| 
|
Topic: concurrency control by sequencers
topics >
computer science >
distributed database
Topic:
concurrency
Topic:
concurrency control by monitors
Topic:
critical regions
Topic:
function call
Topic:
implementing distributed systems and applications
Topic:
index sequence for array access
Topic:
synchronized processing
Topic:
timestamps
Summary
Interacting multi-processors will retain consistency if sequence constraints are followed. Unique sequences can be generated by a time clock with location stamp. This is analogous to the ticket machines found in bakeries and butcher shops. Sequencers can provide synchronization for distributed database updates. They allow local modification without global synchronization. (cbb 5/80)
Subtopic: assigned order
Quote: a serializer should prevent all other access to a protected resource [»atkiR_1977] Quote: serializability only requires some serial order is equivalent to an interleaved operation [»bernPA5_1978] Quote: a distributed system can use a global virtual clock for synchronization and global progress [»jeffDR7_1985] Quote: V uses time servers to serialize access to resources; better than monitors since independent of client [»cherDR4_1984] Quote: CBCAST for casual broadcast to maintain a partial ordering system wide [»birmKP2_1987]
Subtopic: event counts
Quote: synchronization by controlling the sequence of significant events instead of by using mutual exclusion [»reedDP2_1979] Quote: with a serializer, processes do to have to signal other processes Quote: with a serializer, the condition for resuming execution is explicitly stated when entering a queue [»atkiR_1977] Quote: a sequencer assigns an order to events occurring in the system Quote: reliable notification by a logger process that handles all messages; logs messages; adds an issue number to published notices (archived) [»cherDR4_1984]
Subtopic: tickets
Quote: an eventcount is a communication path for signaling and observing the progress of concurrent computations [»reedDP2_1979] Quote: eventcount solutions to synchronization often avoid multiple writers to the same variable [»reedDP2_1979] Quote: producer/consumer synchronization uses eventcounts to maintain the relative ordering; instead of mutual exclusion [»reedDP2_1979]
Subtopic: roll-back
Quote: a sequencer delivers a non-decreasing integer for each ticket(s) operation [»reedDP2_1979] Quote: sequencers use a ticket operation that mimics the automatic ticket machines used in bakeries [»reedDP2_1979] Quote: unlike eventcounts, a sequencer does not allow concurrent ticket operations [»reedDP2_1979]
Subtopic: distributed update
Quote: lookahead-rollback synchronization--execute until a conflict is discovered then roll back the offending processes and reexecute [»jeffDR7_1985] Quote: if a process receives an earlier message, it rolls-back and sends out antimessages to cancel its recent communications [»jeffDR7_1985] Quote: in Time Warp, every message has an antimessage that annihilates it if both are in the same queue [»jeffDR7_1985]
Subtopic: update problems
Quote: a single transaction generates one local action followed by as many update actions as there are sites [»bernPA5_1978] Quote: site n updates its local copy of a data item iff its local timestamp is older than the requesting transaction's timestamp [»bernPA5_1978] Quote: a site requesting an update action, waits for ACCEPT messages with later timestamps from all other sites; ensures all up-to-date [»bernPA5_1978] QuoteRef: bernPA5_1978 ;;157 four different protocols for different levels of synchronization; simplest depends on pipelined messages Quote: with protocol's P2 & P3, a site broadcasts an update request; while waiting, it can process unrelated transactions [»bernPA5_1978, OK] Quote: SDD-1 assigns a transaction protocol by a protocol table that embodies a detailed analysis of how classes interfere with each other [»bernPA5_1978]
Subtopic: crowds vs. counters
Quote: cache coherence by ordering writes to the same location does not satisfy sequential memory consistency since different locations may use different orderings [»adveSV12_1996]
Quote: if use crowds instead of counters for synchronization, can determine which processes are currently using a resource
[»atkiR_1977]
Group: distributed database (6 topics, 194 quotes)
Topic: concurrency (33 items)
Topic: concurrency control by monitors (24 items)
Topic: critical regions (58 items)
Topic: function call (28 items)
Topic: implementing distributed systems and applications (41 items)
Topic: index sequence for array access (16 items)
Topic: synchronized processing (35 items)
Topic: timestamps
(19 items)