Power-Aware RAID Charles Weddle, Mathew Oldham, Jin Qian, An-I Andy - - PowerPoint PPT Presentation

▶
power aware raid
SMART_READER_LITE
LIVE PREVIEW

Power-Aware RAID Charles Weddle, Mathew Oldham, Jin Qian, An-I Andy - - PowerPoint PPT Presentation

Nov 13, 2022 167 likes •619 views

PARAID: A Gear-Shifting Power-Aware RAID Charles Weddle, Mathew Oldham, Jin Qian, An-I Andy Wang Florida St. University Peter Reiher University of California, Los Angeles Geoff Kuenning Harvey Mudd College 1 Motivation Energy

slide-1
SLIDE 1

1

PARAID: A Gear-Shifting Power-Aware RAID

Charles Weddle, Mathew Oldham, Jin Qian, An-I Andy Wang – Florida St. University Peter Reiher – University of California, Los Angeles Geoff Kuenning – Harvey Mudd College

slide-2
SLIDE 2

2

Motivation

 Energy costs are rising

 An increasing concern for servers

 No longer limited to laptops

 Energy consumption of disk drives

 24% of the power usage in web servers  27% of electricity cost for data centers  More energy  more heat  more cooling  lower

computational density  more space  more costs

 Is it possible to reduce energy consumption

without degrading performance while maintaining reliability?

PARAID: A Gear-Shifting Power-Aware RAID

slide-3
SLIDE 3

3

Challenges

 Energy

 Not enough opportunities to spin down RAIDs

 Performance

 Essential for peak loads

 Reliability

 Server-class drives are not designed for

frequent power switching

PARAID: A Gear-Shifting Power-Aware RAID

slide-4
SLIDE 4

4

Existing Work

 Most trade performance for energy

savings directly

 e.g. vary speed of disks

 Most are simulated results

PARAID: A Gear-Shifting Power-Aware RAID

slide-5
SLIDE 5

5

Observations

 RAID is configured for peak performance

 RAID keeps all drives spinning for light loads

 Unused storage capacity

 Over-provision of storage capacity  Unused storage can be traded for energy savings

 Fluctuating load

 Cyclic fluctuation of loads  Infrequent on-off power transitions can be effective

PARAID: A Gear-Shifting Power-Aware RAID

slide-6
SLIDE 6

6

Performance vs. Energy Optimizations

 Performance benefits

 Realized under heavy loads

 Energy benefits

 Realized instantaneously

slide-7
SLIDE 7

7

Power-Aware RAID

 Skewed striping for energy savings  Preserving peak performance  Maintaining reliability  Evaluation  Conclusion  Questions

PARAID: A Gear-Shifting Power-Aware RAID

slide-8
SLIDE 8

8

Skewed Striping for Energy Saving

 Use over-provisioned spare storage

 Organized into hierarchical overlapping subsets

PARAID: A Gear-Shifting Power-Aware RAID RAID 1 2 3 4 5

slide-9
SLIDE 9

9

Skewed Striping for Energy Saving

 Each set analogous to gears in automobiles

PARAID: A Gear-Shifting Power-Aware RAID RAID Gears 1 2 3 1 2 3 4 5

slide-10
SLIDE 10

10

Skewed Striping for Energy Saving

 Soft states can be reclaimed for space

 Persist across reboots

PARAID: A Gear-Shifting Power-Aware RAID RAID Soft States Gears 1 2 3 1 2 3 4 5

slide-11
SLIDE 11

11

Skewed Striping for Energy Saving

 Operate in gear 1  Disks 4 and 5 are powered off

PARAID: A Gear-Shifting Power-Aware RAID RAID Gears 1 2 3 1 2 3 4 5 Soft States

slide-12
SLIDE 12

12

Skewed Striping for Energy Saving

 Approximate the workload  Gear shift into most appropriate gear

 Minimize the opportunity lost to save power Energy ( Powered On Disks ) Workload ( Disk Parallelism ) Conventional RAID PARAID workload

