Load Index Characterization: The Scientific Challenge, and a Survey Approach

Cloud computing configurations became the major facility to support execution and storage of different types of applications. The fact that the diversity of requirements for executions became too much complex, an understanding of load index characterization and knowledge of existing approaches lead us to research these two topics. Load balancing is crucial to enhance performance in distributed systems. The load balancing algorithms intend to distribute as equally as possible the load among all machines. In order to perform this task, they require information about the load exerted on each machine involved in the load balancing process. This information is abstracted by a load index. Thus, the efficiency of the load balancing algorithm is highly dependent on the quality of the load index. However, essential information for current distributed systems and cloud computing ‑ like power consumption ‑ are not fully represented by the current load indices. It is imperative to know what are the most relevant works and where the research is moving towards, to provide new alternatives that better represent today's load balancing goals. This research work uses a formal method to search, select and synthesize the most prominent publications in order to present a differentiated survey of the most relevant load indices available in the literature. Then, it is proposed a set of categories to classify those publications, indicating the state of art of this field.