green computing metrics: power, temperature, co2, … computing system: many-cores, clusters, grids...
Post on 13-Dec-2015
218 Views
Preview:
TRANSCRIPT
Green Computing
• Metrics: Power, Temperature, CO2, …• Computing system: Many-cores, Clusters,
Grids and Clouds• Algorithm and model: task scheduling, CFD
model, …• Middleware: auditing & insertion service,
green resource management service, …
Power aware virtual machine scheduling in a DVFS cluster
• Virtual machine in Grids and Clouds• Dynamic Voltage Frequency Scheduling • Objective: dynamically scale voltages for
virtual machines in a cluster
Virtual machines in compute cluster
vm vm
Computenode
vm
Compute node
vmFile
serverHeadnode
vmvm
job
Start a vm
Execute job in a vm
Schedule virtual machines
PE
PE
PE
PE
PE
PE
Scheduling
algorithm
VM
cluster
queue
Power aware scheduling algorithm
1. Sort VMs in a decreasing order of required CPU speed
2. Set PEs to lowest voltages3. put VMs to PEs4. If cannot accommodate, level up PE voltages5. Level down PE voltages whenever it is
possible to accommodate VMs
Simulation Results
36%
2%5%23%
34%
_x0010_0.6 GHz, 0.956 V_x0010_0.8 GHz, 1.180 V_x0010_1.0 GHz, 1.308 V_x0010_1.2 GHz, 1.436 V_x0010_1.4 GHz, 1.484 V
10 20 30 40 500%
20%
40%
60%
80%
100%
vm=100vm=200vm=300vm=400vm=500
PE Number
Nor
mal
ized
Pow
er C
onsu
mpti
on
Experimental Results
2 4 80
10
20
30
40
50
60
70
80
90
100
1.6 GHz1.867 GHz2.133 GHz2.533 GHz2.668 GHz
Number of VMs
nBen
ch In
tege
r Ind
ex
2 4 80
50
100
150
200
250
1.6 GHz1.867 GHz2.133 GHz2.533 GHz2.688 GHz
Number of VMs
Pow
er C
onsu
mpti
on (W
atts)
Thermal aware workload scheduling in data centers
• Job-temperature model• Data center resource model• Thermal aware scheduling algorithm (TASA)• Thermal aware workload scheduling
framework• Simulation
Job-temperature profile
YZ
X
Rack
Hot
air
Hot
air
Rack RackNode(x,y,z)
Data center model (1)
Data center model (2)
Task-temperature profile
Online task-temperature
calculation
RC-thermal model
Workload model
Thermal map
Datacenter model
Thermal aware workload scheduling
algorithm
Profiling tool
Monitoring service
Cooling system control
CFD model
Workload placement
Thermal aware scheduling framework
Thermal aware scheduling algorithm (TASA)
1. Get thermal field of data center2. Get compute node temperature 3. Put hottest job to coldest resources4. Predict the compute node temperature after
job execution5. If a compute node temperature > redline, set
it idle6. thermal aware backfilling when it is possible
Simulation
• Real workload logged in CCR @ Buffalo Univ.• Temperature logged • FCFS in CCR @ Buffalo Univ.• TASA• Discussion
Workload in CCR
CCR Temperature
Simulation Result (1)
• Reduce max temperature: 6 F• Reduce average temperature: 15 F• Reduce power consumption 4000 kW/h• Reduce CO2 emission 19 000 kg
Simulation Result (2)
• Response time increase 13%
Green Data Center Computing: concept
Software sensor
Auditing & Insertion service
Physical sensor
Monitoring service
CFD model
Auditing & Insertion service
Cooling system and compute resources in a data center
Thermal aw
are resource m
anagement
Workload m
odel
Command Line
InformationTask Submission
Cyberaide Shell
Authentication and Authority
Java CoG Kit
Secure Web Service
Secure Web Service
Cyberaide Portal Python Client Client Layer
MiddlewareLayer
Resource Layer
Workflow Informationcollector
Data Center Data Center
Cyberaide Green: Software achitecture
Thermal-Aware Meta Scheduler
top related