VNFs are replicated into at least as many VNF instances as there are sub-flows, resulting in parallel sub-SFCs; and iii) all sub-flows are distributed over multiple shortest paths and processed in parallel by parallel sub-SFCs. We study service reliability as a function of flow and SFC parallelism and placement of parallel active and backup sub-SFCs within DCN. Based on the probability theory and by considering both server and VNF failures, we analytically derive for each studied VNF placement method the probability that all sub-flows can be successfully processed by the parallelized SFC without service interruption. We evaluate the amount of backup VNFs required to protect the parallelized SFC with a certain level of service reliability. The results show that the proposed flow and SFC parallelism in DCN can significantly increase end-to-end service reliability, while reducing the amount of backup VNFs required, as compared to traditional SFCs with serial traffic flows.

cs.NI↗
arXiv2017

Balancing the Migration of Virtual Network Functions with Replications in Data Centers

The Network Function Virtualization (NFV) paradigm is enabling flexibility, programmability and implementation of traditional network functions into generic hardware, in form of the so-called Virtual Network Functions (VNFs). Today, cloud service providers use Virtual Machines (VMs) for the instantiation of VNFs in the data center (DC) networks. To instantiate multiple VNFs in a typical scenario of Service Function Chains (SFCs), many important objectives need to be met simultaneously, such as server load balancing, energy efficiency and service execution time. The well-known \emph{VNF placement} problem requires solutions that often consider \emph{migration} of virtual machines (VMs) to meet this objectives. Ongoing efforts, for instance, are making a strong case for migrations to minimize energy consumption, while showing that attention needs to be paid to the Quality of Service (QoS) due to service interruptions caused by migrations. To balance the server allocation strategies and QoS, we propose using \emph{replications} of VNFs to reduce migrations in DC networks. We propose a Linear Programming (LP) model to study a trade-off between replications, which while beneficial to QoS require additional server resources, and migrations, which while beneficial to server load management can adversely impact the QoS. The results show that, for a given objective, the replications can reduce the number of migrations and can also enable a better server and data center network load balancing.

cs.NI↗
arXiv2017

Exploiting Parallelism in Optical Network Systems: A Case Study of Random Linear Network Coding (RLNC) in Ethernet-over-Optical Networks

As parallelism becomes critically important in the semiconductor technology, high-performance computing, and cloud applications, parallel network systems will increasingly follow suit. Today, parallelism is an essential architectural feature of 40/100/400 Gigabit Ethernet standards, whereby high speed Ethernet systems are equipped with multiple parallel network interfaces. This creates new network topology abstractions and new technology requirements: instead of a single high capacity network link, multiple Ethernet end-points and interfaces need to be considered together with multiple links in form of discrete parallel paths. This new paradigm is enabling implementations of various new features to improve overall system performance. In this paper, we analyze the performance of parallel network systems with network coding. In particular, by using random LNC (RLNC), - a code without the need for decoding, we can make use of the fact that we have codes that are both distributed (removing the need for coordination or optimization of resources) and composable (without the need to exchange code information), leading to a fully stateless operation. We propose a novel theoretical modeling framework, including derivation of the upper and lower bounds as well as an expected value of the differential delay of parallel paths, and the resulting queue size at the receiver. The results show a great promise of network system parallelism in combination with RLNC: with a proper set of design parameters, the differential delay and the buffer size at the Ethernet receiver can be reduced significantly, while the cross-layer design and routing can be greatly simplified.

cs.PF↗
arXiv2017

Replication of Virtual Network Functions: Optimizing Link Utilization and Resource Costs

Network Function Virtualization (NFV) is enabling the softwarization of traditional network services, commonly deployed in dedicated hardware, into generic hardware in form of Virtual Network Functions (VNFs), which can be located flexibly in the network. However, network load balancing can be critical for an ordered sequence of VNFs, also known as Service Function Chains (SFCs), a common cloud and network service approach today. The placement of these chained functions increases the ping-pong traffic between VNFs, directly affecting to the efficiency of bandwidth utilization. The optimization of the placement of these VNFs is a challenge as also other factors need to be considered, such as the resource utilization. To address this issue, we study the problem of VNF placement with replications, and especially the potential of VNFs replications to help load balance the network, while the server utilization is minimized. In this paper we present a Linear Programming (LP) model for the optimum placement of functions finding a trade-off between the minimization of two objectives: the link utilization and CPU resource usage. The results show how the model load balance the utilization of all links in the network using minimum resources.

