Notes - MIECT
Computação Distribuída
Notes - MIECT
Computação Distribuída
  • Computação Distribuída
  • Introduction / Architecture
    • Distributed Systems
    • Architecture
    • Middleware Organizations
    • Processes
    • Threads
    • Virtualization
    • Clients
    • Servers
    • Migration
  • Communications
    • OSI Model
    • Middleware Layer
    • Types of Communication
    • Remote Call Procedure (RPC)
    • Sockets
    • Application-level Multicasting
  • Naming
    • Names
    • Addresses
    • Identifiers
    • Naming Systems
      • Flat Naming
      • Structured Naming
    • Internet Domain Name System (DNS)
    • Attribute-based naming - LDAP
  • Coordination
    • Clocks
      • Synchronizing without UTC
    • Reference Broadcast Synchronization – RBS
    • Happened-Before Relation
      • Logical Clocks
      • Vector Clocks
    • Mutual Exclusion Algorithms
    • Election Algorithms
    • Distributed Events Correspondance
  • Consistency & Replication
    • Replication
    • Performance and Scalability
    • Client-centric models
    • Replicates
    • Unicasting vs. Multicasting
    • Continuous Consistency
    • Protocols
  • Flaw Tolerance
    • Dependability
    • Terminology
    • Confidence vs. Security
    • Halting failures
    • Redundancy to mask failures
    • Consensus
      • Realistic
      • Consensus in arbitrary failures
      • Achieving failure tolerance
      • Distributed consensus
    • Failure Detection
    • Reliable RPCs
    • Distributed commit protocols
  • Python asyncio & Friends
    • Async
    • Sync vs. Async
    • Tools
  • Flask
    • Introduction
    • Python Requests
  • Containers
    • VM's vs Containers
    • OS Support
    • Building a container
    • Tools
    • Portability
    • Docker
      • Container
  • Map Reduce
    • Map Recude
    • Hadoop
    • Software Architecture
    • Task Scheduling
    • Comparison With Traditional Models
  • Cloud Computing
    • Cloud Computing
    • IaaS – Infrastructure as a Service
    • PaaS – Platform as a Service
    • SaaS – Software as a Service
    • Business Models
Powered by GitBook
On this page
  • Pre-requisites
  • Reformulation
  • Flooding
  • System model
  • Algorithm (round based)
  1. Flaw Tolerance

Consensus

Pre-requisites

In a group of failure-tolerant processes, each process that is not failing executes the same commands and in the same order as any other process that is also not failing.

Reformulation

Non-failing members of the group need to reach a consensus about which command to execute next.

Flooding

System model

Consider a group of processes P = {P1, ..., Pn}

Failure semantics Fail-stop, meaning, with detection of reliable failures.

A client contacts Pi asking to execute a command.

Each Pi holds a list of proposed commands.

Algorithm (round based)

In round r, Pi multicasts its set of Cmdr_i to everybody else.

At the end of r, each Pi merges every received command in a new Cmdr+1_i.

The next Cmd_i command is selected by a globally shared deterministic function.

Example

P2 receives all the proposed commands by all the other processes -> makes a decision.

P3 can detect that P1 crashed, but cannot know if P2 has received anything, i.e, P3 has no way of knowing if it has the same information as P2 -> cannot make a decision (same happens with P4).

PreviousRedundancy to mask failuresNextRealistic

Last updated 1 year ago