The numerous technological advances in data acquisition techniques allow the massive production of enormous amounts of data in diverse fields such as astronomy, health and social networks. Nowadays, only a small part of this data can be analysed because of the lack of computational resources. High Performance Computing (HPC) strategies represent the single choice to analyse such overwhelming amount of data. However, in general, HPC techniques require the use of big and expensive computing and storage infrastructures, usually not affordable or available for most users.
Cloud computing, where users pay for the resources they need and when they actually need them, appears as an interesting alternative. Besides the savings in hardware infrastructure, cloud computing offers further advantages such as the removal of installation, administration and supplying requirements. In addition, it enables users to use better hardware than the one they can usually afford, scale the resources depending on their needs, and a greater fault-tolerance, amongst others. The efficient utilisation of HPC resources becomes a fundamental task, particularly in cloud computing. We need to consider the cost of using HPC resources, specially in the case of cloud-based infrastructures, where users have to pay for storing, transferring and analysing data. Therefore, it is really important the usage of generic tasks scheduling and auto-scaling techniques to efficiently exploit the computational resources. It is equally important to make these tasks user-friendly through the development of tools/applications (software clients), which act as interface between the user and the infrastructure.
Buscar en Dimension