Memorandum of Understanding Chapter 8. Introduction. Processes/sites in distributed systems often compete and cooperate to achieve a common goal. Mutual trust/agreement is in high demand. Solution to the problem of bizantin agreement • First defined and solved by the report. • Source Sends its initial value to all other processors. • Processors pass their values to other processors and pass the received values to others. • During execution, faulty processors can merge by sending conflicting values. • However, if faulty processors dominate in numbers, they can prevent non-faulty processors from reaching an agreement.
• No faulty processor should exceed certain limits. • Pease has shown that it is impossible to reach an agreement in a fully connected network if the number of faulty processors exceeds “m” (n-1)/3. n = number of processors Introduction • Processes/Localization in distributed systems are often competing and cooperating to achieve a common goal. • Mutual trust and agreement are in high demand. • In distributed databases, there may be a situation in which data managers must decide whether to fix or cancel the transaction• If there is no error, it is easy to reach an agreement. • However, in the event of a failure, processes must exchange their values with other processes and transmit the values received by others several times to isolate the effect of a faulty processor. • Memoranda of Understanding make it possible to reach an agreement in case of failure. Performance aspects of memoranda of understanding • The following metrics are used: • Time: number of turns needed to reach an agreement • Message traffic: Number of messages exchanged to reach an agreement. • Memory overload: amount of information that needs to be stored in processors while the protocol is running. Classification of agreement problems. • In all the problems mentioned above, all non-defective processors must reach an agreement • For Byzantine and consensual problems, the agreement is of a single value • In the interactive problem of consistency, the agreement is on a number of common values.
• In the bizantine match problem, only one processor initializes the value, as in two other cases, each processor has its own initial value. Classification of tuning problems • There are three known convergence problems in distributed systems: • Byzantine agreement problem: • A single value must be agreed. • The agreed value is initialized by any processor and all non-defective processors must agree on this value. 2. Consensus problem: • Each processor has its own initial value and all non-faulty processors must agree on a single common value. 3. Interactive consistency issue: • Each processor has its own initial value and all non-faulty processors must agree on a set of common values. Tuning algorithm applications • Tolerant touch synchronization • Distributed systems require synchronization of physical clocks • Physical clocks have drift problems • TUNING protocols can help achieve a common cadence value.
. . .