X

Download A Birds-Eye View Over Faults PowerPoint Presentation

SlidesFinder-Advertising-Design.jpg

Login   OR  Register
X


Iframe embed code :



Presentation url :

Home / Forest & Animals / Forest & Animals Presentations / A Birds-Eye View Over Faults PowerPoint Presentation

A Birds-Eye View Over Faults PowerPoint Presentation

Ppt Presentation Embed Code   Zoom Ppt Presentation

PowerPoint is the world's most popular presentation software which can let you create professional A Birds-Eye View Over Faults powerpoint presentation easily and in no time. This helps you give your presentation on A Birds-Eye View Over Faults in a conference, a school lecture, a business proposal, in a webinar and business and professional representations.

The uploader spent his/her valuable time to create this A Birds-Eye View Over Faults powerpoint presentation slides, to share his/her useful content with the world. This ppt presentation uploaded by onlinesearch in Forest & Animals ppt presentation category is available for free download,and can be used according to your industries like finance, marketing, education, health and many more.

About This Presentation

A Birds-Eye View Over Faults Presentation Transcript

Slide 1 - Winter Semester 2010 ”Politehnica” University of Timisoara Course No. 4: Expanding Bio-Inspiration: Towards Reliable MuxTree Memory Arrays – Part 1 – Emerging Systems
Slide 2 - 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)
Slide 3 - 1.1. Introduction Performance VS Fault Tolerance ↓ Performance AND Fault Tolerance Hot issues in bio-inspiration: mimicking nature, also mimic results? Intersection point: where? exporting biological features toward computer engineering: possible? Chapter 1: Bio-Inspired Reliability (1)
Slide 4 - 1.2. A Plea for Bio-Inspiration Existing: intrinsic robustness of living beings Sustaining many minor wounds/illnesses: self-diagnosing Quick healing: growing/replacing damaged tissues Perfected in time Wanted: robust computing systems Self-testing and self-repair mechanisms through redundancies Immunotronics: bio-inspired immune system for silicon devices Chapter 1: Bio-Inspired Reliability (2)
Slide 5 - 1.2. A Plea for Bio-Inspiration HW redundancy Spares take over faulty resources * Process supervised centrally Overall reliability limited by centralized unit Bio-inspired redundancy (Embryonics) Spares take over faulty resources * Process distributed Multiple level redundancy Chapter 1: Bio-Inspired Reliability (3)
Slide 6 - 1.3. Embryonics and Stem Cells Human brain Exponent of epigenetic systems Complexity given by 1010 cells and 1014 interconnections All from a single, original cell: the zygote Stem cells Recently discovered Division potential apparently unlimited Offspring: identical or able to specialize Anyone potentially becoming a fetus Chapter 1: Bio-Inspired Reliability (4)
Slide 7 - 1.3. Embryonics and Stem Cells Similarities: Cells made by molecules. No limits on cell number or cellular dimensions Indefinite cellular division. Process not material, but informational Specialization via gene expression: information governs over functionality Chapter 1: Bio-Inspired Reliability (5)
Slide 8 - 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)
Slide 9 - 2.1. Bio-Inspired Storage Genetic program stored into cyclic memory structure – macro-cell Large genomes → large macro-cells → increased error likelihood Two-level self-repair mechanism refers to system functionality only Data integrity extremely important, functionality governed by genome No data integrity protection implemented Chapter 2: A Birds-Eye View Over Faults (1)
Slide 10 - 2.2. Fault Taxonomy Over Time Chapter 2: A Birds-Eye View Over Faults (4) Permanent faults: affect normal device operation constantly, over an indefinite period of time; solid or hard fails Non-permanent faults: random occurrence, effect over indefinite but finite periods of time; Intermittent: non-environmental causes (parameter variations, timing, loose connections, etc); difficult debugging process Transient: environmental causes (atmospheric parameters, supply characteristics, cosmic rays); very difficult/impossible to model ↓ Soft fails, soft errors or single event upsets (SEUs): transient type
Slide 11 - 2.3. Single Event Upsets: An Analysis Chapter 2: A Birds-Eye View Over Faults (6) Affect normal execution process Mostly caused by electronic noise: ionization electron-hole pairs Particles capable of generating, energetic nucleons and nuclear fragments: α–particles radioactive isotopes other particles grouped under the more general term of cosmic rays
Slide 12 - 2.3. Single Event Upsets: An Analysis Chapter 2: A Birds-Eye View Over Faults (7) Radiation affecting devices long known 1978, Ziegler (IBM): if α–particles affect semiconductor devices, then other particles from the outer space (cosmic rays) also might 1979, May and Woods (Intel): radioactive isotopes affect memories Confirmation under device irradiating Studied since 1980, hot field today
