Abstract: Early detection of anomalies in data centers is important to reduce downtimes and the costs of periodic maintenance. However, there is little research on this topic and even fewer on the fusion of sensor data for the detection of abnormal events. The goal of this paper is to propose a method for anomaly detection in data centers by combining sensor data (temperature, humidity, power) and deep learning models. The model described in the paper uses one autoencoder per sensor to reconstruct the inputs. The auto-encoders contain Long-Short Term Memory (LSTM) layers and are trained using the normal samples of the relevant sensors selected by correlation analysis. The difference signal between the input and its reconstruction is then used to classify the samples using feature extraction and a random forest classifier. The data measured by the sensors of a data center between January 2019 and May 2020 are used to train the model, while the data between June 2020 and May 2021 are used to assess it. Performances of the model are assessed a posteriori through F1-score by comparing detected anomalies with the data center’s history. The proposed model outperforms the state-of-the-art reconstruction method, which uses only one autoencoder taking multivariate sequences and detects an anomaly with a threshold on the reconstruction error, with an F1-score of 83.60% compared to 24.16%.
Abstract: Industry data centers often need to sync data changes reliably and instantly from a large-scale of heterogeneous autonomous relational databases accessed via the not-so-reliable Internet, for which a practical generic sync middleware of low maintenance and operation costs is most wanted. To this demand, this paper presented a generic sync middleware system (GSMS), which has been developed, applied and optimized since 2006, holding the principles or advantages that it must be SyncML-compliant and transparent to data application layer logic without referring to implementation details of databases synced, does not rely on host computer operating systems deployed, and its construction is light weighted and hence of low cost. Regarding these hard commitments of developing GSMS, in this paper we stressed the significant optimization breakthrough of GSMS sync delay being well below a fraction of millisecond per record sync. A series of ultimate tests with GSMS sync performance were conducted for a persuasive example, in which the source relational database underwent a broad range of write loads (from one thousand to one million intensive writes within a few minutes). All these tests showed that the performance of GSMS is competent and smooth even under ultimate write loads.
Abstract: In this paper, we present a Platform-as-a-Service (PaaS) model for controlling the privacy enforcement mechanisms applied on user data when stored and processed in Cloud data centers. The proposed architecture consists of establishing user configurable ‘sticky’ policies on the Graphical User Interface (GUI) data-bound components during the application development phase to specify the details of privacy enforcement on the contents of these components. Various privacy classification classes on the data components are formally defined to give the user full control on the degree and scope of privacy enforcement including the type of execution containers to process the data in the Cloud. This not only enhances the privacy-awareness of the developed Cloud services, but also results in major savings in performance and energy efficiency due to the fact that the privacy mechanisms are solely applied on sensitive data units and not on all the user content. The proposed design is implemented in a real PaaS cloud computing environment on the Microsoft Azure platform.
Abstract: Nowadays, data center industry faces strong challenges for increasing the speed and data processing capacities while at the same time is trying to keep their devices a suitable working temperature without penalizing that capacity. Consequently, the cooling systems of this kind of facilities use a large amount of energy to dissipate the heat generated inside the servers, and developing new cooling techniques or perfecting those already existing would be a great advance in this type of industry. The installation of a temperature sensor matrix distributed in the structure of each server would provide the necessary information for collecting the required data for obtaining a temperature profile instantly inside them. However, the number of temperature probes required to obtain the temperature profiles with sufficient accuracy is very high and expensive. Therefore, other less intrusive techniques are employed where each point that characterizes the server temperature profile is obtained by solving differential equations through simulation methods, simplifying data collection techniques but increasing the time to obtain results. In order to reduce these calculation times, complicated and slow computational fluid dynamics simulations are replaced by simpler and faster finite element method simulations which solve the Burgers‘ equations by backward, forward and central discretization techniques after simplifying the energy and enthalpy conservation differential equations. The discretization methods employed for solving the first and second order derivatives of the obtained Burgers‘ equation after these simplifications are the key for obtaining results with greater or lesser accuracy regardless of the characteristic truncation error.
Abstract: Cloud computing can be defined as one of the prominent technologies that lets a user change, configure and access the services online. it can be said that this is a prototype of computing that helps in saving cost and time of a user practically the use of cloud computing can be found in various fields like education, health, banking etc. Cloud computing is an internet dependent technology thus it is the major responsibility of Cloud Service Providers(CSPs) to care of data stored by user at data centers. Scheduling in cloud computing environment plays a vital role as to achieve maximum utilization and user satisfaction cloud providers need to schedule resources effectively. Job scheduling for cloud computing is analyzed in the following work. To complete, recreate the task calculation, and conveyed scheduling methods CloudSim3.0.3 is utilized. This research work discusses the job scheduling for circulated processing condition also by exploring on this issue we find it works with minimum time and less cost. In this work two load balancing techniques have been employed: ‘Throttled stack adjustment policy’ and ‘Active VM load balancing policy’ with two brokerage services ‘Advanced Response Time’ and ‘Reconfigure Dynamically’ to evaluate the VM_Cost, DC_Cost, Response Time, and Data Processing Time. The proposed techniques are compared with Round Robin scheduling policy.
