Title :
PCFS: Power Credit Based Fair Scheduler Under DVFS for Muliticore Virtualization Platform
Author :
Wen, Chengjian ; He, Jun ; Zhang, Jiong ; Long, Xiang
Author_Institution :
Sch. of Comput. Sci. & Technol., Beihang Univ., Beijing, China
Abstract :
In the area of system architecture, there are two most significant trends: multi-core and system virtualization technology. Both of them have quite close relationship with energy efficient computing. Industry turns to integrate more cores on a single chip instead of increasing frequency to solve the heat problem. Meanwhile system virtualization could decrease the total power consumption by sharing the same platform among different operating systems. So the necessity of power management on the multi-core virtualization platform has become increasingly evident. However, traditional virtual machine monitor schedulers could not make efficient use of DVFS, and thus could not take it into considering that the guest OSes may run at different frequency. In order to address this problem, this paper designs a power efficient scheduler which uses the load and the power level of the guest OSes as feedback, and implements a prototype based on Xen virtual machine monitor. This scheduler allocates power credit to VCPU of each guest OS, accounts the power consumption of VCPU sat different speed levels and makes scheduler decision by integrating it to the credit oriented to CPU time slice sharing. It also uses utilization of processors as feedback and sets frequency according to the load change trends instead of the simple static relationship between load and frequency policies, so as to decrease the speed steps required by response to burst load change. Experiment results show that the scheduling fairness of guest OS improved when using DVFS as main power saving method, and the power consumption of the whole system can be reduced by 5 percent - 30 percent. Therefore, this framework for feedback scheduling could make efficient use of varieties of power saving methods and maintain ideal balance between overall system power saving and single core over heat.
Keywords :
computerised monitoring; multiprocessing systems; operating systems (computers); power aware computing; scheduling; virtual machines; virtualisation; CPU time slice sharing; DVFS; Xen virtual machine monitor; energy efficient computing; guest OS; multicore virtualization platform; power consumption; power credit based fair scheduler; power management; system virtualization technology; virtual machine monitor scheduler; Heating; Multicore processing; Power demand; Processor scheduling; Temperature measurement; Topology; Virtual machine monitors; DVFS; Multicore; Xen; fair scheduling; power management; virtual machine monitor;
Conference_Titel :
Green Computing and Communications (GreenCom), 2010 IEEE/ACM Int'l Conference on & Int'l Conference on Cyber, Physical and Social Computing (CPSCom)
Conference_Location :
Hangzhou
Print_ISBN :
978-1-4244-9779-9
Electronic_ISBN :
978-0-7695-4331-4
DOI :
10.1109/GreenCom-CPSCom.2010.126