Heterogeneous Bandwidth Research Articles

Resource learning aware algorithm for route prediction and resources allocation in elastic optical networks with network analytics

We have proposed two algorithms in this paper for dynamic traffic in elastic optical networks (EON). Algorithm1 considers the spatial access blocking probability metric (SABPM) during dynamic operation and route traffic to a path with lower SABPM. Therefore, it considers a path with more contiguous blocks to support heterogeneous bandwidths (BW) and modulation schemes. Algorithm1 keeps a balance of fragmentation on available routes. Algorithm2 makes routing decision for dynamic traffic based on the maximum idle slots as well as minimum SABPM. Therefore, it selects a less fragmented path with more number of available contiguous blocks to support heterogeneous BW. Algorithm2 balances the path utilization and avoids over utilization of congested route. We have evaluated the performance of both algorithms in terms of spatial external fragmentation, contiguity ratio, contiguous aligned spectrum ratio, SABPM, and BW blocking probability (BwBP). Python with seaborn, pandas, and other libraries are used for network analytics. Results show that resource utilization improves with the proposed algorithms. It has been shown that algorithm2 and algorithm1 achieve 27% and 18% traffic gains respectively over the alternate routing.

Distributed virtual selective-forwarding units and SDN-assisted edge computing for optimization of multi-party WebRTC videoconferencing

Network service providers (NSP) have growing interest in placing network intelligence and services at network edges by deploying software-defined network (SDN) and network function virtualization infrastructure. In multi-party WebRTC videoconferencing using scalable video coding, a selective forwarding unit (SFU) provides connectivity between peers with heterogeneous bandwidth and terminals. An important question is where in the network to place the SFU service in order to minimize end-to-end delay between all pairs of peers. Clearly, there is no single optimal place for a cloud SFU for all possible peer locations. We propose placing virtual SFUs at network edges leveraging NSP edge datacenters to optimize end-to-end delay and usage of overall network resources. The main advantage of the distributed edge-SFU framework is that each peer video stream travels the shortest path to reach other peers similar to mesh connection model, whereas each peer uploads a single stream to its edge-SFU avoiding the upload bottleneck. While the proposed distributed edge-SFU framework applies to both best-effort and managed service models, this paper proposes a premium managed, edge-integrated multi-party WebRTC service architecture with bandwidth and delay guarantees within access networks by SDN-assisted slicing of edge networks. The performance of the proposed distributed edge-SFU service architecture is demonstrated by means of experimental results.

RAGraph: A Region-Aware Framework for Geo-Distributed Graph Processing

In many global businesses of multinational enterprises, graph-structure data is usually geographically distributed in different regions to support low-latency services. Geo-distributed graph processing suffers from the Wide Area Networks (WANs) with scarce and heterogeneous bandwidth, thus essentially differs from traditional distributed graph processing. In this paper, we propose RAGraph, a Region-Aware framework for geo-distributed graph processing. At the core of RAGraph, we design a region-aware graph processing framework that allows advancing inefficient global updates locally and enables sensible coordination-free message interactions. RAGraph also contains an adaptive hierarchical message interaction engine to switch interaction modes adaptively based on network heterogeneity and fluctuation, and a discrepancy-aware message filtering strategy to filter important messages. Finally, the experiments show that RAGraph can achieve 1.69X - 40.53X speedup and 20.9% - 97% WAN cost reduction compared with state-of-the-art systems.

