Distributed consensus

PDF Publication Title:

Distributed consensus ( distributed-consensus )

Previous Page View | Next Page View | Return to Search List

Text from PDF Page: 144

144 8.3. IMPLICATIONS OF CONTRIBUTIONS Likewise, we see another example of this multi-pathway approach in Paxos revision C (Chapter 5), where a proposer in phase one of epoch e can satisfy its intersection require- ment with the phase two of a previous epoch f either: • by receiving promises from at least one acceptor in each quorum Q ∈ Qf2 ; or • by receiving a promise with a proposal from an epoch g where f ≤ g ≤ e. In section 3.10, we allow proposers the option to copy an existing proposal instead of starting a new proposal. In section 7.4, we propose a hybrid approach consisting of using epoch allocation by an allocator, value mapping or recovery for the minimum epoch and epoch allocation by pre-allocation or voting for all other epochs. 8.3.2 New progress guarantees Paxos focuses on a single progress property: guaranteed progress regardless of an algo- rithm’s current state. Whilst useful for comparing fault-tolerance of algorithms under worst case conditions, this gives us little information regarding the overall availability of the algorithms. Over the course of this thesis, we have demonstrated algorithms with new progress properties depending on the system state. In this section, we will consider various examples. A proposer can terminate in one round trip to a phase two quorum of acceptors if phase one has been completed and no acceptors in the quorum promise or accept since phase one (§4.1). At the extreme, this quorum may only contain only one acceptor, as described in §4.3.2. The tradeoff for optimising the phase two quorum is decreased performance and availability for the phase one quorum. This tradeoff may be worthwhile when combined with Multi-Paxos, which seldom executes phase one compared to phase two (§4.3.3). A proposer can terminate in one round trip to a phase two quorum of acceptors if it is the first to propose and has been allocated emin, since this proposer is able to bypass phase one. More generally, since each proposer during phase one is required to intersect only with the phase two quorums of previous epochs, the intersection requirements build up as epochs increase (§4.2.3). Our progress guarantee for Classic Paxos relies on a single proposer executing the proposer algorithm. In practice, this is often achieved by designating one proposer as distinguished and thus relying on synchrony to detect failure of the designated proposer. In sections 7.2 and 7.3, we proposed the allocation of epochs to values and epochs allocation by recovery. Both these new algorithms can guarantee termination when multiple proposers are executing the proposer algorithm with the same candidate value. For example, this was illustrated in Figure 7.5.

PDF Image | Distributed consensus

PDF Search Title:

Distributed consensus

Original File Name Searched:

UCAM-CL-TR-935.pdf

DIY 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)