Topic:
computational mail
Topic:
replicated data
Topic:
updating information in Clearinghouse and Grapevine
Topic:
updating information in a distributed system
| |
Summary
Rumor mongering is repeated updates until almost everyone knows. Clearinghouse compared databases at random in case rumor mongering failed.
Two-phase gossip uses timestamps and read quorums to survive site crashes, network partitions, and communication errors. (cbb 8/06)
Subtopic: eventual update 
| Quote: rumor mongering -- mail updates until they're old news at multiple sites [»demeA8_1987]
| | Quote: Clearinghouse's update algorithm only requires eventual delivery of repeated messages [»demeA8_1987]
|
Subtopic: deletion
| Quote: use death certificates to delete items -- retain dormant certificates [»demeA8_1987]
|
Subtopic: random, exact updates
| Quote: use anti-entropy as a backup for rumor mongering [»demeA8_1987]
| | Quote: anti-entropy compares databases at random -- is much slower than direct mail [»demeA8_1987]
| | Quote: anti-entropy by recent updates and comparing checksums [»demeA8_1987]
|
Subtopic: random updates
| Quote: epidemic algorithm for peer-to-peer broadcast; each node randomly forwards messages a limited number of times [»eugsPT5_2004]
|
Subtopic: locality
| Quote: favor nearby servers for anti-entropy and rumor mongering updates [»demeA8_1987]
|
Subtopic: two phase gossip
| Quote: two phase gossip--replicated data objects with checkpoint and event history; roll-forward an object when all sites have complete histories [»heddA10_1989]
| | Quote: two phase gossip best for systems that use distributed and replicated event histories for large data objects [»heddA10_1989]
| | Quote: in two phase gossip, each site keeps timestamp vectors that indicate the completeness of its histories and every other site's histories [»heddA10_1989]
| | Quote: two phase gossip uses quorum consensus; a query uses a quorum to gather a complete history for an object; an update writes a quorum [»heddA10_1989]
| | Quote: two phase gossip survives site crashes, network partitions, an message delays, reorderings and losses
|
Related Topics
Topic: computational mail (11 items)
Topic: replicated data (51 items)
Topic: updating information in Clearinghouse and Grapevine (11 items)
Topic: updating information in a distributed system
(50 items)
|