PipePar: Enabling fast DNN pipeline parallel training in heterogeneous GPU clusters

Recently, pipeline parallelism for large-scale Deep Neural Network (DNN) training has been developed, which partitions the DNN model across multiple devices (e.g., GPUs) and improves the training efficiency by processing data divided into minibatches as a pipeline. However, existing model partitioning algorithms are mostly designed for homogeneous clusters with the same GPU devices and network connections (e.g., bandwidths), while heterogeneous GPU clusters are widely used in mainstream computing infrastructures. In heterogeneous environment, devices are equipped with different GPUs and network connections, and the efficiency of previous approaches is unsatisfactory due to the unbalanced load of the pipeline stages. In this paper, we propose PipePar, a model partitioning and task placement algorithm for pipeline parallel DNN training in heterogeneous GPU clusters. PipePar is based on dynamic programming with search space pruning that takes into consideration both the heterogeneity of GPUs and network bandwidth. PipePar can profile the DNN model for each type of GPU and conduct model partitioning and task placement based on given GPUs and network connections, which can optimize pipeline load balancing in heterogeneous environments and thus improve training efficiency. We design and implement a pipeline-based distributed deep learning training system in a heterogeneous GPU cluster and show through extensive experiments that PipePar outperforms the baseline approaches in the speed of large-scale DNN training.

Boosting Erasure-Coded Multi-Stripe Repair in Rack Architecture and Heterogeneous Clusters: Design and Analysis

Large-scale distributed storage systems have introduced erasure codes to guarantee high data reliability, yet inevitably at the expense of high repair costs. In practice, storage nodes are usually divided into different racks, and data blocks in storage nodes are often organized into multiple stripes independently manipulated by erasure code. Due to the scarcity and heterogeneity of the cross-rack bandwidth, the cross-rack network transmission dominates the entire repair costs. We argue that when erasure code is deployed in a rack architecture, existing repair techniques are limited in different aspects: neglecting the heterogeneous cross-rack bandwidth, less consideration for multi-stripe failure, and no special treatment on repair link scheduling. In this paper, we present CMRepair, a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Cross-rack Multi-stripe Repair</i> technique that aims to reduce the repair time for multi-stripes failure repair in heterogeneous erasure-coded clusters. CMRepair first carefully chooses the nodes for reading/repairing blocks and searches for the multi-stripe repair solution. It adopts different algorithms to adjust per stripe solution, including the Computation Time Priority (CTP) algorithm based on the greedy idea and the Repair Time Priority (RTP) algorithm based on the meta-heuristics idea. Furthermore, CMRepair selectively schedules the execution orders of cross-rack links, with the primary objective of saturating the unused upload/download bandwidth resources and avoiding network congestion. The experiments show that CMRepair with the CTP algorithm can reduce 27.59%-58.12% of the repair time while only introducing negligible computation overhead, and CMRepair with the RTP algorithm can reduce 33.52%-97.75% of the repair time in an acceptable computation time, over existing repair techniques.

Bandwidth-Aware Scheduling Repair Techniques in Erasure-Coded Clusters: Design and Analysis

Erasure codes offer a storage-efficient redundancy mechanism for maintaining data availability guarantees in storage clusters, yet also incur high network traffic consumption and recovery time in failure repair. Extensive research has been carried out to reduce the recovery time. However, previous works either target specific erasure code constructions which are not commonly used in today’s distributed storage clusters or neglect the heterogeneous bandwidth property in real network environments. Since erasure-coded clusters are typically composed of multi-node with heterogeneous bandwidth and accessed in parallel, the whole recovery time is mainly restricted by the low-bandwidth links. In this article, we propose SMFRepair, a single-node multi-level forwarding repair technique that is designed to improve the performance in heterogeneous networks based on Reed-Solomon codes for general fault tolerance. SMFRepair carefully selects the helper nodes and uses idle nodes to bypass low-bandwidth links. Idle nodes have sufficient and unused network bandwidth. It also pipelines the repair links that are optimized by idle nodes. Furthermore, a multi-node scheduling repair technique, called MSRepair, is proposed. MSRepair carefully schedules the multi-node repair link to saturate the most unoccupied bandwidth and transfers data from as large-bandwidth links as possible, with the primary objective of minimizing the recovery time. Large-scale simulation and Amazon EC2 real experiments show that compared to state-of-the-art repair techniques, SMFRepair can accelerate the single-node recovery by up to 47.69%, and MSRepair can reduce the multi-node recovery time by 33.78% <inline-formula><tex-math notation="LaTeX">$\sim$</tex-math></inline-formula> 67.53%.

