Energy Efficient Prefetching with Buffer Disks for Cluster
Energy Efficient Prefetching with Buffer Disks for Cluster File Systems Adam Manzanares and Xiao Qin Department of Computer Science and Software Engineering Auburn University http: //www. eng. auburn. edu/~xqin@auburn. edu 1/5/2022 1
Motivation Using 2010 Historical Trends Scenario ◦ Server and Data Centers Consume 110 Billion k. Wh per year ◦ Assume average commercial end user is charged 9. 46 k. Wh ◦ Disk systems can account for 27% of the energy cost of data centers 1/5/2022 2
Buffer Disk Architecture Energy-Related Reliability Model Prefetching Data Partitioning Security Model Disk Requests RAM Buffer Disk Controller Load Balancing Power Management m buffer disks 1/5/2022 n data disks 3
Energy Saving Principles Energy Saving Principle One ◦ Increase the length and number of idle periods larger than the disk break-even time TBE Energy Saving Principle Two ◦ Reduce the number of power-state transitions 1/5/2022 4
Paramaters Tested 1/5/2022 Parameter Values Data Size 1, 5, 10, 25 MB # of Data Disks 4, 8, 12 Inter-arrival Delay 0, 0. 1, 0. 5, 1 S Hit Rate 85, 90, 95, 100% 5
Energy Savings Hit Rate 85% 1/5/2022 6
State Transitions 1/5/2022 7
Why a Cluster File System • Block level prefetching difficult • Natural place to track file accesses • Control placement of data among storage nodes, and data disks • Tiered approach simplifies management of files and disk states • Eliminates some shortcomings of modeling and simulation 1/5/2022 8
Energy Efficient Virtual File System 1/5/2022 9
EEVFS Process Flow 1/5/2022 10
EEVFS Testbed Parameter Storage Server Storage Node Type 1 Storage Node Type 2 CPU P 4 2. 0 GHz P 4 3. 2 GHz P 4 2. 4 GHz Memory (MB) 2000 1000 512 Network Interconnect 1000 100 Disk Type SATA ATA/133 Disk Capacity 120 GB 80 GB Disk Bandwidth 100 MB/s 58 MB/s 34 MB/s 1/5/2022 11
Energy Savings 1/5/2022 12
State Transitions 1/5/2022 13
Response Times 1/5/2022 14
Berkeley Web Trace 1/5/2022 15
EEVFS Summary • Knowledge of requests assumed and may be hard to come by • Performance tied to one of the buffer disks 1/5/2022 16
Parallel Striping Groups File 1 Buffer Disk Group 1 File 3 File 2 Disk 1 Disk 2 Buffer Disk Storage Node 1 Buffer Disk 3 Disk 5 File 4 Disk 6 Storage Node 3 Buffer Disk 4 Storage Node 2 1/5/2022 Group 2 Disk 7 Disk 8 Storage Node 4 17
Striping Within a Group Buffer Disk 1 2 Disk 1 3 5 7 9 Disk 2 4 6 8 10 Disk 4 4 6 8 10 Storage Node 1 Buffer Disk 1 2 Disk 3 3 5 7 9 Storage Node 2 1 File 1 1/5/2022 Group 1 18 File 22 2
Striping Within a Group • Number of disks in a group can be matched to nearest bottleneck • Striping within the group maintains relatively high performance • Allows us to use a buffer disk for each storage node, while still maintaining file striping level 1/5/2022 19
Testbed 1/5/2022 Parameter Storage Server Storage Node CPU Celeron 2. 2 GHz Memory (MB) 2000 Network Interconnect 1000 Disk Type SATA Disk Capacity 160 GB 480 GB Disk Bandwidth 126 MB/s 20
Measured Results 1/5/2022 21
Measured Results 1/5/2022 22
Berkeley Web Trace 1/5/2022 23
Response Time Comparison Parameter Striping No Striping Energy Consumption (J) 2, 088, 113 2, 100, 243 Response Time (S) 2. 78 13. 87 • Energy efficiency is slightly improved • Response time gain is significant 1/5/2022 24
Summary • Improves the energy efficiency and performance of a storage system • Designed to scale – Needs to be tested on large scale storage system 1/5/2022 25
Future Work • Improve the EEVFS prototype for production use • Run EEVFS on large scale storage system – Investigate scaling effects 1/5/2022 26
Questions http: //www. eng. auburn. edu/~xqin/presentations
- Slides: 27