BUDW: Energy-Efficient Parallel Storage Systems with Write-Buffer Disks
A critical challenge with modern parallel I/O systems is that parallel disks consume a significant amount of energy in servers and high-performance computers. To conserve energy consumption in parallel I/O systems, one can immediately spin down disks when disk are idle; however, spinning down disks might not be able to produce energy savings due to penalties of spinning operations. Unlike powering up CPUs, spinning down and up disks need physical movements. Therefore, energy savings provided by spinning down operations must offset energy penalties of the disk spinning operations. To reduce the penalties incurred by disk spinning operations, we describe in this talk an approach to conserving energy of parallel I/O systems with write buffer disks, which are used to accumulate small writes using a log file system. Data sets buffered in the log file system can be transferred to target data disks in a batch way. Thus, buffer disks aim to serve a majority of incoming write requests, attempting to reduce the large number of disk spinning operations by keeping data disks in standby for long period times. Interestingly, the write buffer disks not only can achieve high energy efficiency in parallel I/O systems, but also can shorten response times of write requests. To evaluate the performance and energy efficiency of our parallel I/O systems with buffer disks, we implemented a prototype using a cluster storage system as a testbed. Experimental results show that under light and moderate I/O load, buffer disks can be employed to significantly reduce energy dissipation in parallel I/O systems without adverse impacts on I/O performance.
Recommandé
Recommandé
Contenu connexe
Tendances
Tendances (20)
Similaire à BUDW: Energy-Efficient Parallel Storage Systems with Write-Buffer Disks
Similaire à BUDW: Energy-Efficient Parallel Storage Systems with Write-Buffer Disks (20)
Plus de Xiao Qin
Plus de Xiao Qin (20)
Dernier
Dernier (20)
BUDW: Energy-Efficient Parallel Storage Systems with Write-Buffer Disks
- 1. Energy-Efficient Parallel Storage Systems with Write-Buffer Disks XiaojunRuan and Xiao Qin Computer Science and Software Engineering Samuel Ginn College of Engineering Auburn University
- 2. My Research Group: 2011
- 3. Overview of the Project Energy Efficiency Security Solid State Drives Performance of Secure Disk Systems [IEEE NAS09] Design, Model, Simulate, And Evaluate Disk Systems with Buffer Disks [ACM SAC09][ICPP09] Enhancing Internal Parallelism of SSDs [To Be Submitted11] Message Passing Interface with Enhanced Security [IPCCC 2010] Energy-Efficient Distributed Storage Systems [IPCCC10] Energy-Efficient Dynamic Voltage Scaling[ICCCN07] Xiaojun Ruan 3 BUD
- 4. 4 6/10/2011 Electricity Usage in Data Centers Annual Data Center Electricity Usage and Electricity Price increase Every year
- 5. The average power consumption of TOP10 supercomputing systems is 1.32 Mwatt. Dell’s Texas Data Center 5 6/10/2011 Energy Efficiency of Supercomputers
- 7. Server and Data Centers Consume 110 Billion kWh per year
- 8. Assume average commercial end user is charged 9.46 kWh
- 9. Disk systems can account for 27% of the energy cost of data centers6/10/2011 6 Server and data centers may have an electrical cost of 10.4 billion dollars.
- 10. 7 6/10/2011 Energy Consumption of Disks
- 11. Power States of Disks Active State: high energy consumption Active Standby State transition penalty Standby State: low energy consumption 8
- 12. A Hard Disk Drive A10000RPM Hard Drive may take 10.9 seconds to wake up! 9
- 13. Parallel Disks Performance Energy Efficiency
- 17. Reduce Status Transitions as many as possible12
- 18. IBM Ultrastar 36Z15 6/10/2011 13
- 19. A Parallel Disk System with a Write Buffer Disk
- 20. The BUD Architecture Data Disks can serve requests without buffer disks when workload is high 15
- 22. SRB is set by administrators
- 24. Example Buffer Disk Requests Queue 18
- 25. Auburn University Xiaojun Ruan 19 From Design to Simulation
- 27. Auburn University 21 Workloads
- 28. Impact of SRB—Low Workload, UltraStar 22
- 29. Auburn University Xiaojun Ruan 23 Non-Buffer Experiments
- 30. Auburn University 24 BUD with IBM 40GNX TravalStar
- 31. Buffer Disk Number and Workload-- UltraStar 25
- 32. Auburn University 26 Energy Consumption E = Active Energy Consumption + Standby Energy Consumption + Transition Penalty
- 33. Auburn University Xiaojun Ruan 27 From Simulation to Real Implementation
- 34. An Energy-Efficient Cluster Storage System 28
- 35. Implementation (no buffer disks) 29
- 38. Data Disk 1: WesternDigital 400, 20GB
- 39. Data Disk 2: WesternDigital 400, 20GB
- 40. Disk Category I/O Node 2
- 41. Data Disk 1: WesternDigital 400, 20GB
- 42. Data Disk 2: Maxtor D740X-6L, 20GB31
- 44. Buffer Disk: Maxtor DiamondMax Plus 9
- 45. Data Disk 1: WesternDigital 400, 20GB
- 46. Data Disk 2: WesternDigital 400, 20GB
- 47. Disk Category I/O Node 2
- 48. Buffer Disk: Seagate Barracuda 7200
- 49. Data Disk 1: WesternDigital 400, 20GB
- 50. Data Disk 2: Maxtor D740X-6L, 20GB32
- 51. 33
- 52. 34 idle time gap is 200s idle time gap is 100s
- 53. Previous Research GreenFS [ACM EuroSys 2008] Massive Arrays of Idle Disks [SC 2002] Popular Data Concentration [ACM ICS 2004] 35 6/10/2011
- 54. Download the presentation slideshttp://www.slideshare.net/xqin74 Google: slideshare Xiao Qin ‹#›
- 57. Download Slides at slidesharehttp://www.slideshare.net/xqin74
- 58. Auburn University 40 Questions?
Notes de l'éditeur
- Web Servers they account for 24% of the energy, 77% proxy server, 27% in data center
- , then spin the data disk on, otherwise we always keep data disks in standby mode.
- SRB is small, transfer data from buffer disk to data disk for too many times. Small SRB, more spin up and spin down times.
- Trend is similar, but ultrastar penalty is higher
- MAID:GreenFS:Benefits: low energy cost, Flaws: low capacity, low reliability