Adaptive learning-based resource management strategy in fog-to-cloud
- Autores: Souvik Sengupta
- Directores de la Tesis: Jordi García Almiñana (dir. tes.), Xavier Masip Bruin (codir. tes.)
- Lectura: En la Universitat Politècnica de Catalunya (UPC) ( España ) en 2020
- Idioma: español
- Materias:
- Texto completo no disponible (Saber más ...)
- Resumen
Technology in the twenty-first century is rapidly developing and driving us into a new smart computing world, and emerging lots of new computing architectures. Fog-to-Cloud (F2C) is among one of them, which emerges to ensure the commitment for bringing the higher computing facilities near to the edge of the network and also help the large-scale computing system to be more intelligent. As the F2C is in its infantile state, therefore one of the biggest challenges for this computing paradigm is to efficiently manage the computing resources. Mainly, to address this challenge, in this work, we have given our sole interest for designing the initial architectural framework to build a proper, adaptive and efficient resource management mechanism in F2C.
F2C has been proposed as a combined, coordinated and hierarchical computing platform, where a vast number of heterogeneous computing devices are participating. Notably, their versatility creates a massive challenge for effectively handling them. Even following any large-scale smart computing system, it can easily recognize that various kind of services is served for different purposes. Significantly, every service corresponds with the various tasks, which have different resource requirements.
So, knowing the characteristics of participating devices and system offered services is giving advantages to build effective and resource management mechanism in F2C-enabled system. Considering these facts, initially, we have given our intense focus for identifying and defining the taxonomic model for all the participating devices and system involved services-tasks.
In any F2C-enabled system consists of a large number of small Internet-of-Things (IoTs) and generating a continuous and colossal amount of sensing-data by capturing various environmental events. Notably, this sensing-data is one of the key ingredients for various smart services which have been offered by the F2C-enabled system. Besides that, resource statistical information is also playing a crucial role, for efficiently providing the services among the system consumers. Continuous monitoring of participating devices generates a massive amount of resource statistical information in the F2C-enabled system.
Notably, having this information, it becomes much easier to know the device's availability and suitability for executing some tasks to offer some services. Therefore, ensuring better service facilities for any latency-sensitive services, it is essential to securely distribute the sensing-data and resource statistical information over the network. Considering these matters, we also proposed and designed a secure and distributed database framework for effectively and securely distribute the data over the network.
To build an advanced and smarter system is necessarily required an effective mechanism for the utilization of system resources.
Typically, the utilization and resource handling process mainly depend on the resource selection and allocation mechanism. The prediction of resources (e.g., RAM, CPU, Disk, etc.) usage and performance (i.e., in terms of task execution time) helps the selection and allocation process. Thus, adopting the machine learning (ML) techniques is much more useful for designing an advanced and sophisticated resource allocation mechanism in the F2C-enabled system. Adopting and performing the ML techniques in F2C-enabled system is a challenging task. Especially, the overall diversification and many other issues pose a massive challenge for successfully performing the ML techniques in any F2C-enabled system. Therefore, we have proposed and designed two different possible architectural schemas for performing the ML techniques in the F2C-enabled system to achieve an adaptive, advance and sophisticated resource management mechanism in the F2C-enabled system. Our proposals are the initial footmarks for designing the overall architectural framework for resource management mechanism in F2C-enabled system.
