1 / 21

Winter Semester 2010

Emerging Systems. Course No. 5: Expanding Bio-Inspiration: Towards Reliable MuxTree  Memory Arrays – Part 2 –. ”Politehnica” University of Timisoara. Winter Semester 2010. Presentation Outline.

Download Presentation

Winter Semester 2010

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. Emerging Systems Course No. 5: Expanding Bio-Inspiration: Towards Reliable MuxTree Memory Arrays – Part 2 – ”Politehnica” University of Timisoara Winter Semester 2010

  2. Presentation Outline • Chapter 1: Bio-Inspired Reliability(With a plea for bio-inspiration and a comparison between artificial Embryonics cells and the stem cells from biology) • Chapter 2: A Bird’s Eye View Over Faults (Includes fault tolerance motivation, causes of unexpected, soft errors and a description of the physical phenomena involved) • Chapter 3: Embryonics and SEUs (Particularities of the project, datapath model in memory structures, and reliability analysis)

  3. Chapter 3: Embryonics and SEUs (1) Current state-of-the-art • Bio-inspired memory for Embryonics genome storage • Genome storage critical: • drives actual hardware (polymerase and ribosomic genome) • contains instructions on how additional hardware will be driven (operative genome) • No memory protection mechanismscurrently • Both desirable and feasible

  4. Chapter 3: Embryonics and SEUs (1) 3.1. Error-Type Distribution • “By far the most common type of chip failure is a soft error of a single cell on a chip” • Multiple bit flips 1÷7% of total soft fails recorded • Double bit-flips under 5% of the total events • 2 cases of quadruple bit flip events witnessed; predicted rate 1 in 65 years per device

  5. Chapter 3: Embryonics and SEUs (2) 3.2. Datapath Model for Memory Structures • 3D matrix; M rows and N columns of physically identical storage molecules, of F 1-bit memory cells each • Data synchronously circled

  6. Chapter 3: Embryonics and SEUs (3) 3.2. Datapath Model for Memory Structures • For each Li,j a vicinity V(Li,j) = Lx,y Li,j Lz,w defined • Data shifting process:

  7. Chapter 3: Embryonics and SEUs (4) 3.2. Datapath Model for Memory Structures • Useful for error injection testing

  8. Chapter 3: Embryonics and SEUs (5) 3.3. Reliability Analysis • Basic assumption: failures exponentially distributedinside a molecule • Similar assumptions found to work well

  9. Chapter 3: Embryonics and SEUs (6) 3.4. Error Coding • Failure situations: • Single failure; recovery by parity-based coding • Double failure; core affected by at least one error, at most two errors on the same row; recovery by Hamming-like codes • Multiple failure; same as previous, likelihood found to be minimal • Terminal failure; too many faults, cannot be recovered • No failures detected; either normal operating or undetectable combination of errors; does not require/ cannot be established recovery measures

  10. Chapter 3: Embryonics and SEUs (7) 3.4. Error Coding • Strategies of tolerating faults in Embryonics • Fault tolerance at the molecular level Advantage: isolating faulty molecules possible, use of the transparent reconfiguration process; Disadvantage: considerable portion of molecular core affected for redundant coding • Fault tolerance at the macro-cell level Advantage: separate macro-cells for redundant coding and additional logic Disadvantage: reconfiguration process quite difficult due to lack of addressing

  11. Chapter 3: Embryonics and SEUs (8) 3.5.1. Macro-Cell Level, Classic SEC

  12. Chapter 3: Embryonics and SEUs (9) 3.5.1. Macro-Cell Level, Classic SEC

  13. Chapter 3: Embryonics and SEUs (10) 3.5.1. Macro-Cell Level, Protochip SEC • Faults in a row superimposed onto a protochip • In each protochip, independent Poisson processes formed by failure types • a the probability for a type Afailure

  14. Chapter 3: Embryonics and SEUs (11) 3.5.1. Macro-Cell Level, Protochip SEC

  15. Chapter 3: Embryonics and SEUs (12) 3.5.2. Macro-Cell Level, Protochip DEC • a the probability for a type Afailure

  16. Chapter 3: Embryonics and SEUs (13) 3.5.2. Macro-Cell Level, Protochip DEC

  17. Chapter 3: Embryonics and SEUs (14) 3.6. Molecular Level • Molecular reliability λ known

  18. Chapter 3: Embryonics and SEUs (15) 3.6. Molecular Level • Reliability>90%: 28.4 million hours (SEC) VS 63.3 million hours (DEC) periods; • Reliability=50% reached after 89.8 million hours (SEC) VS 154.5 million hours (DEC)

  19. Chapter 3: Embryonics and SEUs (16) 3.7. Conclusions • Final expressions of R and MTTF quite complicated • Failure rate λ essentially empirical • determined through extensive measurements • may be affected by aggressive environments • constant → variable

  20. Chapter 3: Embryonics and SEUs (17) 3.7. Conclusions • Unfortunately, no accurate model for cosmic rays • Understanding causes and modeling soft fails hot field of research • Stochastic nature of soft fails

  21. Chapter 3: Embryonics and SEUs (18) 3.7. Conclusions • Different macro-cell configurations; may prove too small for real applications • Classic reliability analysis difficult, based on non-stochastic parameters • Protochip-based analysis with similar results, better suited to other influences (such as cosmic rays)

More Related