Slide 13 - 2.3. Single Event Upsets: An Analysis Chapter 2: A Birds-Eye View Over Faults (8) Protective techniques used in modern devices Focus on memories, prone to soft fails: Techniques well understood, relatively easy to implement, not expensive Memory area in systems proportionally significant Combinational logic much less susceptible to soft errors SER not constant, dependant on technological advancement
Slide 14 - 2.3.1. Radioactive Isotopes Chapter 2: A Birds-Eye View Over Faults (9) Due to proximity contamination in semiconductor facilities Famous “HERA” problem, 1987: IBM produced LSI memories hit by anomalous behavior Chips produce in US and Europe, but only US-produced batches affected Package responsibility ruled out
Slide 15 - 2.3.1. Radioactive Isotopes Chapter 2: A Birds-Eye View Over Faults (10) Finally, traces of Po210 discovered Product of the uranium decay chain Strangely, other expected particles totally missing After months of searching, the culprit found as a bottle of nitric acid, used in the process Intel also hit by problems: 2107 series, 16Kb DRAM memories Cause: radioactive contamination Reason: factory built downstream, close to an old uranium mine
Slide 16 - 2.3.2. Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (11) Strong electromagnetic perturbations induced by any ionizing particle hitting an electronic device Disturbances translate into electron–hole pairs If local fields sufficiently strong, pairs cannot recombine; instead, find a way out to the nearest appropriate contact Charge collection provoke soft errors
Slide 17 - 2.3.2. Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (12) Particles entering atmosphere collide: elastic and inelastic collisions Secondary nuclear fragments generated in avalanche-type phenomena: nucleons, pions, light ions (2H, 3H, 3He and α–particles), and heavy residual nuclei (O, C, Mg) Disturbances created by cosmic ray collision with semiconductor nuclei High-energy particle flux born, reaching the surface
Slide 18 - 2.3.2. Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (13) Final flux distribution affected by: Altitude: Due to atmospheric filtering effect, the lower the altitude, the smaller particle rate Geomagnetic region (GMR): shielding of Earth‘s magnetic field; cosmic rays penetration smallest around equator, largest at poles 11-year solar cycle: strongly affects particle flux; increased sun activity leads to lower rate of particles, because shielding effect of magnetic field around the earth also strengthened
Slide 19 - 2.3.2. Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (14) Attempts on flux measurement as early as 1980s (IBM). At least 3 categories of cosmic rays Primary cosmic rays: lurking particles in the outer space, eventually may hit the planet; product of intense reactions
Slide 20 - 2.3.2. Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (15) Sun also major source of primary cosmic rays Measured flux of primary cosmic rays about 105/m2·s Secondary cosmic rays: born by collisions when primary cosmic rays enter atmosphere; also known as “cascade particles”
Slide 21 - 2.3.2. Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (16) Terrestrial cosmic rays: particles actually reaching sea-level max. 1% originate from primary cosmic rays the rest cascade-generated particles, from the 3rd to 7th generations extremes in sea-level particle flux lag solar cycle by 1-2 years Final flux made of hadrons (collide due to atmospheric density), leptons and photons
Slide 22 - 2.3.2. Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (18) Flux measurement : 360/m2·s Most particles decay spontaneously or reach thermal energies (absorbed by the atmosphere) Peak of cascading density at an altitude of about 15 km, usually referred to as the Pfotzer point Altitude used by many commercial aircraft This is where the fail rate of electronic devices is about 100 times worse than at terrestrial altitudes
Slide 23 - 2.3.2. Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (19) After protons hit atmosphere most resulting fragments decay or absorbed Proton flux largely affected by interactions with atmospheric electrons Muons dominate the medium/ high energy particle spectrum Energetic levels of most concern for SER between 200 and 3000MeV
Slide 24 - 2.3.3. Modelling Cosmic Ray Influence Chapter 2: A Birds-Eye View Over Faults (21) Significant effort towards particle measurements Vast effort in providing computational models CREME96 (Cosmic Ray Effects on Micro Electronics Code): numerical models for ionizing radiations (near-Earth orbits) and effects in spacecraft NUSPA (NUclear SPAllation): initial focus on sea-level interactions (where protons and neutrons play the most significant role); extended model interactions at high altitudes, including pion interactions