cs.NI↗
arXiv2016

What is a Fog Node A Tutorial on Current Concepts towards a Common Definition

Fog computing has emerged as a promising technology that can bring the cloud applications closer to the physical IoT devices at the network edge. While it is widely known what cloud computing is, and how data centers can build the cloud infrastructure and how applications can make use of this infrastructure, there is no common picture on what fog computing and a fog node, as its main building block, really is. One of the first attempts to define a fog node was made by Cisco, qualifying a fog computing system as a mini-cloud, located at the edge of the network and implemented through a variety of edge devices, interconnected by a variety, mostly wireless, communication technologies. Thus, a fog node would be the infrastructure implementing the said mini-cloud. Other proposals have their own definition of what a fog node is, usually in relation to a specific edge device, a specific use case or an application. In this paper, we first survey the state of the art in technologies for fog computing nodes as building blocks of fog computing, paying special attention to the contributions that analyze the role edge devices play in the fog node definition. We summarize and compare the concepts, lessons learned from their implementation, and show how a conceptual framework is emerging towards a unifying fog node definition. We focus on core functionalities of a fog node as well as in the accompanying opportunities and challenges towards their practical realization in the near future.

cs.NI↗
arXiv2016

Practical Privacy in WDM Networks with All-Optical Layered Encryption

Privacy in form of anonymous communication could be comparably both faster and harder to break in optical routers than in today's anonymous IP networks based on The Onion Routing (Tor). Implementing the practical privacy alloptically,however, is not straightforward, as it requires key generation in each anonymization node to avoid distribution of long keys, and layered encryption, both at the optical line rate. Due to the unavailability of cryptographically strong optical key generation and encryption components, not only a layered encryption is a challenge, but an optical encryption in general. In this paper, we address the challenges of optical anonymous networking for the first time from the system's perspective, and discuss options for practical implementation of all-optical layered encryption. To this end, we propose an optical anonymization component realized with the state-of-the-art optical XOR logic and optical Linear Feedback Shift Registers (oLFSRs). Given that LFSR alone is known for its weak cryptographic security due to its linear properties, we propose an implementation with parallel oLFSRs and analyze the resulting computational security. The results show that proposed optical anonymization component is promising as it can be practically realized to provide a high computational security against deanonymization (privacy) attack.

cs.CR↗
arXiv2016

On the IP Traffic Matrix Problem in Hybrid SDN/OSPF Networks

IP networks with a hybrid control plane deploy a distributed routing protocol like OSPF and the centralized paradigm of SDN in parallel. The advantages and disadvantages of a hybrid control plane have already been extensively discussed in the literature. This paper examines to what extent hybrid SDN/OSPF can solve the IP traffic matrix and related monitoring problems, inherent to the IP layer. The problem persists in hybrid networks, as the required SDN byte counters for a complete traffic matrix may not be sufficiently implemented (or even not at all), or the deployed SDN nodes may be too few, or not adequately located in the network. For such cases, we propose to augment the SDN traffic statistics with SNMP-based measurements on IP backup links. We address implementation and network function virtualization aspects of the required hybrid monitoring infrastructure and discuss the timing issues of the measurements based on hands-on experiences in our lab. We furthermore provide a placement algorithm for SDN nodes and backup links that can guarantee a complete IP traffic matrix.

cs.NI↗
arXiv2016

VNF Placement with Replication for Load Balancing in NFV Networks

Network Function Virtualization (NFV) is a new paradigm, enabling service innovation through virtualization of traditional network functions located flexibly in the network in form of Virtual Network Functions (VNFs). Since VNFs can only be placed onto servers located in networked data centers, which is the NFV's salient feature, the traffic directed to these data center areas has significant impact on network load balancing. Network load balancing can be even more critical for an ordered sequence of VNFs, also known as Service Function Chains (SFCs), a common cloud and network service approach today. To balance the network load, VNF's can be placed in a smaller cluster of servers in the network thus minimizing the distance to the data center. The optimization of the placement of these clusters is a challenge as also other factors need to be considered, such as the resource utilization. To address this issue, we study the problem of VNF placement with replications, and especially the potential of VNFs replications to help load balance the network. We design and compare three optimization methods, including Linear Programing (LP) model, Genetic Algorithm (GA) and Random Fit Placement Algorithm (RFPA) for the allocation and replication of VNFs. Our results show that the optimum placement and replication can significantly improve load balancing, for which we also propose a GA heuristics applicable to larger networks.

cs.NI↗