Xempupil performance-aware power capping orchestrator for the Xen hypervisor
Data di pubblicazione
Data di priorità
Politecnico di Milano
DIPARTIMENTO DI ELETTRONICA, INFORMAZIONE E BIOINGEGNERIA
Marco Santambrogio; Marco Arnaboldi; Matteo Ferroni
In the era of Cloud Computing, services and computational power are provided in an as a Service (aaS) fashion, reducing the need of buying, building and maintaining proprietary systems. In the last few years, many services moved from being proprietary to the aaS paradigm: this, together with virtualization techniques, allows multiple applications to easily run on the same machine. However, the burden of costs optimization is left to the Cloud Provider that still faces the problem of consolidating multiple workloads on the same infrastructure. As power consumption remains one of the most impacting costs of any digital system, several approaches have been explored in literature to cope with power caps, trying to maximize the performance of the hosted applications.
We propose a methodology for hardware and software power capping in virtualized environment. It is hybrid as it makes use of the RAPL hardware interface to set a strict limit on the processor’s power consumption, while a software level ODA loop structure performs an exploration of the available resource allocations, to find the most power efficient one for the running workload. Of course, the innovation does not lie in the exploitation of the well-known ODA loop structure, but in the adoption of a hybrid power capping approach in a virtualized environment. An overview of the system is presented in Figure 1.
The proposed methodology allows either to maximize the performance of the workloads running in the system while respecting a given power limit or minimizing the power consumption while respecting a given SLA (Service Level Agreement) on the running workloads.
Campo di applicazione
This patent application targets datacenters, in particular datacenters owners, seeking for an effective methodology in order to respect power budget constraints on the:
- whole infrastructure;
- single clusters;
- single machines.
The methodology main advantage is the adoption of a hybrid power capping approach in a virtualized environment that permits to maintain standard performances with low power consumption. Such approach can be applied also to any other hypervisor with the main goal of trying to maximize the performance of the running workloads in the system and at the same time strictly enforcing and respecting a defined power budget.
Stadio di sviluppo