Energy efficient storage in vm

Energy-Efficient Storage in
Virtual Machine Environments

Lei Ye, Gen Lu, Sushanth Kumar,
Chris Gniady, John H. Hartman
Department of Computer Science
University of Arizona

Energy is Critical

App App App App App App
… … …
OS1 OSn OS1 OSn OS1 OSn
VMM VMM VMM

 Energy consideration is critical for system design
 Virtualization is becoming common in all platforms
 Integration of virtualization and energy management is critical
for overall system efficiency

Xen Summit at AMD April 28-29, 2010 1

Current Environment for Energy Management

Applications

Operating
Systems

Physical
Hardware

 OS is the central point for energy management
 Global views of application execution
 OS directly interacts with the hardware
 Detailed hardware knowledge allows sophisticated energy
management


Abstraction in Virtual Machines

Virtual Machine Monitor

 Where should we manage energy?
 OS only sees the virtualized hardware
 VMM is unaware of the application execution
 Separation of hardware and OS makes energy management a
challenge

Cross-layer Energy Optimization
 Goals
 Preserve VM abstractions
 Adapt current energy optimization

 Approach
 Shape I/O accesses at VMM
 Provide hints from VMM to the VMs

 Outcome
 Integrate existing disk energy management into VM


Outline
 Introduction
 Existing Mechanisms
 Design
 Evaluation
 Conclusion


Existing Disk Energy Management
Last I/O New I/O
Spin-down

I/O Idle sleep I/O

Spin-up
Timeout Timeout
starts expires
 Goal: Maximize length of sleeping periods
 Must exceed break-even time
 Approach: Shaping write activities
 Consequence:
 Maximize disk off time
 Reduce disk spin-ups


Buffer Cache

File Buffer
Applications
System Cache

Linux Kernel
 Function
 Store data temporarily (default 30 seconds)
 Perform data transfer for block device
 Advantage
 Trade disk accesses for memory accesses
 Disadvantage
 Loss of buffered data in case of system crash


Buffer Cache Complicates Timeout

App I/O Buffer cache Data delivered
to buffer cache flush to disk
30 seconds

I/O Idle Idle sleep I/O

Spin-up
Timeout Timeout
starts expires

 Timeout might occur when there are pending writes
 Disk spin-up might occur even in absence of application I/O


Early Flush

OS
I/O Idle Buffer Flush
Idle Idle Idle
Early Flush

Physical
Disk state
I/O Idle I/O sleep

 Do buffer flushes prior to shutdown
 No impact on desired I/O reliability


Extended Buffering

OS
Idle Buffer Flush
Idle Idle
Buffer Flush I/O Idle

30 seconds

Physical Buffer being flushed
Disk state
sleep sleep I/O sleep

Extended idle time

 Can extend idle time under writes
 May impact I/O reliability


Buffer Flushes in Virtual Machines
VM1
I/O Idle Idle Buffer Flush Idle Idle IdleIdle
VM2
I/O Idle I/O Idle Buffer Flush Idle

Physical
Disk state
I/O Idle I/O Idle I/O Idle
sleep sleep sleep
 Guest OS is unaware of physical disk state
 Buffer flushes from VMs are unsynchronized


Early Flush in Virtual Machines
VM1
I/O Idle Idle
Buffer Flush Idle Idle Idle Idle
Idle
VM2
I/O Idle Buffer Flush
Idle Idle Idle
Early Flush Notification
Physical
Disk state
I/O Idle I/O sleep

 Hints signal disk shutdown and trigger flush
 Disk spin-ups avoided and disk idle time maximized


Buffering in Virtual Machine Monitor
 Extend disk idle time
 Buffer writes from VMs
 Only when the disk is sleeping
 Only one copy of data
 Buffer is flushed when:
 VMM buffer timeout (30 seconds)
 Read request from a VM
 Reads and sync from VMs are processed immediately
 Impact I/O reliability


Evaluation Methodology
 Trace collection
 8 concurrent VMs as remote desktops on Xen
 Users interact with VMs remotely via VNC
 Rsyslog: timestamps of I/O activity, I/O type, access size,
domain identifier, process identifier, process name, and file
inode
 Traces: 12 hours per VM
 Common interactive applications
 Disk drive:
 WD2500JD: High performance drive, 9 s. long spin-up time


Disk I/O Trace Statistics

Number of Number of Number of Number of Total Idle
VMs Reads Writes Idle Periods Time [s]
1 62951 1339 255 40071
2 64886 17031 435 37470
4 83050 59407 601 31263
8 160049 103840 566 20450

 Configuration
 1 VM – 1 user
 2 VMs – 2 concurrent users


Distribution of Idle Periods

1 VM 2 VMs 4 VMs 8 VMs
700
600
Number of Idle Periods

500
400
300
200
100
0
2 4 8 16 32 64 128
Time [Break-Even Period]

Energy Consumption

S-Standard, B-Buffering, F-Early Flush, BF-Buffering&Early Flush
500
Sleep
Energy Consumption (KJoules)

400 Power-Cycle
Active
300

200

100

0
S B F BF S B F BF S B F BF S B F BF


Execution Time

14
Spin-up Idle > BE Active
12
Execution Time (Hours)

10
8
6
4
2
0



Reducing Spin-ups
600
Buffered Write
500 Write
Number of Spin-ups

400 Read

300

200

100

0


Energy Delay Product

1.2
Normalized Energy Delay Product

1

0.8

0.6

0.4

0.2

0


Buffering in VMM with Immediate Sleep
VM1
Idle Buffer Flush Idle
VM2 Writes being buffered
Idle Buffer Flush Idle
VMM Writes being buffered
Delayed Writes
Physical
30 seconds Buffer being flushed
Disk state
sleep sleep I/O sleep

Extended idle time
 The disk is put to sleep immediately after a buffer flush
 VMM buffer flush independent from VM I/Os

Immediate Sleep Optimization

B-Buffering, BF-Buffering&Early Flush, T-Timeout, I-Immediate
400
Energy Consumption (KJoules)

Sleep
300 Power-Cycle
Active
200

100

0
T I T I T I T I T I T I T I T I
B BF B BF B BF B BF


Writes in Buffering and Early Flush
VMs Mechanisms Write Delay Delayed Writes Early Flush
Participants
1 Buffering 14.5s 3.5 N/A
Early Flush N/A N/A 1.0
Buffering & Early Flush 10.6s 1.9 1.0
 Buffering in the VMM can result in additional delays
 The longer the write resides the buffer, the higher the impact
on reliability

Conclusion
 Existing energy optimizations require modification to be
effective in a VM environment
 OS only sees the virtualized hardware
 VMM is unaware of the application execution

 Early Flush reduces spin-ups due to buffer flushes
 Reduce energy by 10.5% compared with standard case in 8 VMs

 VMM buffering extends disk idle time by delaying VM writes
 Reduce energy by 7.7% compared with standard case in 8 VMs

 Combination reduces energy consumption by 14.8% in 8 VMs


Q&A

Thank You


Energy efficient storage in vm

More Related Content

What's hot (20)

Similar to Energy efficient storage in vm (20)

More from The Linux Foundation (20)

Recently uploaded (20)

Energy efficient storage in vm