Slide 25 - 2.3.4. Introduction to Particle Physics Chapter 2: A Birds-Eye View Over Faults (22) All known particles made of 2 types of building bricks: leptons and quarks Leptons: do not interact by strong nuclear force electrons, muons, tau particles, and their associated neutrinos believed to be point-like particles Hadrons: interact by strong nuclear force Mesons: made by 2 quarks (pions) Baryons: made by 3 quarks (protons and neutrons)
Slide 26 - 2.3.4. Introduction to Particle Physics Chapter 2: A Birds-Eye View Over Faults (23) At sea-level approx. 94% of hadron total flux made by neutrons, 2% protons Neutrons (uncharged particles) usually go through electrical circuits with no interactions Unless combined into a nucleus, a free neutron follows a β-decay process (10.3 min half-life)
Slide 27 - 2.3.4. Introduction to Particle Physics Chapter 2: A Birds-Eye View Over Faults (24) Only 1 neutron out of 4·104 hit a silicon nucleus Hit very likely to produce soft fail About 105 neutrons/cm2·year at sea level above 20MeV Neutron flux exponentially influenced by altitude
Slide 28 - 2.3.4. Introduction to Particle Physics Chapter 2: A Birds-Eye View Over Faults (24) Pions (π-mesons): unstable particles, 135MeV mass, 2% of total hadron flux at sea-level neutral pion (its anti-particle being itself), with a lifetime of 10-16s positive pion (its anti-particle being the negative pion), longer lifetimes of about 26ns (26·10-9s) SER contribution expected negligible due to small numbers compared to other nucleons
Slide 29 - 2.3.4. Introduction to Particle Physics Chapter 2: A Birds-Eye View Over Faults (25) Particles released during pion decay: electron, positron, gamma ray, muon, and neutrino Charged pions can interact with matter Low-energy positive pions repelled by the nucleus
Slide 30 - 2.3.4. Introduction to Particle Physics Chapter 2: A Birds-Eye View Over Faults (27) Energetic, positive pion may lose kinetic energy to a nucleus Pion capture, most contributive to SER Pion-capture: entire pion mass is transformed into nucleonic energy, provoking a nuclear fision
Slide 31 - 2.3.4. Introduction to Particle Physics Chapter 2: A Birds-Eye View Over Faults (28) Every negative pion capture within the active volume of an electronic circuit leads to a soft fail Pion capture into silicon about 8.5/cm3·year at sea-level For intermediate energies (between 100 and 250 MeV) pion contributions to SER of modern chips (16-Mb DRAMs) about 5 times greater that SER caused by protons Pions may have significant impact at aircraft altitudes Studies of pion-induced soft errors still in progress
Slide 32 - 2.3.4. Introduction to Particle Physics Chapter 2: A Birds-Eye View Over Faults (29) Muons (μ-leptons): lifetime of 2.2 us Produced in the upper atmosphere by pion decay Two effects leading to soft fails: electrostatic muon scattering from nuclei muon capture (about 510/cm3·year): orbiting a Si nucleus; most of the initial mass-energy, the neutrino
Slide 33 - 2.3.5. Ion-Induced SEUs Chapter 2: A Birds-Eye View Over Faults (30) Discrete satellite components highly resistant to radiations; modern satellites sensitive Energetic heavy ions (100MeV Fe) at least partly responsible for reported fails Croley et al.: roughly two-thirds of the fails due to ions with Z ≥ 6 (C, O, Fe, N, Ne, Mg, Si, S, Ar, Ca) At satellite altitudes heavy energetic ions responsible for at least 45% (roughly equal proportions with the proton-induced) of SEUs
Slide 34 - 2.3.6. Neutron-Induced SEUs Chapter 2: A Birds-Eye View Over Faults (31) Pathway of electron–hole pairs due to the energetic impact of a high-energy neutron striking a p-n junction Minimum charge resulting in a soft error called critical charge (QCRIT) Two types of neutron interactions: Elastic scattering (target nucleus not excited, until QCRIT becomes smaller that 50fC minor role in SER) Inelastic scattering
Slide 35 - 2.3.6. Neutron-Induced SEUs Chapter 2: A Birds-Eye View Over Faults (32) Inelastic scattering nucleon + target → X1 + X2 + …+ Xn + residual nucleus much higher exchange energies (order of MeV or larger) identity of incoming particle lost pions may also be produced Example of inelastic scattering:
Slide 36 - 2.3.7. α-Induced SEUs Chapter 2: A Birds-Eye View Over Faults (32) Natural decay processes (of U, Th, Ac and Np) generate helium ions 4He or α–particles α–particles large mass and charge, most upsetting α–decay may release energetic α–particles
Slide 37 - 2.3.7. α-Induced SEUs Chapter 2: A Birds-Eye View Over Faults (33) Also due to impurities in semiconductor chip materials and packaging Technological progress, careful supervision of materials quality: α–particle SER momentum lost α–particles also emitted by at most 3% of captured muons
Slide 38 - 2.3.8. Proton-Induced SEUs Chapter 2: A Birds-Eye View Over Faults (33) Particle distribution change with altitude Inelastic scattering example: Uncertainties concerning proton influences over silicon