Figure - available from: Journal of Optical Communications and Networking
This content is subject to copyright. Terms and conditions apply.
Source publication
The requirements for large-scale computing, storage, and network capabilities by the business and scientific communities have led to the development of the grid/cloud network. Grid network users can access a shared set of resources for scientific computing tasks. Cloud tenants are offered IT infrastructure through infrastructure as a service. An ef...
Similar publications
Network Function Virtualization (NFV) is an enabling technology that brings together automated network service management and corresponding virtualized network functions that use an NFV Infrastructure (NFVI) framework. The Virtual Network Function Manager (VNFM) placement in a large-scale distributed NFV deployment is therefore a challenging task d...
Citations
... The investigation outcome reveals the performance through better trade-o® among cost and QoS in minimized computational time. Ding and Ramamurthy [85] developed a joint resource allocation using MILP with best-¯t and Tabu search models. The empirical outcome obtains minimized blocking rate and time-e®ective performance. ...
... Research objectives [23] Minimize storage consumption and maximize user experience satisfaction [24] Enhance the social welfare and pro¯ts of cloud resource providers [25] Improving the overall revenue for users and providers [26] Reduce hardware investment and power consumption [28] Pro¯t maximization [30] Accomplish scalable and energy-e±cient network topology [31] Estimate the public cloud [32] Optimize the reliability, performance [33] Improve user satisfaction [34] Increase the desired amount of network tra±c o®load and meet content requests [35] Minimize SLA violations [36] Dynamic resource allocation and power minimization [38] Optimal power-savings [39] Improve the entire utility of users [41] Maximize the revenue for resource providers and applicants [42] Maximize resource consumption [44] Provide good quality resource allocation and user satisfaction [45] Balance the workload of physical networks and reduce response delay [46] Provide energy e±ciency through minimizing performance overhead, power consumption and SLA violation [47] Optimally matching service providers and seekers [50] Provide high computational and storage capabilities [51] Achieve fairness and security [52] Achieve pro¯t maximization [56] Maximize market surplus and surplus strength [57] Maximize social welfare and pro¯ts [60] Predictable network performance with maximum network utilization and low management overhead [61] Maximize user satisfaction and network revenue [62] Improve the performance by balancing the load [63] Maximize the bene¯ts of service providers [65] Minimize communication tra±c [66] Better adaptability [67] Throughput enhancement [69] SLA constraints [70] Save time as well as enhance the pro¯ts of the cloud provider [72] Maximize cluster consumption and minimize the functional [73] Minimize the e±ciency loss and maximize the fairness [74] Pro¯t maximization [75] For attaining the distribution of optimum resources [76] Optimal resource allocation [77] To ensure the security requests [79] Reduce the wastage of the resources and enhance performance [80] E±ciently allocate VM to attain a reduced response time [81] Minimize the degree of load imbalance [82] Provide good performance by increasing accuracy [83] Develop BoT models constrained using budgets and deadlines [85] Enhance the resource utilization and minimize the capital expense objectives in terms of the parameters mentioned above. Analysis of makespan and energy consumption has been illustrated in Tables 2 and 3. From the table, it can be noticed that SGO [67] requires very less makespan of 10 s, whereas ICSA [64] requires a higher makespan of 1400 s. ...
