PDF Publication Title:
Text from PDF Page: 016
16 1.3. MOTIVATION provide key-value stores and coordination services3. 1.3 Motivation Despite becoming the de facto approach to consensus in distributed systems, Paxos is not without its limitations. Firstly, Paxos is notoriously difficult to understand, leading to much follow up work, ex- plaining the algorithm in simpler terms [PLL97, Lam01a, OO14, VRA15] and filling the gaps in the original description, necessary for constructing practical implementa- tions [CGR07, BBH+11]. This dissonance between the theory and the systems communities is best illustrated by the following quotes: The Paxos algorithm, when presented in plain English, is very simple. [Paxos] is among the simplest and most obvious of distributed algorithms. - Leslie Lamport [Lam01a] Paxos is exceptionally difficult to understand. The full explanation is noto- riously opaque; few people succeed in understanding it, and only with great effort. . . . In an informal survey of attendees at NSDI 2012, we found few people who were comfortable with Paxos, even among seasoned researchers. We concluded that Paxos does not provide a good foundation either for system building or for education. - Diego Ongaro and John Ousterhout [OO14] Secondly, the reliance on majority agreement means that the Paxos algorithm is slow to reach decisions, as each requires a round trip to/from many participants. By involving most participants in each decision, a high load is placed upon the network between participants and the leader itself. As a result, systems are limited in scale, often to three or five participants4, as each additional participant substantially decreases overall performance5. It is widely understood that Paxos is unable to reach an agreement if the majority of participants have failed. However, this is only part of the overall picture, failure to reach agreement can result not only from unavailable hosts but also network partitions, slow hosts, network congestion, contention for resources such as persistent storage, clock skew, packet loss and countless other scenarios. Such issues are commonplace in some systems, 3Applications include databases such as HBase (hbase.apache.org) or MongoDB (mongodb.com) and orchestration tools such as Kubernetes (kubernetes.io), Docker Swarm (github.com/docker/swarm) and Mesos (mesos.apache.org) 4For example, Chubby reaches consensus between a small set of servers, typically five [CGR07]. Likewise, Raft clusters typically contain five servers [OO14, §5.1] 5This effect can be seen for example in [MJM08, Figure 8]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 | RSS | AMP |