مدل سازی چگونگی تکامل منابع و خدمات برای تطبیق زمان اجرای موثر
Abstract
1- Introduction
2- Related work
3- Overview of the proposed approach
4- Triggering time for runtime adaptations based on resources and services evolvement
5- Evaluation
6- Conclusions
References
Abstract
Regardless of their implementation aspects and distribution elements, i.e. centralized or distributed, service-based environments such as cloud computing and edge/fog infrastructures, enable the provisioning of services addressing a wide range of application domains. The key requirement for users and consumers of such services refers to the corresponding levels of quality, which is affected both by the real-world dynamics – given the non-deterministic use of services, and by the underlying resources state – given the typically virtualized sharing nature of the resources. In this paper, an approach is presented that aims at estimating the evolvement of services and resources state in order to provide insights for runtime adaptations, as required to ensure services quality. The state refers to different metrics/parameters such as memory, number of users, throughput, etc, and can be extended and applied to different ones. The proposed approach exploits polynomial regression and prediction to identify the aforementioned state evolvement by mapping the two first monitoring data points for each metric/parameter to the corresponding function that depicts their evolvement. The latter provides added value in different cases, including among others the adaptation of monitoring time intervals, the estimation of the potential breach of quality thresholds, and the prediction of the time for runtime adaptations and scaling decisions. The effectiveness of the implemented approach is demonstrated and evaluated through a set of different scenarios.
Introduction
The continuous-changing landscape of the services provisioning space as well as the realization of advanced communication and networking paradigms - such as SDN/NFV and 5G, drives the emergence of new holistic environments. While these holistic environments integrate the aforementioned advanced communication paradigms, they also exploit computing infrastructures such as cloud, edge and fog, in order to provide added-value services to a wide set of users and consumers. Currently, service providers go beyond the SPI model (Software, Platform, and Infrastructure as a Service) [1], aiming at the delivery of different assets as a service. Furthermore, new patterns of mobility and the wide deployment of Internet of Things (IoT) environments contribute towards the compilation of new services and products through edge / fog computing models. In this context, cloud infrastructures actually reflect a baseline utility for any IT-based service delivery environment. As a result, the infrastructure space of emerging service-based environments includes different computing, storage and communication elements that serve the needs of applications and users in a holistic way, expressed as “complete computing” [2]. On the applications space, the aforementioned communication and IoT environments act as enablers for the provision of added-value services in combination with data management and computing “backbone” infrastructures. All in one, they allow the realization and offerings of composite applications that consist of application service components (i.e. micro-services) – often of different nature. These application service components provide specific functionality, contributing to the overall application’s one, and may be offered by different providers. It has to be noted that this composite application paradigm is also applied across the service stack since besides service components (on the application layer), different infrastructure services (e.g., networking or storage resources) may also be offered within the overall applications by exploiting the concept of containerization [3].