FAULTTOLERANT COMPUTING

1
FAULT-TOLERANT COMPUTING

• Fall 2006
• Daniel Ortiz-Arroyo
• Computer Science and Engineering Department
• Aalborg University, Esbjerg

2
About the Course

• This is a short introductory course
• 5 classes
• Each class 2 sessions/45 min each
• Exercises
• Prerequisites (DE education)
• Concepts of Probability
• Computer Architecture
• Reliability
• Software Systems

3
Course Contents
4
Reading Material

• Textbook No textbook required
• Optional Reference Books
• Software Fault Tolerance Techniques and
  Implementation by Laura L. Pullum ISBN
  1580531377 Publisher Artech House Computer
  Security Series, 2001
• Reliability of Computer Systems and Networks
  -Fault Tolerance Analysis and Design, M.L.
  Shooman, Wiley 2002
• Papers listed on courses web page

5
Today

• Review of Fault Tolerant Concepts
• Overview of the course/topics
• Topic list course format discussion

6
Course Goals

• Provide an overview of Fault Tolerant Computing
  (FTC)
• Hardware and software
• Models
• Implementation mechanisms
• Discuss research and real implementation cases
• Hugh area with more than 50 years of
  research/development

7
Motivation

• What is Fault-Tolerance?
• A fault-tolerant system is one that continues
  to perform at desired level of service in spite
  of failures in some components that constitute
  the system.
• What is Fault tolerant computing?
• Is the art and science of building computing
  systems that continue to operate satisfactorily
  in the presence of faults
• Computing correctly despite the existence of
  errors in the system

8
Motivation

• Why FTC is important
• FTC techniques are the foundation for other areas
  e.g. FT Control
• Techniques invented in FTC have been transferred
  to other fields

9
Motivation

• Dependable Systems are essential to humans
  Dependability is the ability of a system to
  deliver a service that can justifiably be trusted

Measured by
Means to achieve dependability
10
Motivation (contd.)

• Approaches to design fault tolerant computer
  systems
• Bottom-up designing fault tolerant components
  to integrate them into a fault tolerant system
• Top-down designing a fault tolerant system
  using components with little or not fault
  tolerance
• Top down is the most used approach

11
Motivation (contd.)

• Challenge of Fault Tolerant Computing using the
  top-down approach
• Given that both hardware and software components
  are unreliable, how do we build reliable systems
  from these unreliable components?

12
Motivation (contd.)

• A FTC system may be able to tolerate one or more
  fault-types including
• HW transient, intermittent or permanent hardware
  faults,
• HW SW design errors,
• operator errors, or
• externally induced upsets or physical damage.

13
Motivation (contd.)

• Examples of FTC mechanisms/systems at different
  levels
• PCs RAMs with parity checks and Error Correcting
  Codes (ECC) (HW)
• Workstations error detection (HW), occasional
  corrective action (SW), keeping logs (SW)
• RAID (Redundant Array of Inexpensive Disks)
• Distributed Systems

14
Introduction

• Historical Perspective
• Theory established by J. von Neumann, 1956
• Probabilistic logic and synthesis of reliable
  organism from unreliable components, Annals of
  mathematical studies, Princeton University Press
• The SAPO computer built in Prague, Czechoslovakia
  in 19501954 under the supervision of A. Svoboda
  was probably the first fault-tolerant computer.
• Used relays and a magnetic drum memory.
• The processor used triplication and voting (TMR),
  and the memory implemented error detection with
  automatic retries when an error was detected.

15
Introduction

• Historical Perspective
• Over the past 30 years, a number of
  fault-tolerant computers have been developed that
  fall into three general types
• long-life, unmaintainable computers,
• ultradependable, real-time computers, and
• high-availability computers.

16
Introduction

• Long-life, un-maintainable computers.
• Spacecrafts require computers to operate for long
  periods of time without external repair. Typical
  requirements are a probability of 95 that the
  computer will operate correctly for 510 years.
• JPL Self-Testing-and-Repairing (STAR) computer
  was the next fault-tolerant computer, developed
  by NASA in the late 1960s for a 10-year mission
  to the outer planets.