Multi-path provisioning in elastic optical network with dynamic on-request optimal defragmentation strategy

Elastic Optical Networks (EONs) are extensively accepted promising solution for future ultra-high-speed optical transmission, due to their capability of efficient and flexible assignment of immense optical bandwidth to the heterogeneous traffic demand. However, it undergoes through an aggravated form of the spectrum fragmentation problem due to the heterogeneous bandwidth requirements, spectrum continuity and contiguity constraints. Several techniques have been presented in the literature to get the optimal solution of this problem while multi-path provisioning comes out to be very compelling among them. Although, it creates further challenges due to high hardware requirement and more spectrum consumption due to guard-bands. In this paper, we propose a novel scheme, called optimal spectrum defragmentation (DF) strategy with split spectrum approach (OSDSSA), to tackle this limitation of multi-path provisioning by utilizing minimum number of spectral paths, which improves the scalability, reduces the extra spectrum consumption by guard-bands and avoids the exhaustion of transponders at nodes. The proposed scheme initially serves a connection request using k-shortest routing paths and first-fit spectrum allocation policy in a best possible way. In case, if a connection request cannot be served then the proposed mechanism proceeds with the triggering of Process Defragmentation on the link disjoint shortest paths followed by splitting of traffic demand on corresponding link-disjoint paths. The Process Defragmentation is triggered on the active connections at the immediate right and left of the largest continuous available spectrum block on the link-disjoint shortest paths sequentially until the requested amount of frequency slots (FSs) are achieved. In this work, reconfiguration includes spectrum re-tuning without changing the routing paths, thereby reducing the unnecessary service disruption, DF complexity and cost, leading to improvement in quality of service (QoS) of the network. The performance of the proposed scheme has been assessed through simulation results where single path and multi-path provisioning algorithms have been considered as benchmarks. Through our proposed scheme bandwidth blocking performance, spectrum utilization ratio, and degree of fragmentation have also been improved significantly.

A New Method for Reconstructing Data on a Single Failure Node in the Distributed Storage System Based on the MSR Code

As a storage method for a distributed storage system, an erasure code can save storage space and repair the data of failed nodes. However, most studies that discuss the repair of fault nodes in the erasure code mode only focus on the condition that the bandwidth of heterogeneous links restricts the repair rate but ignore the condition that the storage node is heterogeneous, the cost of repair traffic in the repair process, and the influence of the failure of secondary nodes on the repair process. An optimal repair strategy based on the minimum storage regenerative (MSR) code and a hybrid genetic algorithm is proposed for single‐node fault scenarios to solve the above problems. In this work, the single‐node data repair problem is modeled as an optimization problem of an optimal Steiner tree with constraints considering heterogeneous link bandwidth and heterogeneous node processing capacity and takes repair traffic and repair delay as optimization objectives. After that, a hybrid genetic algorithm is designed to solve the problem. The experimental results show that under the same scales used in the MSR code cases, our approach has good robustness and its repair delay decreases by 10% and 55% compared with the conventional tree repair topology and star repair topology, respectively; the repair flow increases by 10% compared with the star topology, and the flow rate of the conventional tree repair topology decreases by 40%.

An Efficient Data Partitioning Method in Distributed Heterogeneous Bandwidth Environment

An Efficient Data Partitioning Method in Distributed Heterogeneous Bandwidth Environment

A New Incentive Mechanism to Detect and Restrict Sybil Nodes in P2P File-Sharing Networks with a Heterogeneous Bandwidth

In cooperative P2P networks, there are two kinds of illegal users, namely free riders and Sybils. Free riders are those who try to receive services without any sort of cost. Sybil users are rational peers which have multiple fake identities. There are some techniques to detect free riders and Sybil users which have previously been proposed by a number of researchers such as the Tit-for-tat and Sybil guard techniques. Although such previously proposed techniques were quite successful in detecting free riders and Sybils individually, there is no technique capable of detecting both these riders simultaneously. Therefore, the main objective of this research is to propose a single mechanism to detect both kinds of these illegal users based on Game theory. Obtaining new centrality and bandwidth contribution formulas with an incentive mechanism approach is the basic idea of the present research’s proposed solution. The result of this paper shows that as the life of the network passes, free riders are identified, and through detecting Sybil nodes, the number of services offered to them will be decreased.