service (ExaaS) and anything as a service (Xaas) are also added [89, 90, 92, 97]. With cloud computing and wireless networks, a numerous variation of resources are shared by several users at a time. Here, allocating resources to the appropriate users pro-ciently is a challenging task, as it has to be utilized e±ciently with minimized wastage and over-consumption [93]. In recent years, a lot of research studies were performed to clear up utilization issues in cloud resource allocation. In cloud-data centers, the majority of this IoT data will be collected and analyzed. Cloud service providers face substantial hurdles in processing and analyzing this massive amount of data. The IoT cloud platform enables tenants' devices and apps to connect to cloud applications and other devices safely. The IoT broker service in a typical IoT cloud enables secure connections to devices and applications and a publish and subscribe environment. Security, scal-ability, performance, and others are all major IoT cloud service requirements. Service level agreements (SLAs) are commonly used to communicate these service standards to renters [150]. As a result, arti¯cial intelligence (AI), machine learning (ML) and meta-heuristic algorithms are recognized as an e®ective methodology to e±ciently allocate the cloud resources to the applicants [94, 95, 138-140]. Generally, the cloud resources include I/O devices like CPU, RAM/memory, storage, network, bandwidth, quality of experience (QoE), load, virtual machine (VM), quality of service (QoS), power and so on. Further, resource scheduling, pro¯t maximization, energy minimization, work-load balancing, etc., are all dealt with the above-mentioned algorithms [96, 98, 100, 141]. Typically, an exact and e±cient resource allocation paradigm enhances the resource utilization and the performance of the entire cloud system. However, it su®ers from accuracy, stability, user satisfaction, resource utilization, trust analysis and so on. For this reason, improvements are necessitated to develop better resource allocation models [99, 102]. There are several varieties of cloud systems available based on the requirements of the organization implementing clouds such as collab-orative cloud, vehicular cloud, grid cloud, MCC, green cloud, heterogeneity cloud, CRAN and edge cloud [101, 104]. Besides, the challenge exists by continually providing the best QoS without violating the SLAs. Single-objective optimization algorithms were proposed [103, 105] in order to provide optimal cloud resources to all the applicants simultaneously. Still, the limitations like resource failure, dynamicity, the large availability of resource utilities and heterogeneity among resources, etc. lead to evolving multi-objective optimization (MOO) models. In the cloud environment, the computing resources a®ecting the cloud performance and the marketing demands due to the globalized commercial scenario force the cloud manufacturers to establish collaboration between the cloud resources [108, 110]. Moreover, tra±c among the networks is evolving as a major risk in cloud computing in the areas of ultra-high mobility, ultra-high residential places, etc. Generally, cloud computing o®ers a``pay-as-you-go" strategy for the users, which enables them to utilize the resources concerning the cost paid [106, 107]. Therefore, users can utilize the resource virtually and globally
... The investigation outcome reveals the performance through better trade-o® among cost and QoS in minimized computational time. Ding and Ramamurthy [85] developed a joint resource allocation using MILP with best-¯t and Tabu search models. The empirical outcome obtains minimized blocking rate and time-e®ective performance. ...
... Research objectives [23] Minimize storage consumption and maximize user experience satisfaction [24] Enhance the social welfare and pro¯ts of cloud resource providers [25] Improving the overall revenue for users and providers [26] Reduce hardware investment and power consumption [28] Pro¯t maximization [30] Accomplish scalable and energy-e±cient network topology [31] Estimate the public cloud [32] Optimize the reliability, performance [33] Improve user satisfaction [34] Increase the desired amount of network tra±c o®load and meet content requests [35] Minimize SLA violations [36] Dynamic resource allocation and power minimization [38] Optimal power-savings [39] Improve the entire utility of users [41] Maximize the revenue for resource providers and applicants [42] Maximize resource consumption [44] Provide good quality resource allocation and user satisfaction [45] Balance the workload of physical networks and reduce response delay [46] Provide energy e±ciency through minimizing performance overhead, power consumption and SLA violation [47] Optimally matching service providers and seekers [50] Provide high computational and storage capabilities [51] Achieve fairness and security [52] Achieve pro¯t maximization [56] Maximize market surplus and surplus strength [57] Maximize social welfare and pro¯ts [60] Predictable network performance with maximum network utilization and low management overhead [61] Maximize user satisfaction and network revenue [62] Improve the performance by balancing the load [63] Maximize the bene¯ts of service providers [65] Minimize communication tra±c [66] Better adaptability [67] Throughput enhancement [69] SLA constraints [70] Save time as well as enhance the pro¯ts of the cloud provider [72] Maximize cluster consumption and minimize the functional [73] Minimize the e±ciency loss and maximize the fairness [74] Pro¯t maximization [75] For attaining the distribution of optimum resources [76] Optimal resource allocation [77] To ensure the security requests [79] Reduce the wastage of the resources and enhance performance [80] E±ciently allocate VM to attain a reduced response time [81] Minimize the degree of load imbalance [82] Provide good performance by increasing accuracy [83] Develop BoT models constrained using budgets and deadlines [85] Enhance the resource utilization and minimize the capital expense objectives in terms of the parameters mentioned above. Analysis of makespan and energy consumption has been illustrated in Tables 2 and 3. From the table, it can be noticed that SGO [67] requires very less makespan of 10 s, whereas ICSA [64] requires a higher makespan of 1400 s. ...
