PARAID: The Gear-Shifting Power-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

Motivation Energy costs are rising Energy costs are rising An increasing concern for servers An increasing concern for servers No longer limited to laptops No longer limited to laptops Energy consumption of disk drives Energy consumption of disk drives 24% of the power usage in web servers 24% of the power usage in web servers 27% of electricity cost for data centers 27% of electricity cost for data centers Root to other issues, e.g. server room cooling Root to other issues, e.g. server room cooling Is it possible to reduce energy consumption without degrading performance? Is it possible to reduce energy consumption without degrading performance? The Power Aware RAID –

Challenges Energy Energy Not enough opportunities to spin down RAIDs Not enough opportunities to spin down RAIDs Performance Performance Essential for peak loads Essential for peak loads Reliability Reliability Server-class drives are not designed for frequent power switching Server-class drives are not designed for frequent power switching The Power Aware RAID –

Challenges Existing Work Existing Work Most trade performance for energy savings directly. Most trade performance for energy savings directly. e.g. vary speed of disks e.g. vary speed of disks Most are simulated results Most are simulated results The Power Aware RAID –

Observations Over provisioning of resources Over provisioning of resources RAID is configured for peak performance RAID is configured for peak performance RAID keeps all drives spinning for light loads RAID keeps all drives spinning for light loads Unused storage capacity Unused storage capacity Over-provision of storage capacity Over-provision of storage capacity Unused storage can be traded for energy savings Unused storage can be traded for energy savings Fluctuating load Fluctuating load Cyclic fluctuation of loads Cyclic fluctuation of loads Infrequent on-off power transitions can be effective Infrequent on-off power transitions can be effective The Power Aware RAID –

Power-Aware RAID Skewed striping for energy savings Skewed striping for energy savings Preserving peak performance Preserving peak performance Maintaining reliability Maintaining reliability Evaluation Evaluation Conclusion Conclusion Questions Questions The Power Aware RAID –

Skewed Striping for Energy Saving Use over-provisioned spare storage Use over-provisioned spare storage Can use fewer disks for light loads Can use fewer disks for light loads The Power Aware RAID – gear 1 Soft-State Block Replication disk 1 gear 2 RAID Layout disk 2disk 3disk 4disk 5

Skewed Striping for Energy Saving Operate in gear 1 Operate in gear 1 Disks 4 and 5 are powered off Disks 4 and 5 are powered off The Power Aware RAID – gear 1 Soft-State Block Replication disk 1 gear 2 RAID Layout disk 2disk 3disk 4disk 5

Skewed Striping for Energy Saving Approximate the workload Approximate the workload Gear shift into most appropriate gear Gear shift into most appropriate gear Minimize the opportunity lost to save power Minimize the opportunity lost to save power The Power Aware RAID – Energy ( Powered On Disks ) Workload ( Disk Parallelism ) Conventional RAIDPARAID workload

Skewed Striping for Energy Saving Adapt to cyclic fluctuating workload Adapt to cyclic fluctuating workload Gear shift when gear utilization threshold is met Gear shift when gear utilization threshold is met The Power Aware RAID – time load utilization threshold gear shift

Preserving Peak Performance Operate in the highest gear Operate in the highest gear When the system demands peak performance When the system demands peak performance Maximize parallelism within each gear Maximize parallelism within each gear Load is balanced on each gear Load is balanced on each gear Uniform striping pattern within each gear Uniform striping pattern within each gear Delay block replication until gear shifts Delay block replication until gear shifts Capture block writes Capture block writes The Power Aware RAID –

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

Maintaining Reliability Busy disk stay powered on, idle disks stay powered off Busy disk stay powered on, idle disks stay powered off Outside disks are role exchanged with middle disks Outside disks are role exchanged with middle disks The Power Aware RAID – busy disks power cycled disks idle disks role exchange Disk 1 Gear 1 Gear 2 Gear 3 Disk 2Disk 3Disk 4Disk 5Disk 6

PARAID Implementation Implemented in Linux Implemented in Linux ≈ 3500 lines of code ≈ 3500 lines of code The Power Aware RAID – File System RAID PARAID Block Handler Disk Device Driver PARAID User Administration PARAID Monitor PARAID Disk Manager PARAID Reliability Manager User Space Linux kernel Software RAID

PARAID Evaluation Questions to answer Questions to answer Can energy consumption be reduced? Can energy consumption be reduced? Can performance be maintained? Can performance be maintained? Other questions to answer Other questions to answer Do the power measurements reflect workload as expected? Do the power measurements reflect workload as expected? The Power Aware RAID –

PARAID Evaluation Measurement framework Measurement framework The Power Aware RAID – multimeter USB cable client server power supply 12v & 5v power lines power measurement probes SCSI cable crossover cable Xeon 2.8 Ghz, 2 GB RAM 36.7 GB 10k RPM SCSI P4 2.8 Ghz, 1 GB RAM 160 GB 7200 RPM SATA RAID BOOT

PARAID Evaluation FSU web trace workload FSU web trace workload Obtained from Florida State University CS web servers - two months of traces Obtained from Florida State University CS web servers - two months of traces September , 2004, 17,000 requests September , 2004, 17,000 requests Captured web server trace files and a snapshot of the file system Captured web server trace files and a snapshot of the file system No Worries, file names and content was encrypted No Worries, file names and content was encrypted The Power Aware RAID –

Energy Savings The Power Aware RAID – 256x 128x 64x 32xSpeed-Up Energy Savings 256x3.3% 128x1.2% 64x12% 32x19%

Disk Utilization The Power Aware RAID – 256x 128x 64x 32x

Performance – Peak Hour Latency RAID 5 PARAID 256x85%76% 126x79%83% 64x87%88% 32x92%92% The Power Aware RAID –

Performance – Peak Hour Total Completion Time RAID 5 PARAID 256x79%71% 128x75%78% 64x81%81% 32x87%86% The Power Aware RAID –

Postmark Benchmark Postmark benchmark Postmark benchmark Popular synthetic benchmark Popular synthetic benchmark Stresses peak read/write performance of storage device Stresses peak read/write performance of storage device The Power Aware RAID –

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

Questions PARAID: The Gear-Shifting Power-Aware RAID For more information For more information Contact Contact Charles Weddle – Charles Weddle – The Power Aware RAID –

Postmark Benchmark- Total Completion Time The Power Aware RAID – RAID 5 PARAID High Gear PARAID Low Gear 1k Files, 50k Trans 5s6s4.8s 20k Files, 50k Trans 64.8s58.8s130.4s 20k Files, 100kTrans147.8s136.2s261.2s

Postmark Benchmark Power – 20k Files, 100k Transactions The Power Aware RAID – Start in High Gear Start in Low Gear