Elastic optical network with spectrum slicing for fragmented bandwidth allocation

Elastic Optical Networks (EONs) allow the channel spacing and the spectral width of an optical signal to be dynamically adjusted and hence have become an important paradigm in managing the heterogeneous bandwidth demands of optical backbone networks. The entire available optical spectrum is divided into some spectrum slots which define the smallest granularity of bandwidth and optical signals with variable bandwidths can occupy different number of such slots. The constraints imposed by the physical layer of an EON require that the slots occupied by an optical signal from source to destination have to be consecutive and contiguous in terms of their relative position in the optical spectrum. Furthermore, the same spectrum slots need to be reserved throughout the entire optical signal's path from its source to destination. The above constraints make the routing and spectrum allocation (RSA) in EONs very challenging because unavailability of enough spectrum slots that together equals the spectral width of the optical signal associated with an end-to-end request, will result in blocking of the request. Recent developments in the physical layer technologies have made all-optical ‘slicing’ of a request possible and make the request to be ‘fit’ into multiple non-consecutive spectral slots in an EON. But these all-optical ‘slicers’ employ complex technologies and can be very costly to employ. In this paper, we propose a spectrum allocation scheme for an EON node architecture with these ‘slicers’ and we also formulate a modified RSA scheme for EONs employing slicers, both as a mixed-integer linear programming (MILP) model and a heuristic algorithm. Our main aim is to analyze the tradeoff between the number of slicers that can be used per node versus the spectrum utilization and bandwidth blocking rate. The numerical results show that the proposed scheme with slicers can significantly improve bandwidth blocking rate, compared to the conventional scheme without slicer.

Spatio-temporal heterogeneous bandwidth allocation mechanism against DDoS attack

As the spring up of Internet of Things (IoT), IoT devices are naturally advantageous to create botnets for Distributed Denial of Service (DDoS) attacks. To mitigate the effect of IoT-powered DDoS attacks, multiple bandwidth reservation approaches have been explored. Nonetheless, spreading capability and prompt defense are not fully considered in these approaches. In this paper, we propose STBA, a spatio-temporal heterogeneous bandwidth allocation mechanism, combining with existing protocols and algorithms. We leverage network capability and domain characteristics to achieve reasonable bandwidth allocation among diverse domains in STBA. Our scheme introduces centrality strategy searching for capable spreaders to ensure dependable link access of influential domains. Based on mathematical synthesis, STBA can achieve automatic shift to instantly isolate legitimate and malicious traffic. Finally, STBA realizes better performance than related schemes in comparative experiments.

Joint Online Coflow Optimization Across Geo-Distributed Datacenters

Growing data volumes have been generated, stored and processed across geographically distributed datacenters. Processing such geo-distributed datasets should consider task or job placement, heterogeneous available link bandwidth, and routing decisions. Thus optimizing the transmissions of inter-datacenter flows, especially coflows that capture application-level semantics, is important. However, prior solutions on coflow scheduling either merely ignore routing or fix the endpoint placement. This article focuses on the problem of jointly considering endpoint placement, coflow scheduling, and coflow routing to minimize the average coflow completion time (CCT) across geo-distributed datacenters. This paper first presents the system model and problem formulation, then derives a randomized approximation algorithm for a single coflow. After that it also proposes an online algorithm to handle the complex cases of multiple coflows. Through theoretical analysis, a proof that the proposed algorithms have a non-trivial competitive ratio is given. Results from extensive simulations demonstrate that the proposed algorithms can significantly reduce the CCT of coflows and at the same time, have similar algorithm run time, compared with the state-of-the-art solutions.

