PDF Publication Title:
Text from PDF Page: 120
120 7.3. EPOCHS BY RECOVERY 7.3.2 Algorithm Algorithm 24: Acceptor algorithm for epochs by recovery. 1 2 3 4 5 6 7 8 9 10 11 while true do switch do case prepare(e) received from proposer if epro =nil∨e≥epro then epro ← e send promise(e,eacc,vacc) to proposer case propose(e,v) received from proposer if epro =nil∨e≥epro ∧(e̸=eacc ∨v=vacc)then epro ← e vacc ←v,eacc ←e send accept(e,v) to proposer Algorithm 24 outlines an acceptor algorithm for epochs by recovery. The only two differ- ences between this and the Classic Paxos acceptor algorithm are that accept messages now include the value (line 11) and an extra condition on accepting proposals (line 8). Specifically, acceptors do not overwrite accepted proposals with proposals of the same epoch but a different value. This is implemented on line 8 whereupon receiving a propose, the acceptor must check that they have not already accepted a proposal with this number but a different value. Algorithm 25 outlines a phase one for the proposer algorithm for Revision A with epochs allocated by recovery6. We have switched to epoch agnostic, quorum-based value selection (§6.1) as this approach is better suited to efficiently expressing epochs by recovery. There are three key differences between this algorithm and Revision A with quorum-based value selection (Algorithm 16)7. Firstly, as proposers are no longer required to choose from disjoint sets of epochs and track which have been used, E has been removed. Instead, epoch e is initially set to nil and is incremented before each use (line 4-7, Algorithm 25)89. Secondly, our implementation of possibleValues (Algorithm 26) includes an extra case to 6For simplicity, we are not varying quorums depending upon epoch thus Revisions B and C do not apply. 7Whilst not explicitly represented in the pseudocode, this algorithm also requires that phase two quorums of a given epoch must intersect. 8In contrast to our previous algorithms which were general to any epoch set E, it is now the case that E = N0. This approach has been chosen for simplicity however the algorithms could easily be generalised to any epoch set E. 9The current epoch e does not need to be in persistent storage for correctness, however, it would help proposers recovery quickly after failure.PDF Image | Distributed consensus
PDF Search Title:
Distributed consensusOriginal File Name Searched:
UCAM-CL-TR-935.pdfDIY PDF Search: Google It | Yahoo | Bing
Cruise Ship Reviews | Luxury Resort | Jet | Yacht | and Travel Tech More Info
Cruising Review Topics and Articles More Info
Software based on Filemaker for the travel industry More Info
The Burgenstock Resort: Reviews on CruisingReview website... More Info
Resort Reviews: World Class resorts... More Info
The Riffelalp Resort: Reviews on CruisingReview website... More Info
CONTACT TEL: 608-238-6001 Email: greg@cruisingreview.com (Standard Web Page)