PARAID: A Gear-Shifting Power-Aware RAID

slide-13
SLIDE 13

13

Skewed Striping for Energy Saving

 Adapt to cyclic fluctuating workload  Gear shift when gear utilization threshold is met

time load utilization threshold gear shift

PARAID: A Gear-Shifting Power-Aware RAID

slide-14
SLIDE 14

14

Preserving Peak Performance

 Operate in the highest gear

 When the system demands peak performance  Uses the same disk layout

 Maximize parallelism within each gear

 Load is balanced  Uniform striping pattern

 Delay block replication until gear shifts

 Capture block writes

PARAID: A Gear-Shifting Power-Aware RAID

slide-15
SLIDE 15

15

Maintaining Reliability

 Reuse existing RAID levels (RAID-5)

 Also used in various gears

 Drives have a limited number of power

cycles

 Ration number of power cycles

PARAID: A Gear-Shifting Power-Aware RAID

slide-16
SLIDE 16

16

Maintaining Reliability

 Busy disk stay powered on, idle disks stay powered off  Outside disks are role exchanged with middle disks

busy disks power cycled disks idle disks

role exchange Disk 1 Gear 1 Gear 2 Gear 3 Disk 2 Disk 3 Disk 4 Disk 5 Disk 6

PARAID: A Gear-Shifting Power-Aware RAID

slide-17
SLIDE 17

17 File system RAID PARAID block mapping Disk device driver

User space Linux kernel

Soft RAID Reliability manager Load monitor Gear manager Admin tool

Logical Component Design

PARAID: A Gear-Shifting Power-Aware RAID

slide-18
SLIDE 18

18

Data Layout

PARAID: A Gear-Shifting Power-Aware RAID Disk 1 Disk 2 Disk 3 Disk 4 Disk 5 Gear 1 RAID-5 (1-4) 8 12 ((1-4),8,12) 16 20 (16,20,_) _ Gear 2 RAID-5 1 2 3 4 (1-4) 5 6 7 (5-8) 8 9 10 (9-12) 11 12 13 (13-16) 14 15 16 (17-20) 17 18 19 20

 Resembles the data flow of RAID 1+0  Parity for 5 disks does not work for 4 disks

 For example, replicated block 12 on disk 3

slide-19
SLIDE 19

19

Data Layout

PARAID: A Gear-Shifting Power-Aware RAID Disk 1 Disk 2 Disk 3 Disk 4 Disk 5 Gear 1 RAID-5 (1-4) 8 12 ((1-4),8,12) 16 20 (16,20,_) _ Gear 2 RAID-5 1 2 3 4 (1-4) 5 6 7 (5-8) 8 9 10 (9-12) 11 12 13 (13-16) 14 15 16 (17-20) 17 18 19 20

 Cascading parity updates

 For example, updating block 8 on disk 5

slide-20
SLIDE 20

20

Update Propagation

 Up-shift propagation (e.g. shifting from 3 to

5 disks)

 Full synchronization  On-demand synchronization

 Need to respect block dependency

 Downshift propagation

 Full synchronization

PARAID: A Gear-Shifting Power-Aware RAID

slide-21
SLIDE 21

21

Asymmetric Gear-Shifting Policies

 Up-shift (aggressive)

 Moving utilization average + moving standard

deviation > utilization threshold

 Downshift (conservative)

 Modified utilization moving average + moving

standard deviation < utilization threshold

 Moving average modified to account for fewer drives

and extra parity updates

PARAID: A Gear-Shifting Power-Aware RAID

slide-22
SLIDE 22

22

Implementation

 Prototyped in Linux 2.6.5

 Open source, software RAID

 Implemented block I/O handler, monitor,

disk manager

 Implemented user admin tool to configure

device

 Updated Raid Tools to recognize PARAID

level

PARAID: A Gear-Shifting Power-Aware RAID

slide-23
SLIDE 23

23

Evaluation

 Challenges

 Prototyping PARAID  Commercial machines  Conceptual barriers  Benchmarks designed to measure peak

performance

 Trace replay  Time consuming

PARAID: A Gear-Shifting Power-Aware RAID

slide-24
SLIDE 24

24

Evaluation

multimeter USB cable client server

power supply

12v & 5v power lines power measurement probes SCSI cable crossover cable

Xeon 2.8 Ghz, 512 MB RAM 36.7 GB 15k RPM SCSI P4 2.8 Ghz, 1 GB RAM 160 GB 7200 RPM SATA RAID RAID RAID RAID RAID BOOT

PARAID: A Gear-Shifting Power-Aware RAID

 Measurement framework

slide-25
SLIDE 25

25

Evaluation

 Three different workloads using two different

RAID settings

 Web trace - RAID level 0 (2-disk gear 1, 5-disk gear 2)