Meta-heuristic procedure for enhanced spectrum fragmentation assessment in elastic optical networks

This work proposes the use of a procedure for improving the loss of capacity assessment of heterogeneous bandwidth requests in the min slot-continuity capacity loss (MSCL) spectrum allocation algorithm. The particle swarm optimization (PSO) method is used to determine the best relationship between the number of requested slots and the amount of contiguous free slots in the requested and interfering paths. We also incorporate in the MSCL the influence of the number of hops of a route in its capacity evaluation. Considerable reductions in the path request blocking probability are achieved when the technique is applied, particularly when the number of hops is included to the evaluation.

PAPR Reduction With Mixed-Numerology OFDM

High peak-to-average power ratio (PAPR) is a critical problem in orthogonal frequency-division multiplexing (OFDM). The fifth-generation New Radio (5G NR) facilitates the utilization of multiple heterogeneous bandwidth parts (BWPs), which complicates the PAPR problem even further and introduces inter-numerology interference (INI) between the BWPs. This paper proposes two novel schemes to reduce the PAPR of mixed-numerology OFDM signals. The first scheme is an original enhanced iterative clipping-and-error-filtering (ICEF) approach that cancels efficiently the INI along with PAPR reduction. This allows to achieve efficient PAPR reduction while being compatible with well-known windowed overlap-and-add (WOLA) processing. The second scheme is based on fast-convolution (FC) processing, where PAPR reduction is embedded in the FC filtering carried out using overlapping processing blocks. This allows one to use any existing block-wise PAPR reduction method to reduce the composite signals' PAPR with arbitrary BWP waveforms, and it is thus especially well suited for processing mixed-numerology composite waveforms carrying multiple BWPs with different OFDM numerologies. The performance of the proposed algorithms is evaluated in the 5G NR context, and the essential performance advantages are demonstrated and quantified.

Scheduling Dynamic Multicast Requests in Advance Reservation Environment for Enterprise Video Conferencing Systems

In an enterprise video conferencing system, the enterprise network can be shared by hundreds of video conferences. On the other hand, participants may not arrive at the same time and stay until the end of the conference. The abilities to reserve resources in advance, as well as effective dynamic multicast when participants can join and leave the conference at any time, are essential in the distributed multi-party video conferencing systems. However, the effective advance reservation strategies of the dynamic multicast requests for a heavy traffic case still remains open. In this paper, we investigate the problem of Maximizing the number of admitted Dynamic Multicast requests in the Advance Reservation environment (MDMAR) for the enterprise video conferencing system. We take two path schemes of a fixed path and variable paths, as well as a heterogeneous bandwidth reservation model into account. Firstly, we prove that the MDMAR problem is NP-complete and formulate it mathematically as an integer linear program (ILP) for small networks. Then, we develop greedy algorithms and simulated annealing (SA) algorithms for enterprise networks. Comparative simulations are performed to evaluate the heuristic algorithms for both small networks and enterprise networks. We find that the SA algorithms can provide within 6% lower optimal solutions than the ILP algorithms for our small network, and up to 10% improvement over the greedy algorithms for the large campus or enterprise network.

Integration of 6LOWPAN and Cloud Data Centers: The Role of Link Capacity between Sensor Networks and Cloud Data Centers

Internet of Things (IoT) and Internet of Mobile Things (IoMT) acquired widespread popularity by its ease of deployment and support for innovative applications. The sensed and aggregated data from IoT and IoMT are transferred to Cloud through Internet for analysis, interpretation and decision making. In order to generate timely response and sending back the decisions to the end users or Administrators, it is important to select appropriate cloud data centers which would process and produce responses in a shorter time. Beside several factors that determine the performance of the integrated 6LOWPAN and Cloud Data Centers, we analyze the available bandwidth between various user bases (IoT and IoMT networks) and the cloud data centers. Amidst of various services offered in cloud, problems such as congestion, delay and poor response time arises when the number of user request increases. Load balancing/sharing algorithms are the popularly used techniques to improve the performance of the cloud system. Load refers to the number of user requests (Data) from different types of networks such as IoT and IoMT which are IPv6 compliant. In this paper we investigate the impact of homogeneous and heterogeneous bandwidth between different regions in load balancing algorithms for mapping user requests (Data) to various virtual machines in Cloud. We investigate the influence of bandwidth across different regions in determining the response time for the corresponding data collected from data harvesting networks. We simulated the cloud environment with various bandwidth values between user base and data centers and presented the average response time for individual user bases. We used Cloud- Analyst an open source tool to simulate the proposed work. The obtained results can be used as a reference to map the mass data generated by various networks to appropriate data centers to produce the response in an optimal time.

