Abstract Over the last few years, big data have emerged as a paradigm for processing and analyzing a large volume of data. Coupled with other paradigms, such as cloud computing, service computing, and Internet of Things, big data processing takes advantage of the underlying cloud infrastructure, which allows hosting and managing massive amounts of data, while service computing allows to process and deliver various data sources as on‐demand services. This synergy between multiple paradigms has led to the emergence ofbig services, as a cross‐domain, large‐scale, and big data‐centric service model. Apart from the adaptation issues (e.g., need of high reaction to changes) inherited from other service models, the massiveness and heterogeneity of big services add a new factor of complexity to the way such a large‐scale service ecosystem is managed in case of execution deviations. Indeed, big services are often subject to frequent deviations at both the functional (e.g., service failure, QoS degradation, and IoT resource unavailability) and data (e.g., data source unavailability or access restrictions) levels. Handling these execution problems is beyond the capacity of traditional web/cloud service management tools, and the majority of big service approaches have targeted specific management operations, such as selection and composition. To maintain a moderate state and high quality of their cross‐domain execution, big services should be continuously monitored and managed in a scalable and autonomous way. To cope with the absence of self‐management frameworks for large‐scale services, the goal of this work is to design an autonomic management solution that takes the whole control of big services in an autonomous and distributed lifecycle process. We combine autonomic computing and big data processing paradigms to endow big services withself‐* andparallel processingcapabilities. The proposed management framework takes advantage of the well‐known MapReduce programming model and Apache Spark and manages big service's related data usingknowledge graph technology. We also define ascalable embedding modelthat allows processing and learning latent big service knowledge in a distributed manner. Finally, acooperative decision mechanismis defined to trigger non‐conflicting management policies in response to the captured deviations of the running big service. Big services' management tasks (monitoring, embedding, and decision), as well as the core modules (autonomic managers' controller, embedding module, and coordinator), are implemented on top of Apache Spark as MapReduce jobs, while the processed data are represented as resilient distributed dataset (RDD) structures. To exploit the shared information exchanged between the workers and the master node (coordinator), and for further resolution of conflicts between management policies, we endowed the proposed framework with a lightweight communication mechanism that allows transferring useful knowledge between the running map‐reduce tasks and filtering inappropriate intermediate data (e.g., conflicting actions). The experimental results proved the increased quality of embeddings and the high performance of autonomic managers in a parallel and cooperative setting, thanks to the shared knowledge.

Abstract Recent years have witnessed the emergence ofbig services, as a large‐scale big data‐centric service model, that resulted from the synergy between powerful computing paradigms (big data processing, service and cloud computing, Internet of Things, etc.). Big services are seen as a heterogeneous combination of physical and virtualized domain‐specific resources, with a huge volume of data and complex functionalities, all encapsulated and offered as services. This complexity of big services (composition units' heterogeneity, cross‐domain orientation, data massiveness), coupled with other environmental factors (cloud dynamicity, providers' policies, customer requirements) makes their management tasks beyond humans' capability. Therefore, endowing big service ecosystems with self‐adaptive behavior is a natural solution. To achieve this goal, this article models big services asautonomic computing systems, and structures their behavioral aspects (functional behavior, quality of service/data levels, management policies) as amulti‐view knowledge graph. To infer useful knowledge (e.g., conflicts between policies) for the autonomic big service's management tasks, we process the big service's knowledge graph (BSKG) via agraph neural network‐based graph embedding model. This latter is reinforced by anincremental learningmethod, that helps capturing the big services' frequent changes (e.g., QoS deviations, service failures, new policies), and drives autonomic managers to continuously update and enrich their knowledge w.r.t. the managed big service's current state. Finally, a flexibledecision mechanismexplores the BSKG structure and the latent knowledge, to locate and trigger the appropriate management policies, according to the big service's produced events.

Big services have recently emerged from the synergy between big data and cloud computing paradigms. This new big data-centric service model aims to provide customer-oriented massive services by combining both physical and virtualized resources from different domains. Although such complex ecosystem is able to process, encapsulate and offer huge volumes of data as services, its management operations are beyond the ability of human administrators, due to several challenges including the big services’ large-scale nature and complexity, the heterogeneity of their components (e.g., services, data sources, connected things), the dynamicity and uncertainty of their hosting cloud environments. To cope with the lack of understanding regarding big services capabilities, we propose to describe them using a novel meta-model for the quality of big services (QoBS). We also take advantage of a recent technology called knowledge graphs, to represent the big service information (service descriptions, services’ and data sources’ quality levels, management policies) as a heterogeneous information network. Finally, a multi-view representation learning approach is proposed to infer additional knowledge regarding big services capabilities.