 Mostly read activity

 Cello99 - RAID level 5 (3-disk gear 1, 5-disk gear 2)

 I/O-intensive workload with writes

 PostMark - RAID level 5

 Measure peak performance and gear shifting overhead

 Speed up trace playback

 To match hardware  Explore range of speed up factors and power savings

PARAID: A Gear-Shifting Power-Aware RAID

slide-26
SLIDE 26

26

Web Trace

UCLA CS Dept Web Servers (8/11/2006 – 8/14/2006)

File system: ~32 GB (~500k files)

Trace replay: ~95k requests with ~4 GB data (~260 MB unique)

PARAID: A Gear-Shifting Power-Aware RAID

0.1 0.2 0.3 0.4 0.5 0.6 1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96 hours GB/hour

slide-27
SLIDE 27

27

Web Trace Power Savings

10 20 30 40 50 60 5 10 15 20 25 30 hours watts RAID-0 PARAID-0

10 20 30 40 50 60 5 10 15 20 25 30 hours watts RAID-0 PARAID-0

PARAID: A Gear-Shifting Power-Aware RAID 64x – 60 requests/sec 128x – 120 requests/sec 256x – 240 requests/sec

10 20 30 40 50 60 5 10 15 20 25 30 hours watts RAID-0 PARAID-0

64x - 34% 128x - 28% 256x - 10% Energy Savings

slide-28
SLIDE 28

28

Web Trace Latency

0.2 0.4 0.6 0.8 1 1 10 100 1000 10000 100000 msec RAID-0 PARAID-0 0.2 0.4 0.6 0.8 1 1 10 100 1000 10000 100000 msec RAID-0 PARAID-0

PARAID: A Gear-Shifting Power-Aware RAID 256x 128x 64x

0.2 0.4 0.6 0.8 1 1 10 100 1000 10000 100000 msec RAID-0 PARAID-0

256x - within 2.7% 64x - 240% 80ms vs. 33ms

Overhead

slide-29
SLIDE 29

29

Web Trace Bandwidth

  • 20

30 80 130 180 5 10 15 20 25 30 hours MB/sec RAID-0 PARAID-0

  • 20

30 80 130 180 5 10 15 20 25 30 hours MB/sec RAID-0 PARAID-0

PARAID: A Gear-Shifting Power-Aware RAID 256x 128x 64x

20 40 60 80 100 120 140 160 180 5 10 15 20 25 30 hours MB/sec RAID-0 PARAID-0

256x - within 1.3% in high gear

Overhead

slide-30
SLIDE 30

30

Cello99 Trace

 Cello99 Workload

 HP Storage Research Labs  50 hours beginning on 9/12/1999  1.5 million requests (12 GB) to 440MB of unique blocks  I/O-intensive with 42% writes

PARAID: A Gear-Shifting Power-Aware RAID

slide-31
SLIDE 31

31

Cello99 Power Savings

10 20 30 40 50 10 20 30 40 50 hours watts RAID-5 PARAID-5

5 10 15 20 25 30 35 40 45 50 10 20 30 40 50 hours watts RAID-5 PARAID-5

PARAID: A Gear-Shifting Power-Aware RAID 128x – 1000 requests/sec 32x – 270 requests/sec 64x – 550 requests/sec