Impairment Aware Routing, Bit Loading, and Spectrum Allocation in Elastic Optical Networks

The exponential growth of Internet traffic due to bandwidth hungry applications has led to the design of spectrum efficient, flexible, and scalable elastic optical networks. The scope of high spectral-efficient modulation formats and optical orthogonal frequency-division multiplexing (OFDM) in coherent optical communication systems appears to be significant in catering to the heterogeneous bandwidth demands. Furthermore, OFDM enables each subcarrier to be modulated independently to enhance the spectrum efficiency known as bit loading in OFDM. On the other hand, spectral efficiency and the optical reach of a signal are limited by the transmission impairments such as shot noise, amplified spontaneous emission noise, beat noises due to coherent detection, crosstalk in cross-connect, nonlinear and filter narrowing effects, etc. In this paper, we propose a novel impairment-aware routing, bit loading, and spectrum allocation scheme through a mixed-integer linear programming optimization framework. Within our optimization framework, we ensure that a minimum end-to-end bit error rate for each demand is provisioned in the presence of impairments. We also propose a novel heuristic for realistic networks and compare the performance with existing schemes. Through simulation, we demonstrate the benefits of our approach in terms of achieving lower blocking probability and the spectral fragmentation compared to other existing studies.

Efficient Dynamic Routing and Spectrum Assignment for Multifiber Elastic Optical Networks

Elastic optical networks are seen as a promising solution to improve spectrum utilization efficiency by utilizing flex-grid optical orthogonal frequency division multiplexing technology and facilitating flexible bandwidth allocation to services with heterogeneous demands. With dramatic growth of Internet traffic and imminent fiber capacity exhaustion, multiple fibers per link are required to accommodate increasing demands. In this paper, we investigate path selection and spectrum management for heterogeneous bandwidth requests by taking the network topology and expected traffic pattern into account. A novel and efficient solution for path selection and spectrum assignment that optimizes the state of the network after assignment is proposed. We first develop integer linear programming formulations to calculate path selection probabilities offline. Then, we propose a spectrum partition scheme that provides a particular spectrum range for each request size. A next-state-aware spectrum assignment algorithm with resource sharing among partitions is then introduced. We use the demand blocking ratio for dynamically arriving requests as the performance indicator. Simulation results indicate the effectiveness of our proposed schemes for routing and spectrum assignment, as they perform 1 to 2 orders of magnitude better than several baseline schemes.

Practical User Selection With Heterogeneous Bandwidth and Antennas for MU-MIMO WLANs

User selection is one of the most important components for next generation multi-user multiple-input-multiple-output wireless local area networks. However, state-of-the-art approaches neglect the heterogeneity of users in the available bandwidth and the number of antennas, which diminishes their performance considerably. To tackle this challenge, we formulate a novel integer optimization framework to select the antennas of heterogeneous users simultaneously. With estimated signal-to-interference-and-noise ratio of users via channel vector projection, we propose a low-complexity branch-and-prune algorithm to search for the near-optimal combinations of user antennas. Our algorithm is compatible with legacy 802.11ac and is implemented on the software defined radio system. Extensive experiments show that our algorithm achieves around 95% of the optimal throughput and outperforms a benchmark scheme with a $1.18\boldsymbol \times $ gain in realistic indoor environments.