Cloud Computing, as a distributed computing paradigm, consists of the provisioning of infrastructure, platform, and software resources as services. This paradigm is being increasingly used for the deployment and execution of service-based business processes. To efficiently manage them according to the autonomic computing paradigm, service-based business processes can be associated with autonomic managers that monitor these processes, analyze monitoring data, plan configuration actions, and execute these actions on these processes. Although, during these last years, autonomic management of cloud services has received increasing attention, the optimization of autonomic managers to be assigned to cloud services remains not well explored. In fact, almost all the existing solutions on autonomic computing have been interested in modeling and implementing autonomic mechanisms without making any effort to optimize the number of used autonomic managers. Moreover, when it comes to large service-based business processes, optimization of management resources becomes a critical issue. To overcome this issue, we present in this paper a novel approach to determine how many autonomic managers to use for the management of large service-based business processes in order to minimize their cost while avoiding management bottlenecks. Experiments conducted on three different types of datasets highlight the effectiveness of our approach.

To deliver the required QoS, the cloud provider is asked to efficiently execute the tenants’ tasks and manages a huge amount of distributed and shared data. Hence, task scheduling and data replication are interdependent techniques that can improve the overall system performance and guarantee efficient data accessing. These operations must also preserve the economic profit of the cloud provider, which is very challenging. In this paper, we present a novel combination between a scheduling algorithm called Bottleneck Value Scheduling (BVS) algorithm with a dynamic data replication strategy called Correlation and Economic Model-based Replication (CEMR). Our aim is to improve data access effectiveness in order to meet service level objectives in terms of response time S LORT and minimum availability S LOMA, while preserving the provider profit. Simulation results demonstrate that the proposed scheduling and replication strategies offer better performance compared to existing strategies.

Nowadays with the widespread of cloud computing, a large number of complex networked applications are being deployed in data center networks. The performance and reliability of any particular application may depend on multiple services, spanning many hosts and networks components [8]. In such settings, dynamic discovery of dependencies among services and applications has been recognized as a highly important issue in order to efficiently carry out a range of critical management tasks including fault localization, reconfiguration planning, impact analysis, etc. Previous studies on application and service dependencies mining mainly concern with two specific historical data: i) network level information i.e., packet trace data or ii) application level information. Based on a mining with regard to these two aspects, various works have been proposed. Unfortunately, almost all studies either rely on the former data or on the latter one separately, thus unable to accurately and effectively mine application and service dependencies. The contribution of this paper is twofold. In the first part, we thoroughly study the literature review of discovery of dynamic service dependencies in service oriented systems based on data mining techniques. Furthermore, we introduce SCoRMiner(1), a novel approach to the extraction of service dependencies that applications rely on that considers both network level packet trace data and application level information simultaneously. Unlike almost previous works, which discover service dependencies using a single large graph, relating all network services without considering their application contexts, we construct a graph data base composed by a set of Service Dependency Graphs of running applications. According to the requirement, the graph mining process could be initiated at a single application graph level or at the whole application graph database level. SCoRMiner is non-intrusive approach which processes existing monitoring data generated by server platforms without any additional application or system modification.

Replicating data through different storage elements in order to facilitate their access is a main feature in data grids. This helps reducing execution time and bandwidth consumption, ensuring load balancing, and increasing data availability and service quality. In this regard, several replication strategies have been proposed in the literature. In this work, a new evaluation metric called DISQUEV is proposed and experimentally evaluated. This metric will serve to tackle the evaluation of the impact of a replication strategy on the data distribution quality in the grid. The data distribution is indeed proven to have a close influence on the performances of the grid not only for its short term uses but also for the long term.

Enterprises are more and more using hybrid cloud environments to deploy and run applications. This consists in providing and managing software and hardware resources within the enterprise and getting additional resources provided externally by public clouds whenever this is needed. In this later case, deployment of new applications consists in choosing a placement of some components in the private cloud and some others in the public cloud. To tackle this NP-hard problem, we have proposed in a previous work an approximate approach based on communication and hosting costs induced by the deployment of components in the public cloud. In this paper, we go further and propose a new efficient algorithm adapted for service-based applications modelled that can be not only described as behavior-based but also as architecture-based compositions of services.

Data replication in data grids is an efficient technique that aims to improve response time, reduce the bandwidth consumption and maintain reliability. In this context, a lot of work is done and many strategies have been proposed. Unfortunately, most of existing replication techniques are based on single file granularity and neglect correlation among different data files. Indeed, file correlations become an increasingly important consideration for performance enhancement in data grids. In fact, the analysis of real data intensive grid applications reveals that job requests for groups of correlated files and suggests that these correlations can be exploited for improving the effectiveness of replication strategies. In this paper, we propose a new dynamic periodic decentralized data replication strategy, called RSBMFCP (1), which consider a set of correlated files as granularity. Our strategy gathers files according to a relationship of simultaneous accesses between files by jobs and stores correlated files at the same site. In order to find out these correlations, a maximal frequent correlated pattern mining algorithm of the data mining field is introduced. We choose the all-confidence as correlation measure. The proposed strategy consists of four steps: storing file access history, converting the file access history into a logical history file, applying maximal frequent correlated pattern mining algorithm and performing replication and replacement. Experiments using the well-known data grid simulator Opt or Sim show that our proposed strategy has better performance in comparison with other strategies in terms of job execution time and effective network usage.