| Abstrakt: |
Datacenters are the primary driving force behind the energy shift to renewable sources. This shift limits the environmental impact of non-renewable sources and decarbonizes datacenters. However, renewable energy generation fluctuates with time and location, which introduces uncertainty in fulfilling user requests (URs). Thus, non-renewable energy generation continues to power the datacenters to make stability. Recent works direct to the use of renewable energy generation followed by non-renewable generation while assigning the URs to the resources of the datacenters. These works do not model the uncertainty of renewable and non-renewable energy resources and the level of uncertainty. This paper extends the three benchmark algorithms, namely future-aware best-fit (FABEF), highest available renewable energy first (HAREF) and round-robin (RR), by incorporating uncertainty and its level (UNL), and we call them UNL-FABEF, UNL-HAREF and UNL-RR, respectively. The goal of UNL-FABEF is to minimize the overall cost, whereas UNL-HAREF is to maximize the available renewable energy usage. On the contrary, UNL-RR assigns the datacenter to the URs in a roundabout fashion. This paper also introduces the UNL multi-objective scheduling algorithm (UNL-MOSA) to make a trade-off between UNL-FABEF and UNL-HAREF. UNL-MOSA creates a balance between the overall cost and the available renewable energy usage. All four algorithms consider three UNLs of URs, namely low, medium and high, for renewable energy resources. These algorithms are tested using fifty instances of ten datasets with 200 to 2000 URs and 20 to 200 datacenters and compared using five performance metrics: the overall cost, number of used renewable and non-renewable energy resource slots, and uncertainty cost and time. The performance of four algorithms in these performance metrics is extensively examined to know their applicability. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink