1 / 27

MEMS Based Mass Storage Systems

MEMS Based Mass Storage Systems. What is MEMS?. (M)icro(E)lectric(M)echanical(S)ystems Consist of mech µ(structures, sensors, actuators), electronics, integrated onto same chip Transducer = Sensor / Actuator Smart sensors Cheap Examples. Fender?.

kenyon
Download Presentation

MEMS Based Mass Storage Systems

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. MEMS Based Mass Storage Systems

  2. What is MEMS? • (M)icro(E)lectric(M)echanical(S)ystems • Consist of mech µ(structures, sensors, actuators), electronics, integrated onto same chip • Transducer = Sensor / Actuator • Smart sensors • Cheap • Examples

  3. Fender? • The world's smallest guitar is 10 micrometers long – • Made by Cornell University researchers from crystalline silicon

  4. Example

  5. Why use MEMS? Capacity @ Entry Cost • Cost • Examples 100 GB HARD DISK MEMS 10 GB 1 GB 0.1 GB DRAM CACHE RAM 0.01 GB $1 $100 $10 $1000 Entry Cost

  6. Why use MEMS?(cont.) • Volume • Examples Flash memory, 0.4 µm2 cell 100,000 10,000 3.5” Disk Drive 1000 Occupied volume [cm3] 100 10 Chip-sized data storage @ 10 GByte/cm2 1 0.1 0.1 1 10 100 1000 10,000 Storage Capacity [GByte]

  7. Why use MEMS?(cont.) • Lower data latency • Why not EEPROM? $300 / GB EEPROM (Flash) DRAM $100 / GB Prediction 2008 $30 / GB Cost $ / GB $10 / GB MEMS Worst-Case Access Time (Rotational Latency) $3 / GB HARD DISK $1 / GB 100µs 10ns 1µs 10ms

  8. Storage Device Design • 2 proposed models • Cantilever • “Moving media”

  9. Read/Write tips Actuators Magnetic Media “Moving Media”

  10. Bits stored underneath each tip “Moving Media” Read/write tips Media side view

  11. Logistics • Area = 1 cm2 • 10,000 probe tips • Bit cell of 0.0025-0.0009 µm2  4 – 11 GB • Advantages / disadvantages

  12. Data Layout • Cylinders • Tracks • Sectors • Logical block

  13. Device Performance • timeservice=time seek+latencyrotate+timetransfer • MEMS • timeservice=time seek +timetransfer time seek,acceleration, turnaround time, settling time

  14. Physical Characteristics • Bit Size • Access Velocity • Sled acceleration • Spring stiffness • Number of sleds • Number of active tips • Error rates

  15. Performance Characteristics • Seek time • Settle time • Turnaround time • Peak bandwidth • Capacity • Power • Reliability

  16. Example • Fast read-modify-write • No rotational latency Atlas 10K MEMS Read 0.14 0.13 Reposition 5.98 0.07 Write 0.14 0.13 Total 6.26 0.33

  17. Seek Time From Center

  18. Sustained Data Rate 1.6 Mbits / sec * 1280 tips = 2048 Mbits / sec

  19. Sustained Data Rate

  20. Failure Management • MEMS devices will have internal failures • Tips will break during fabrication/assembly, use • Media can wear ECC can be both horizontal and vertical Could then use spares to regain original level of reliability

  21. Performance Models • Generation 1 • Generation 2 • Generation 3 • Reference disk – Atlas 10k • Super disk

  22. Random Workload - Microbenchmark

  23. Postmark

  24. Power Utilization • Lower operating power • 100 mW for sled positioning • 1 mW per active tip • For 1000 active tips, total power is 1.1 watt • 50 mW standby mode • Fast transition from standby – 0.5 ms

  25. Future Potential • Definite advantages • Portable applications • New low-cost entry point • Archival storage • Active storage devices • Throwaway devices • …

  26. Problems? • Very little has been implemented • Power consumption? • Heat – kinetic energy? • Reliability? • Sturdiness? • Any other alternatives?

  27. Conclusions • Potential to fill the RAM/Disk gap • Simulation results show • reductions in I/O stall times • overall performance improvement We’ll have to wait and see …

More Related