H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS Energy-Efficient Management of Virtual Machines in Data Centers for Cloud Computing Anton Beloglazov Supervisor: Prof. Rajkumar Buyya The Cloud Computing and Distributed Systems (CLOUDS) Lab CIS Department, The University of Melbourne PhD Completion Seminar
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS C LOUD D ATA C ENTERS ◮ Delivering computing resources on-demand over the Internet ◮ Hundreds of thousands of servers worldwide ◮ Amazon EC2 2012 Google’s data center [Google, 2012] ◮ 450,000 servers [Liu, 2012] 300 ◮ 9 regions 250 Billion kWh/year 200 ◮ High energy consumption and 150 CO 2 emissions [Koomey, 2011] 100 ◮ 2005-2010: 56% increase in 50 energy consumption 0 ◮ 2% of global CO 2 emissions 2000 2005 2010 Year [Gartner, 2007] Worldwide data center energy consumption 2000-2010 [Koomey, 2011] 2 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS C LOUD D ATA C ENTERS ◮ Delivering computing resources on-demand over the Internet ◮ Hundreds of thousands of servers worldwide ◮ Amazon EC2 2012 Google’s data center [Google, 2012] ◮ 450,000 servers [Liu, 2012] 300 ◮ 9 regions + Sydney (2012)! 250 Billion kWh/year 200 ◮ High energy consumption and 150 CO 2 emissions [Koomey, 2011] 100 ◮ 2005-2010: 56% increase in 50 energy consumption 0 ◮ 2% of global CO 2 emissions 2000 2005 2010 Year [Gartner, 2007] Worldwide data center energy consumption 2000-2010 [Koomey, 2011] 3 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS S OURCES OF E NERGY W ASTE 1. Infrastructure efficiency ◮ Facebook’s Oregon data center PUE = 1.08 [Open Compute, 2012] ◮ 91% of energy is consumed by the computing resources 2. Resource utilization ◮ Average CPU utilization: < 50% [Barroso, 2007] ◮ Low server dynamic power range: 30% Server power consumption depending [Fan, 2007] on the CPU utilization [Fan, 2007] 4 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS S OURCES OF E NERGY W ASTE 1. Infrastructure efficiency ◮ Facebook’s Oregon data center PUE = 1.08 [Open Compute, 2012] ◮ 91% of energy is consumed by the computing resources 2. Resource utilization ◮ Average CPU utilization: < 50% [Barroso, 2007] ◮ Low server dynamic power range: 30% Server power consumption depending [Fan, 2007] on the CPU utilization [Fan, 2007] Solution – sleep mode! 450 W → 10 W in 300 ms 5 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS A T AXONOMY OF E NERGY -E FFICIENT C OMPUTING Power Management Techniques Static Power Management (SPM) Dynamic Power Management (DPM) Hardware Level Software Level Software Level Hardware Level [Devadas 1995], [Ong 1994], [Benini 2000] [Venkatachalam 2005] [Givargis 2001] Single Server Multiple Servers, Data Circuit Level Logic Level Architectural Level Centers and Clouds OS Level Virtualization Level [Pallipadi 2006] [Wei 2009], [Stoess 2007] 6 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS A T AXONOMY OF E NERGY -E FFICIENT C OMPUTING Power Management Techniques Static Power Management (SPM) Dynamic Power Management (DPM) Hardware Level Software Level Software Level Hardware Level [Devadas 1995], [Ong 1994], [Benini 2000] [Venkatachalam 2005] [Givargis 2001] Single Server Multiple Servers, Data Circuit Level Logic Level Architectural Level Centers and Clouds OS Level Virtualization Level [Pallipadi 2006] [Wei 2009], [Stoess 2007] 7 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS D YNAMIC C ONSOLIDATION OF V IRTUAL M ACHINES ◮ Adjusts the number of User User User active hosts according to VM provisioning SLA negotiation Application requests the resource demand Global resource managers ◮ Improves the power Consumer, scientific and business proportionality Virtual applications machines ◮ 2 basic processes and user applications ◮ VM consolidation ◮ VM deconsolidation Virtualization layer (VMMs, local resource managers) ◮ Nathuji 2007 Raghavendra 2008 Physical compute Verma 2008 nodes Kusic 2009 Power On Power Off Hermenier 2009 8 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS I NFRASTRUCTURE AS A S ERVICE – P ROPERTIES 1. Large scale ◮ Amazon EC2: ≈ 450,000 servers ◮ Rackspace: ≈ 85,000 servers ◮ Scalability and fault-tolerance are required 2. Multiple independent users ◮ On-demand VM provisioning ◮ Full access and permissions ◮ VM provisioning time is unknown 3. Unknown mixed workloads ◮ Web, HPC applications ◮ The provider is unaware of the application workloads 4. Quality of Service (QoS) guarantees ◮ Currently, the performance in IaaS is not guaranteed ◮ Existing metrics: availability, response time, deadlines ◮ Workload independent QoS are required 9 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS R ESEARCH Q UESTIONS 1. How to define workload-independent QoS requirements? 