Abstract: Like most entrepreneurs, data center operators pursue goals such as profit-maximization, improvement of the company’s reputation or basically to exist on the market. Part of those aims is to guarantee a given quality of service. Quality characteristics are specified in a contract called the service level agreement. Central part of this agreement is non-functional properties of an IT service. The system availability is one of the most important properties as it will be shown in this paper. To comply with availability requirements, data center operators can use virtualization technologies. A clear model to assess the effect of virtualization functions on the parts of a data center in relation to the system availability is still missing. This paper aims to introduce a basic model that shows these connections, and consider if the identified effects are positive or negative. Thus, this work also points out possible disadvantages of the technology. In consequence, the paper shows opportunities as well as risks of data center virtualization in relation to system availability.
Abstract: The emergence of Cloud data centers has revolutionized
the IT industry. Private Clouds in specific provide Cloud services
for certain group of customers/businesses. In a real-time private
Cloud each task that is given to the system has a deadline that
desirably should not be violated. Scheduling tasks in a real-time
private CLoud determine the way available resources in the system
are shared among incoming tasks. The aim of the scheduling policy is
to optimize the system outcome which for a real-time private Cloud
can include: energy consumption, deadline violation, execution time
and the number of host switches. Different scheduling policies can be
used for scheduling. Each lead to a sub-optimal outcome in a certain
settings of the system. A Bayesian Scheduling strategy is proposed
for scheduling to further improve the system outcome. The Bayesian
strategy showed to outperform all selected policies. It also has the
flexibility in dealing with complex pattern of incoming task and has
the ability to adapt.
Abstract: We model the process of a data center as a multi- objective problem of mapping independent tasks onto a set of data center machines that simultaneously minimizes the energy consump¬tion and response time (makespan) subject to the constraints of deadlines and architectural requirements. A simple technique based on multi-objective goal programming is proposed that guarantees Pareto optimal solution with excellence in convergence process. The proposed technique also is compared with other traditional approach. The simulation results show that the proposed technique achieves superior performance compared to the min-min heuristics, and com¬petitive performance relative to the optimal solution implemented in UNDO for small-scale problems.
Abstract: Server provisioning is one of the most attractive topics in virtualization systems. Virtualization is a method of running multiple independent virtual operating systems on a single physical computer. It is a way of maximizing physical resources to maximize the investment in hardware. Additionally, it can help to consolidate servers, improve hardware utilization and reduce the consumption of power and physical space in the data center. However, management of heterogeneous workloads, especially for resource utilization of the server, or so called provisioning becomes a challenge. In this paper, a new concept for managing workloads based on user behavior is presented. The experimental results show that user behaviors are different in each type of service workload and time. Understanding user behaviors may improve the efficiency of management in provisioning concept. This preliminary study may be an approach to improve management of data centers running heterogeneous workloads for provisioning in virtualization system.
Abstract: As the data-driven economy is growing faster than
ever and the demand for energy is being spurred, we are facing
unprecedented challenges of improving energy efficiency in data
centers. Effectively maximizing energy efficiency or minimising the
cooling energy demand is becoming pervasive for data centers. This
paper investigates overall energy consumption and the energy
efficiency of cooling system for a data center in Finland as a case
study. The power, cooling and energy consumption characteristics
and operation condition of facilities are examined and analysed.
Potential energy and cooling saving opportunities are identified and
further suggestions for improving the performance of cooling system
are put forward. Results are presented as a comprehensive evaluation
of both the energy performance and good practices of energy
efficient cooling operations for the data center. Utilization of an
energy recovery concept for cooling system is proposed. The
conclusion we can draw is that even though the analysed data center
demonstrated relatively high energy efficiency, based on its power
usage effectiveness value, there is still a significant potential for
energy saving from its cooling systems.
Abstract: Scheduling of diversified service requests in
distributed computing is a critical design issue. Cloud is a type of
parallel and distributed system consisting of a collection of
interconnected and virtual computers. It is not only the clusters and
grid but also it comprises of next generation data centers. The paper
proposes an initial heuristic algorithm to apply modified ant colony
optimization approach for the diversified service allocation and
scheduling mechanism in cloud paradigm. The proposed optimization
method is aimed to minimize the scheduling throughput to service all
the diversified requests according to the different resource allocator
available under cloud computing environment.
Abstract: Power consumption is rapidly increased in data centers
because the number of data center is increased and more the scale of
data center become larger. Therefore, it is one of key research items to
reduce power consumption in data center. The peak power of a typical
server is around 250 watts. When a server is idle, it continues to use
around 60% of the power consumed when in use, though vendors are
putting effort into reducing this “idle" power load. Servers tend to
work at only around a 5% to 20% utilization rate, partly because of
response time concerns. An average of 10% of servers in their data
centers was unused. In those reason, we propose dynamic power
management system to reduce power consumption in green data
center. Experiment result shows that about 55% power consumption is
reduced at idle time.
Abstract: With data centers, end-users can realize the pervasiveness of services that will be one day the cornerstone of our lives. However, data centers are often classified as computing systems that consume the most amounts of power. To circumvent such a problem, we propose a self-adaptive weighted sum methodology that jointly optimizes the performance and power consumption of any given data center. Compared to traditional methodologies for multi-objective optimization problems, the proposed self-adaptive weighted sum technique does not rely on a systematical change of weights during the optimization procedure. The proposed technique is compared with the greedy and LR heuristics for large-scale problems, and the optimal solution for small-scale problems implemented in LINDO. the experimental results revealed that the proposed selfadaptive weighted sum technique outperforms both of the heuristics and projects a competitive performance compared to the optimal solution.