Due to the exponential rise in the usage of the internet and smart devices, there is a demand for enhanced network efficiency and user satisfaction in a cloud computing environment. Moreover, moving to the cloud systems, it mainly focuses on storage, computation and resources. Due to copious growth, there exist more challenges as well. Among those, resource allocation in cloud computing is the main study, which is essential to determine the QoS and improved performance concerning reliability, confidentiality, trust, security, user satisfaction, profits, etc. This paper plans to prepare a detailed review on trust-based resource allocation in the collaborative cloud. The cloud industry has been assessed in terms of trust-based and other important factors to produce a road plan for resource allocation. Many papers are reviewed here and give a substantial evaluation of cloud resources and their resource allocation models using machine learning and optimization models. First, this survey provides an elaborated study concerning the various cloud resources considering the performance and QoS. Eventually, it extends the research based on trust-based approaches, with the intention of motivating the researchers to focus on trust-based resource allocation on collaborative cloud computing (CCC) atmosphere.
... There is also a broad number of earlier technoeconomic studies published in journals and conferences; however, most of them rely on the original equipment data provided by the seminal paper [19] from the STRONGEST project. Remarkable examples of application include the joint optimization of multilayer network design with and without protection [35][36][37][38], the feasibility and cost-saving opportunities of elastic/gridless optical networks [39][40][41], and cloud/grid design over optical WDM networks [42,43]. Other interesting technoeconomic studies on optical access technologies and C-RAN may also be found in [20,[44][45][46][47][48][49]. ...
This paper introduces a novel and simplified cost model for designing and evaluating a Central Office Rearchitected as a Datacenter (CORD). The model includes equipment and elements for the realization of optical, packet switching, and data center parts with a special focus not only on relative costs but also on power consumption figures. The cost model is then applied to the design and comparison of a metropolitan area network (MAN) including both aggregation and metrocore nodes following several MAN node architectures based on CORD-like leaf-and-spine fabric. In particular, equipment disaggregation at the Central Offices, both on the packet-switching and optical components, can provide important cost savings to telco operators. On the other hand, incorporating computing/storage capabilities in the MAN for the realization of multiaccess edge computing (MEC) has a significant impact on the total network cost but especially on power consumption.
... where C node represents the cost related to the utilization of NFV-nodes (e.g., due to power consumption, software licences, etc., [20]), and C wl represents the cost of the wavelengths required for traffic transport, namely, due to the transponders to be installed at the nodes [21]. Note that, in this paper, we do not consider the capital expenditures due to the network and computing equipment (i.e., switches, routers and servers) as we assume it is not affected by the VNF placement strategy being adopted. ...