17
Introduction

• Ultra-dependable, real-time computers Computers
  for which an error or delay can prove to be
  catastrophic.
• They are designed for applications such as
  control of aircraft, mass transportation systems,
  and nuclear power plants.
• One of the first operational machines of this
  type was the Saturn V guidance computer,
  developed in the 1960s. Space shuttle is another
  example
• Fly-by-wire aircraft exhibits a very high degree
  of fault-tolerance in their real-time flight
  control computers. For example the Airbus
  Airliners

18
Introduction

• High-availability computers can tolerate an
  occasional error or very short delays (on the
  order of a few seconds), while error recovery is
  taking place.
• Example applications are telephone switching and
  transaction processing for banks, airline
  reservations, etc.
• Tandem Computers, Inc used a design of a
  distributed system with a sophisticated form of
  duplication.
• SUN's ft-SPARC and the HP/Stratus Continuum 400
  are systems that contain redundant processors,
  disks and power supplies, and automatically
  switch to backups if a failure is detected.

19
Introduction (contd.)

• More recent pushes for FTC
• Moores Law, complexity in processors
• Fault tolerant mechanisms
• in HW (more studied)
• in SW (more recent, more debated)

20
Introduction (contd.)

• Intuitive concepts
• Reliability continues to work
• Availability works when I need it
• Safety does not put me in jeopardy
• Performability maintains same performance in
  spite of failures
• Maintainability do not take much time to repair

21
Introduction (contd.)

• The two most common ways industry expresses a
  systems ability to tolerate failure are
• Reliability
• Availability
• In modern distributed systems other measures can
  be used such as outage time (or down time)

22
Terminology and definitions

• Reliability (time interval)
• R(t) conditional probability that a system is up
  in the interval 0,t given that it was up at
  time 0. Measured by MTBF MTTF MTTR
• Availability (time point)
• A(t) probability that a system is operating
  correctly and is available to perform its
  functions at the instant of time t. Measured by
  MTBF/(MTBFMTTR)
• Availability can be high, even if the system has
  frequent periods of inoperability if time to
  repair is low.

Up means system provides the required
functionality
23
Beyond Fault Tolerance
Server
Hw

• Distributed Systems -while cost of HW and SW
  drops, down time cost increases every year
• Availability is a good metric but outage minutes
  may be more useful in some cases (it can be
  measured)

Sys-sw
App-sw
Network
Hw
Sys-sw
App-sw
Client
Hw
Sys-sw
Industry has focused mainly on Hw faults but
those are not the only ones
App-sw
Customer view of 7x24
24
Break
25
Fundamental Principles

• Redundancy
• Addition of extra parts in a systems design to
  allow it continue functioning as intended in
  spite of failures
• Providing redundancy is key in fault tolerant
  computing
• Hardware redundancy
• Software Redundancy
• Time Redundancy
• Information Redundancy

26
Summary of FTC Techniques
Well cover FTC basic and advanced techniques
some research work and case studies
27
Fundamental Principles (contd.)

• Hardware Redundancy
• Low level
• Logic level - Self checking circuits, parity bit
  code
• High level
• Triplicate or use 5-copies of a computer (as in
  space shuttle)

28
Fundamental Principles (contd.)

• Software Redundancy
• Use two different programs/algorithms
• Time Redundancy
• Re-compute or redo the task and compare the
  results
• May or may not use the same hardware/software
• Information Redundancy
• Backup information
• Use of Error Correcting Codes (ECC)

29
Fault-Error-Failure concept

• Intuitive definitions
• Fault
• An anomalous physical condition caused by a
  manufacturing problem, fatigue, external
  disturbance (intentional or un-intentional),
  design flaw,
• Error - Effect of activation of a fault
• Failure - over-all system effect of an error
• Fault -gt Error -gt Failure

Bit stuck at
Incorrect data at ALU
Incorrect balance, system crash
Not all errors lead to failures!!
30
Faults/Failure Classification
Fault -gt Error -gt Failure
31
Containment Zones

• Containment Zones are used to limit error
  propagation in a system
• Barriers reduce the chance that a fault or error
  in one zone will propagate to another
• A fault-containment zone can be created by
  providing an independent power supply to each
  zone
• The designer tries to electrically isolate one
  zone from another
• An error-containment zone can be created by using
  redundant units and voting on their output
  (details later)

32
Barriers in HW/SW
Barriers constructed by design techniques for
fault avoidance, masking, tolerance
Better name for fault tolerance is error tolerance
33
Fault Modeling

• Fault models at different levels (HW)
• Process level
• Transistor level
• Gate level
• Function level
• System level

VLSI Manufacturing (Important but not covered
here)
We will discuss fault/failure models mainly at
high levels (from function to system level) in
the course
34
System Level Fault Modeling

• Distributed System
• Must consider HW/SW faults communication errors

Communication Channel (HW/SW)
Client HW/SW Redundancy
DataBase HW/SW Redundancy
Server HW/SW Redundancy
Time redundancy Information redundancy
Information redundancy
35
Fault Modeling (contd.)