2. When to migrate VMs? 3. Which VMs to migrate? 4. Where to migrate the VMs selected for migration? 5. When and which physical nodes to switch on/off? 6. How to provide scalability and fault-tolerance? 10 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS T HESIS C ONTRIBUTIONS 1. A taxonomy and survey of energy-efficient computing ◮ Advances in Computers 2011 2. Competitive analysis of dynamic VM consolidation ◮ CCPE 2012 3. Novel heuristics for dynamic VM consolidation ◮ FGCS 2012, CCPE 2012 4. The Markov host overload detection algorithm ◮ TPDS 2013 5. A software framework for dynamic VM consolidation ◮ SPE 2013 (in prep.) 11 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS T HESIS C ONTRIBUTIONS 1. A taxonomy and survey of energy-efficient computing ◮ Advances in Computers 2011 2. Competitive analysis of dynamic VM consolidation ◮ CCPE 2012 3. Novel heuristics for dynamic VM consolidation ◮ FGCS 2012, CCPE 2012 4. The Markov host overload detection algorithm ◮ TPDS 2013 5. A software framework for dynamic VM consolidation ◮ SPE 2013 (in prep.) 12 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS O UTLINE H EURISTICS Distributed Approach Workload Independent QoS Dynamic VM Consolidation Heuristics M ARKOV H OST O VERLOAD D ETECTION Problem Definition The Optimal Offline Algorithm Markov Host Overload Detection (MHOD) Algorithm I MPLEMENTATION Framework for Dynamic VM Consolidation Experimental Evaluation C ONCLUSIONS Summary and Future Directions 13 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS O UTLINE H EURISTICS Distributed Approach Workload Independent QoS Dynamic VM Consolidation Heuristics M ARKOV H OST O VERLOAD D ETECTION Problem Definition The Optimal Offline Algorithm Markov Host Overload Detection (MHOD) Algorithm I MPLEMENTATION Framework for Dynamic VM Consolidation Experimental Evaluation C ONCLUSIONS Summary and Future Directions 14 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS D ISTRIBUTED A PPROACH : 4 S UB - PROBLEMS 1. Host underload detection 2. Host overload detection 3. VM selection 4. VM placement 15 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS D ISTRIBUTED A PPROACH : 4 S UB - PROBLEMS 1. Host underload detection 2. Host overload detection 3. VM selection 4. VM placement Scalability and fault-tolerance → distribution and replication 16 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS O UTLINE H EURISTICS Distributed Approach Workload Independent QoS Dynamic VM Consolidation Heuristics M ARKOV H OST O VERLOAD D ETECTION Problem Definition The Optimal Offline Algorithm Markov Host Overload Detection (MHOD) Algorithm I MPLEMENTATION Framework for Dynamic VM Consolidation Experimental Evaluation C ONCLUSIONS Summary and Future Directions 17 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS O VERLOAD T IME F RACTION (OTF) ◮ u t – the CPU utilization threshold distinguishing the non-overload and overload states of a host ◮ t o – the time, during which the OTF ( u t ) = t o ( u t ) host has been overloaded, which t a is a function of u t ◮ t a – the total time, during which the host has been active 18 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS A GGREGATE O VERLOAD T IME F RACTION (AOTF) ◮ u t – the CPU utilization threshold distinguishing the non-overload and overload states of a host ◮ H – the set of compute hosts � t o ( h , u t ) AOTF ( u t ) = ◮ h – a compute host t a ( h ) h ∈H ◮ t o ( h , u t ) – the overload time of the host h , which is a function of u t ◮ t a ( h ) – the total activity time of the host h 19 / 70
H EURISTICS M ARKOV H OST O VERLOAD D ETECTION I MPLEMENTATION C ONCLUSIONS O UTLINE H EURISTICS Distributed Approach Workload Independent QoS Dynamic VM Consolidation Heuristics M ARKOV H OST O VERLOAD D ETECTION Problem Definition The Optimal Offline Algorithm Markov Host Overload Detection (MHOD) Algorithm I MPLEMENTATION Framework for Dynamic VM Consolidation Experimental Evaluation C ONCLUSIONS Summary and Future Directions 20 / 70
Recommend
More recommend