... y no existe posibilidad de un aumento de longitudes de onda, entonces será necesario reutiliza la longitud de onda. Las redes DWDM al multiplexar mayor cantidad de longitudes de onda tienen un espacio entre canal de 50 a 200 Ghz (0,2 -1,6 nm)[7].Estas redes al trabajar con más longitudes de onda en un rango donde la atenuación es baja (Coeficiente de atenuación bajo -Tercera Ventana) tienen la ventaja de no necesitar WR, pero la desventaja de la aparición de fenómenos no lineales como FWM (Four Wave Mixing -Mezcla de Cuatro Ondas) entren otros, causando problemas en la recepción de la señal óptica[1,4].El problema RWA o enrutamiento y asignación de longitud de onda, se plantea como la búsqueda de una ruta y una longitud de onda asociada que satisfaga la demanda solicitada, como se observa en la ecuación (3).S i h = (n i ,n D ,n CX ,t CX )(3)Donde, S i h representa la h-ésima solicitud entrante al i-ésimo nodo, n i representa el nodo origen, n D representa el nodo destino, n CX representa el número de conexiones solicitadas y n CX representa el tiempo de conexión solicitado o Holding Time en milisegundos[8,10,[13][14]. ...
The heuristics used to solve the problem of routing and wavelength assignment in optical networks in dynamic scenarios had been partially successful and especially do not respond well when subjected to stress. This article describes a new strategy called Snake-Two that uses the algorithm Snake-One with the monitoring of network links, trying to concentrate on the most used traffic areas leaving more openings in the rest of the network is proposed, which allows decreasinge the likelihood of blocking incoming network requests. The results improve the average probability of blocking up to 37.7% of the highest obtained results. However, the use of the network continues to increase. These results ensure that the improved care of requests for lower scenarios to 140 Erlangs.
... The authors propose a decentralized Cloud firewall framework for individual Cloud customers and propose novel queuing models to solve this problem. In our previous work [26], we investigate the cost-optimized joint resource allocation problem over multi-layer network structure (IP/MPLS-over-OTNover-WDM) from the Cloud provider's perspective. Different CapEx model for multi-layer network is defined in the work. ...
... Independent tasks in one job can be executed in parallel, while dependent tasks must be executed sequentially. A job structure can be modeled as a directed multi-stage graph with a single source/destination node (a DAG), as shown in Fig. 2, similar with the structures we used in our previous work [26]. ...
Resource allocation is an important component of many Cloud computing and datacenter management problems. For infrastructure as a service (IaaS) in the Cloud, the Cloud service provider allocates computing resources such as processor, memory and storage. In addition to the computing infrastructures, the Cloud service provider in the future would also allocate bandwidth for some applications that require guaranteed bandwidth service to transmit a large amount of data. This type of guaranteed bandwidth service can be provided by provisioning a distinct connection from end-to-end, e.g. by provisioning wavelength(s) in a wavelength division multiplexed (WDM) wavelength routed network. In this work, we focus on inter-datacenter network-aware optimal resource allocation in the Cloud from the customer’s perspective. We develop a mixed integer linear programming (MILP) optimal mathematical model and heuristics (Best-Fit and Tabu Search) to solve the budget optimized joint-resource allocation problem to minimize the rental cost for each customer. The experimental results show that our heuristics can achieve approximate optimal solution to the MILP solution and can reduce the customer’s rental cost by at least 30%. The Best-Fit heuristic with shortest job execution time first (STF) and simplest job structure first (SSF) scheduling policies have a better performance in terms of the traffic blocking rate. The traffic blocking rates under both scheduling policies are 5%–25% less than other policies. The Tabu search based heuristic with SSF job scheduling policy has a better performance in terms of the traffic blocking rate than other job scheduling policies. In addition, the Tabu Search based heuristic also reduces the blocking rate by 4%–30% compared with the Best-Fit heuristic under any job scheduling policy.
... Independent tasks in one job can be executed in parallel, while dependent tasks must be executed sequentially. A job structure can be modeled as a directed multi-stage graph with a single source/destination node (a DAG), as shown in Fig. 4.2, similar with the structures we used in our previous work [33]. ...
... The results show that both MILP and heuristics work well to solve the problem, and the heuristics are much more time-efficient. In addition, the Tabu search method achieves the optimal resource allocation, and also reaches a lower blocking rate compared to Best-Fit method[33].2. Budget-optimized network-aware resource allocation in Grid/Cloud over optical networks In addition, we focus on network-aware optimal resource allocation in the Cloud from the customer's perspective. ...