• High-level failure models (process or system
  level failure). General classification
• crash failure - a faulty processor or system
  stops permanently
• omission failure - a faulty process omits
  inputs/outputs some times but when it works, it
  works correctly
• timing failure - inputs/outputs are delayed or
  arrive too early
• Byzantine failure (or arbitrary failure) - a
  faulty processor can exhibit arbitrary behavior
  including malicious nature

36
Failure Rate

• Bath tube curve (for Hardware)
• The rate at which a component suffers faults
  depends on its age, the ambient temperature, any
  voltage or physical shocks that it suffers, and
  the technology

Burning in used to avoid this zone
constant
Normal lifetime
20 weeks
5-25 years
37
Failure Rate

• Empirical formula for failure rate (in normal
  lifetime stage) for HW systems
• ? LQ(C1TVC2E)
• LLearning factor (maturity of technology)
• QManufacturing process quality factor
• TTemparature factor
• VVoltage stress factor
• EEnvironmental shock factor
• C1C2 Complexity factor ( gates, pins in
  package)

38
Failure Rate

• In most calculations of reliability, a constant
  failure rate ? is assumed, or equivalently the
  exponential distribution for the component
  lifetime T.
• There are cases in which this simplifying
  assumption is inappropriate
• Example - during the infant mortality and
  wear-out phases of the bathtub curve
• In such cases, the Weibull distribution for the
  lifetime T is often used in reliability
  calculation

39
Fault Tolerance and Reliability

• The effect of a fault tolerant design on
  reliability can be expressed as
• RsysP(no-fault)P(correct-operation
  fault)P(fault)

Maximized by fault intolerant design (proofs of
correct design, high quality components)
Coverage of a fault tolerance design over all
possible faults
For cost effectiveness, fault tolerant design
should target most likely faults
40
Dependability Evaluation

• Once a fault-tolerant system is designed, it must
  be evaluated to determine if its architecture
  meets reliability and dependability objectives
  using
• Analytical models
• Injecting faults

41
Modeling

• Mathematical formulation for quantitative
  analysis
• consider a large experiment with N systems at
  observation at time t
• Nc(t) - number of correctly operating systems
• Nf(t) - number of failed systems
• N Nc(t)Nf(t)
• Hence
• Reliability R(t) Nc(t)/N 1 - Nf(t)/N
• Unreliability Q(t) 1 - R(t)
• Derivative of reliability dR(t)/dt
  -(1/N)(dNf(t)/dt)
• dNf(t)/dt is called instantaneous failure rate of
  the component

42
Modeling (contd.)

• Reliability Modeling
• System model, concentrating on reliability aspect
• The structure of a system is used to model its
  reliability
• Models
• Combinatorial Models
• Markov Models

43
Modeling (contd.)

• Combinatorial Modeling - Probabilistic Techniques
• Idea Express reliability of a system as a
  function of reliability of its components
• Construction models
• series
• All components must work correctly
• No redundancy
• parallel
• Only one of the components must work correctly
• High redundancy

44
Modeling (contd.)

• Markov Models
• Many complex problems cannot be modeled easily
  in combinational fashion
• Use Markov models (aka Markov chains)
• Repair is very difficult to model
  combinatorially
• Markov models can be applied to modeling
  reliability, availability etc.

45
Modeling (contd.)
Markov Models STATE Represents all that must be
known to describe the system at a given instant
in time E.g. for reliability Each state
represents a distinct combination of faulty and
fault-free modules (e.g. 101, 1OK, 0fault)
TRANSITION Changes of state that happen in
system Over time as failures occur, system goes
from one state to another State changes are
given probabilities (e.g. prob. of failure, etc.)
Transitions are probabilities
46
Fault Injection

• Fault injection artificial creation of failure
  sources
• study system behavior under faults/errors
• what is the effect of faults? how fast do they
  propagate?
• is recovery code correct?
• is our failure model accurate?
• if failover / reconfiguration effective?
• Limitations
• Not all failures can be injected
• Year-long MTTFs impossible to study

47
Fault Injection

• Fault Injection is used to test Fault Tolerant
  Systems
• Simulation
• Prototyping

48
Topic List Discussion
Which topics which you havent seen before,
should we cover?
49
Course Format

• Discussion on course format
• Regular lectures
• Student Participation
• Workshop