10 20 30 40 50 10 20 30 40 50 hours watts RAID-5 PARAID-5

32x - 13% 64x - 8.2% 128x - 3.5% Energy Savings

slide-32
SLIDE 32

32

Cello99 Completion Time

0.9 0.92 0.94 0.96 0.98 1 1 10 100 1000 10000 100000 msec RAID-5 PARAID-5

0.9 0.92 0.94 0.96 0.98 1 1 10 100 1000 10000 100000 msec RAID-5 PARAID-5

PARAID: A Gear-Shifting Power-Aware RAID 128x 64x 32x

0.9 0.92 0.94 0.96 0.98 1 1 10 100 1000 10000 100000 msec RAID-5 PARAID-5

32x - 1.8ms, 26% slower due to time spent in low gear Overhead

slide-33
SLIDE 33

33

Cello99 Bandwidth

1 10 100 1000 500000 1000000 1500000 request number MB/sec RAID-5 PARAID-5

1 10 100 1000 500000 1000000 1500000 requests MB/sec RAID-5 PARAID-5

PARAID: A Gear-Shifting Power-Aware RAID

1 10 100 1000 500000 1000000 1500000 request number MB/sec RAID-5 PARAID

64x 32x 128x

Overhead < 1% degra- dation during peak hours

slide-34
SLIDE 34

34

PostMark Benchmark

 Popular synthetic benchmark  Generates ISP-style workloads  Stresses peak read/write performance of storage

device

PARAID: A Gear-Shifting Power-Aware RAID

slide-35
SLIDE 35

35

Postmark Performance

PARAID: A Gear-Shifting Power-Aware RAID 50 100 150 200 1K files, 50K trans 20K files, 50K trans 20K files, 100K trans seconds RAID-5 PARAID-5 high gear PARAID-5 low-gear

slide-36
SLIDE 36

36

Postmark Power Savings

PARAID: A Gear-Shifting Power-Aware RAID

1 20 30 40 50 60 70 80 1 1 1 21 31 41 51 61 71 81 91 1 1 1 1 1 2 1 3 1 4 1 51 1 6 1 71

seconds watts

R A ID 5 PA R A ID

slide-37
SLIDE 37

37

Related Work

 Pergamum  EERAID  RIMAC  Hibernator  MAID  PDC  BlueFS

PARAID: A Gear-Shifting Power-Aware RAID

slide-38
SLIDE 38

38

Future Work

 Try more workloads  Optimize PARAID gear configuration  Explore asynchronous update propagation  Speed up recovery  Live testing

PARAID: A Gear-Shifting Power-Aware RAID

slide-39
SLIDE 39

39

Lessons Learned

 Third version of design, early design too

complicated

 Data alignment problems  Difficult to measure system under normal load  Hard to predict workload transformations due to

complex system optimizations

 Challenging to match trace environments

PARAID: A Gear-Shifting Power-Aware RAID

slide-40
SLIDE 40

40

Conclusion

 PARAID reuses standard RAID-levels without

special hardware while decreasing their energy use by 34%.

 Optimized version can save even more energy

 Empirical evaluation important

PARAID: A Gear-Shifting Power-Aware RAID

slide-41
SLIDE 41

41

Research Theme

 Data flow management

 Storage  MANETs

 Current state

 Reminiscent of plumbing industry 200 years

ago

 Limited interchangeable parts  Poorly understood interactions

slide-42
SLIDE 42

42

Research Areas

 Power-Aware RAID  Electric-field-based routing for MANETs  Conquest disk-persistent-RAM hybrid file

system

 Optimistic replication  Real-time systems

slide-43
SLIDE 43

43

Questions

PARAID: A Gear-Shifting Power-Aware RAID

 Contact

 Andy Wang – awang@cs.fsu.edu

 http://www.cs.fsu.edu/~awang/conquest-2

slide-44
SLIDE 44

44

PARAID Recovery

2.7 times slower than conventional raid

For example, 2 gear PARAID device

First, the soft state must recover

Second, data must be propagated

Third, conventional raid must recover

Recovery not as bad for read intensive workloads

slide-45
SLIDE 45

45

PARAID Gear-Shifting

256x 128x 64x Number of gear switches 15.2 8.0 2.0 % time spent in low gear 52% 88% 98% % extra I/Os for update propagations 0.63% 0.37% 0.21% 128x 64x 32x Number of gear switches 6.0 5.6 5.4 % time spent in low gear 47% 74% 88% % extra I/Os for update propagations 8.0% 15% 21%

Web Trace Gear-Shifting Stats Cello99 Gear-Shifting Stats