Resource allocation is an evolving part of many Cloud computing and data center management problems. For infrastructure as a service (IaaS) in the Cloud, the Cloud service provider allocates virtual machines (VMs) to the customers with required CPU, memory and disk configurations. In addition to the computing infrastructures, the bandwidth resources would also be allocated to customers for data transmission between reserved VMs. In the near future, users may also want to reserve multiple virtual data centers (VDCs) to construct their own virtual Cloud, which could be called data center as a service (DCaaS). For these two types of services, how to provide guaranteed network bandwidth over an optical network and achieve the joint resource allocation is a challenge to the central resource manager.
In this dissertation, we focus on network-aware resource allocation in Cloud/Grid over optical networks first. We investigate this problem from the provider's perspective and user's perspective. A multi-layer (IP-over-OTN-over-WDM) optical network architecture is utilized for reserving network resources. We develop mixed-integer linear programming (MILP) mathematical models and propose different heuristics for the optimal network-aware resource allocation problem from the Cloud/Grid provider's and the customer's perspectives with different targets.
Furthermore, we investigate the network-efficient virtualized cloud infrastructure provisioning (NE-VCIP) problem in IP-over-EON inter-data center network (DCN) based on the DCaaS model. The elastic optical network (EON) is adopted to provide spectrum and cost-efficient networking resources for large bandwidth requests. We develop MILP mathematical models for this problem and propose a cost-optimized heuristic to solve this problem. To investigate the cost and blocking rate for the served demands, different modulation formats and optical transponders are compared in the EON layer, and the sliceable bandwidth variable transponders (SBVT) and optical traffic grooming technology are considered.
Finally the network-efficient virtual resource provisioning is investigated for intra-DCN based on different types of optical intra-DCN architectures: a hybrid packet and circuit switched DCN architecture (HyPaC), a novel optical switching DCN architecture (OSA) with reconfigurable optical switching matrix and a pure optical DCN architecture with fully connected non-blocking optical switching matrix. Multi-objective MILP and mixed-integer quadratic programming (MIQP) models are constructed for the optimal resource provisioning problems for the corresponding DCN architectures.
Adviser: Byrav Ramamurthy
... Pero, el tráfico dinámico hoy el Holding Time es menor al tiempo entre llegada de solicitudes, como consecuencia ocurre que existe una disponibilidad de la red siempre cambiante y dinámica. Este nuevo escenario no permite el uso de algoritmos convencionales dando espacio para el desarrollo de algoritmia heurística y metaheurística (Chiappone et al., 2016;Rodriguez et al., 2015;Pan et al., 2014;Assis, et al., 2010;Zang, et al., 2001). Los indicadores utilizados para evaluar los diferentes algoritmos son la Probabilidad de Bloqueo (PB), la Utilización de la red (UR) y la tasa de Algoritmia Heurística (TAH). ...
... Para efectos de la simulación se utilizó la red óptica NSFNET (Network Science Foundation NETwork, ver Figura 4) con 14 nodos, 21 enlaces y 8 longitudes de onda (ver Figura 7). La demanda de servicios se distribuye uniformemente para todos los nodos, la demanda de solicitud de servicio es de características poisson, el límite máximo de saltos es la cantidad de nodos por la condición del algoritmo Snake One y se niega el reintento de solicitud cuando esta es bloqueada, condiciones similares para los algoritmos que se estudian en este artículo y que se pueden observar en (Rodríguez et al. 2015;Pan et al., 2014;Zang et al., 2010). Información Tecnológica -Vol. ...
