PARAID: The Gear-Shifting Power-Aware RAID

1
PARAIDThe Gear-ShiftingPower-Aware RAID

• Charles Weddle, Mathew Oldham, An-I Andy Wang
  Florida State University
• Peter Reiher University of California, Los
  Angeles
• Geoff Kuenning Harvey Mudd College

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
• Root to other issues, e.g. server room cooling
• Is it possible to reduce energy consumption
  without degrading performance?

The Power Aware RAID www.cs.fsu.edu/weddle/
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

The Power Aware RAID www.cs.fsu.edu/weddle/
4
Challenges

• Existing Work
• Most trade performance for energy savings
  directly.
• e.g. vary speed of disks
• Most are simulated results

The Power Aware RAID www.cs.fsu.edu/weddle/
5
Observations

• Over provisioning of resources
• 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

The Power Aware RAID www.cs.fsu.edu/weddle/
6
Power-Aware RAID

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

The Power Aware RAID www.cs.fsu.edu/weddle/
7
Skewed Striping for Energy Saving

• Use over-provisioned spare storage
• Can use fewer disks for light loads

The Power Aware RAID www.cs.fsu.edu/weddle/
8
Skewed Striping for Energy Saving

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

The Power Aware RAID www.cs.fsu.edu/weddle/
9
Skewed Striping for Energy Saving

• Approximate the workload
• Gear shift into most appropriate gear
• Minimize the opportunity lost to save power

Conventional RAID
PARAID
Energy ( Powered On Disks )
workload
Workload ( Disk Parallelism )
The Power Aware RAID www.cs.fsu.edu/weddle/
10
Skewed Striping for Energy Saving

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

The Power Aware RAID www.cs.fsu.edu/weddle/
11
Preserving Peak Performance

• Operate in the highest gear
• When the system demands peak performance
• Maximize parallelism within each gear
• Load is balanced on each gear
• Uniform striping pattern within each gear
• Delay block replication until gear shifts
• Capture block writes

The Power Aware RAID www.cs.fsu.edu/weddle/
12
Maintaining Reliability

• Distributed parity (RAID-5)
• Tolerate single-disk failures
• Used in soft state
• Drives have a limited number of power cycles
• Form bi-modal distribution of busy/idle drives
• Rotate drives with more power cycles
• Ration number of power cycles

The Power Aware RAID www.cs.fsu.edu/weddle/
13
Maintaining Reliability

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

The Power Aware RAID www.cs.fsu.edu/weddle/
14
PARAID Implementation

• Implemented in Linux 2.6.5
• 3500 lines of code

The Power Aware RAID www.cs.fsu.edu/weddle/
15
PARAID Evaluation

• Questions to answer
• Can energy consumption be reduced?
• Can performance be maintained?
• Other questions to answer
• Do the power measurements reflect workload as
  expected?

The Power Aware RAID www.cs.fsu.edu/weddle/
16
PARAID Evaluation

• Measurement framework

The Power Aware RAID www.cs.fsu.edu/weddle/
17
PARAID Evaluation

• FSU web trace workload
• Obtained from Florida State University CS web
  servers - two months of traces
• September 19 - 21, 2004, 17,000 requests
• Captured web server trace files and a snapshot of
  the file system
• No Worries, file names and content was encrypted

The Power Aware RAID www.cs.fsu.edu/weddle/
18
Energy Savings
256x
32x
128x
Speed-Up Energy Savings
256x 3.3
128x 1.2
64x 12
32x 19
64x
The Power Aware RAID www.cs.fsu.edu/weddle/
19
Disk Utilization
256x
128x
32x
64x
The Power Aware RAID www.cs.fsu.edu/weddle/
20
Performance Peak Hour Latency
RAID 5 PARAID
256x 85 76
126x 79 83
64x 87 88
32x 92 92
The Power Aware RAID www.cs.fsu.edu/weddle/
21
Performance Peak Hour Total Completion Time
RAID 5 PARAID
256x 79 71
128x 75 78
64x 81 81
32x 87 86
The Power Aware RAID www.cs.fsu.edu/weddle/
22
Postmark Benchmark

• Postmark benchmark
• Popular synthetic benchmark
• Stresses peak read/write performance of storage
  device

The Power Aware RAID www.cs.fsu.edu/weddle/
23
Conclusion

• Energy efficiency and performance can be achieved
  simultaneously
• Work in progress
• Understand PARAID under a wider range of
  workloads.
• Explore gear-centric parity schemes
• Optimize gear selection and gear shifting

The Power Aware RAID www.cs.fsu.edu/weddle/
24
Questions

• PARAID
• The Gear-Shifting
• Power-Aware RAID
• For more information
• www.cs.fsu.edu/weddle/
• Contact
• Charles Weddle weddle_at_cs.fsu.edu

The Power Aware RAID www.cs.fsu.edu/weddle/
25
Postmark Benchmark- Total Completion Time
RAID 5 PARAID High Gear PARAID Low Gear
1k Files, 50k Trans 5s 6s 4.8s
20k Files, 50k Trans 64.8s 58.8s 130.4s
20k Files, 100kTrans 147.8s 136.2s 261.2s
The Power Aware RAID www.cs.fsu.edu/weddle/
26
Postmark BenchmarkPower 20k Files, 100k
Transactions
Start in Low Gear
Start in High Gear
The Power Aware RAID www.cs.fsu.edu/weddle/