Amazon Company Reliability

Key problems and solutions of building a highly available, scalable and reliable distributed system by the name Dynamo are addressed by the Author in the article. Reliability will be achieved by using the replication. Sloppy quorum and anti-entropy with Merkle trees will be used for management impermanent, permanent failures. Incremental scalability will be achieved by use of constant hashing to divide the keys. The article addresses the problem that is based on building a production level highly available system for storing key-value pairs as suggested by the title. This paper work is important since of its specific needs which are single-purpose key-value pair storage system which focuses fully on accessibility. From existing approaches …show more content…
In my own opinion, the problems are significant since Amazon company have been facing some of the problems as deliberated by the authors such as bad loading and adding something in the shopping cart. Conversely, the contributions are not significant since as they mention in the article the Dynamo can be characterized as zero-hop DHT, even though it gives them scalability when it comes to adding or removing nodes easily, scalability of the dynamo is limited since they have selected individually node knows about all other DHT nodes. This will be very expensive as it cannot scale to thousands of nodes. The claims presented can be invalid because by allowing the application developer rights to choose value that are for tunable parameters or the best approach for reconciliation as a benefit. But this might be challenging since one should have knowledge on the existing application and parameters so that they can make sensible evaluations and choose the right strategy for her application. For the system to achieve high convenience they have to relax in consistency and this makes dynamo not suitable for any system that requires strong consistency such as banking …show more content…
They claim that Amazon is not consistence, available and scalable to the users. In my opinion, I support the need for alternative applications that can reduce in consistency guarantees so as to achieve more suitability and makes simple key-value inquiries since from the case presented in the article, I do think they could have done it better on their evaluation section but instead they fail to achieve on the investigation on all their target goals. Clarifying, use of homogeneous set of servers and setting heterogeneity as a target in their system has been used in the evaluation section, Moreover, in regard to the argument made, the authors assume that there are no malicious codes and having in mind that buggy codes can be produced by some of the unskilled