This research shows the performance comparison between different heuristico such as genetic algorithms, simulated annealing, Tabu Search, Snake-One and a new metaheuristic called Snake-Two. Varlier were obtained low blocking probability but with a tendency to increase network utilization. The new Snake-Two strategy can prove that the behavior of the blocking probability decreases at low network usage. These results are compared with a conventional algorithm that was used to visualize the improvement of each indicator. The indicators used for comparison, are the blocking probability and the network utilization and a new indicator called Heuristic Algorithm Rate. This strategy proposeo using the congeoted links, allowing increasing the traffic in some areas, and reducing it in other areas. The results are promising to achieve lower blocking probability but the tendency of increasing the use of the network resources continues.
... Thus it is hard to estimate available bandwidth (AB) accurately based on the packet arrival time. An accurate AB can be put to good use in rate-based streaming applications [2], task scheduling in data centers [4], resource allocation in grids/clouds using optical network architecture [61], and congestion control for TCP in data center networks [59,54]. To provide an accurate AB, it is urgent to develop a more efficient AB estimation method that works in high-speed networking and virtual environment. ...
Cloud computing is transforming a large part of IT industry, as evidenced by the increasing popularity of public cloud computingservices, such as Amazon Web Service, Google Cloud Platform, Microsoft Windows Azure, and Rackspace Public Cloud. Manycloud computing applications are bandwidth-intensive, and thus the network bandwidth information of clouds is important for theirusers to manage and troubleshoot the application performance.
The current bandwidth estimation methods originally developed for the traditional Internet, however, face great challenges in clouds dueto virtualization that is the main enabling technique of cloud computing. First, virtual machine scheduling, which is an importantcomponent of computer virtualization for processor sharing, interferes with packet time-stamping and thus corrupts the networkbandwidth information carried by the packet timestamps. Second, rate limiting, which is a basic building block of networkvirtualization for bandwidth sharing, shapes the network packets and thus complicates the bandwidth analysis of the packets. In this dissertation, we tackle the two virtualization challenges to design new bandwidth estimation methodologies for clouds. First, wedesign bandwidth estimation methods for networks with rate limiting, which is widely used in cloud networks. Bandwidth estimation fornetworks with token bucket shapers (i.e., a basic type of rate limiters) has been studied before, and the conclusion is that “bothcapacity and available bandwidth measurement are challenging because of the dichotomy between the raw link bandwidth and thetoken bucket rate”. Our methods are based on in-depth analysis of the multi-modal distributions of measured bandwidths. Second, we expand the design space of bandwidth estimation methods to challenging but not rare networks where accurate and correctpacket time information are hard to obtain, such as in cloud networks with heavy virtual machine scheduling. Specifically, we designand develop a fundamentally new class of sequence-based bandwidth comparison methods that relatively compare the bandwidthinformation of multiple paths instead of accurately estimating the bandwidth information of a single path. By doing so, our methods useonly packet sequence information but not packet time information, and are fundamentally different from the current bandwidthestimation methods that all use packet time information. Furthermore, we design and develop a new class of sequence-basedbandwidth estimation methods by conveying the time information in the packet sequence. Sequence-based bandwidth estimationmethods estimate the bandwidth information of a path using the time information conveyed in the packet sequence from another path.
Adviser: Professor Lisong Xu
... efficient resource scheduling methods for resource allocation in grids/clouds, to improve resource utilization and reduce cost of scheduling. Pan Yi et al [40] emphasize on combined resource allocation in the grid/cloud environment. Also, optical network architecture is used for reserving network bandwidth. ...
Numerous cloud simulator tools and frameworks have been developed to aid the simulation of cloud environments in order to test any newly proposed algorithm, model or concept without having to incur the cost of deploying the same on an actual cloud infrastructure. These cloud simulation frameworks are documented and well-illustrated with examples by their respective authors and there exists several such survey papers which delineate and differentiate the features of these frameworks. In this paper, however, we
cite some of the recent researches taken place in the world of Cloud Computing where some of these cloud simulators were made use of. It can be observed that although most of the cloud simulators and frameworks have similar architectures and functions, they considerably differ when comes to capability and extensibility. It is also observed that few Cloud Computing concepts cannot be satisfactorily simulated by any of these simulators.
