To understand the effectiveness and/or efficiency of a protocol implemented in distributed storage systems, it is often desirable to evaluate it in a quantitative way under a hypothetical model.
Obviously, every model is an approximation to the reality and there is not an all-purpose one. I expect to do a mini survey of the models of distributed storage systems in the literature. Specifically, I am interested in the models on system throughput, read/write rate, read/write pattern, read/write duration (interval), read/write latency, communication delay, failure pattern, and more professional ones such as data consistency, quorum size, recovery mechanism, and so on.
Therefore, I am requesting references, including but not limited to
- statistical reports of commercial/open source distributed storage systems
- hypothetical models used in research papers.
a typical example: costs and availability of replicated services which studies availability, consistency, partition.
- experimental results given in research papers, talks, and technical reports.
typical examples: Dynamo in Amazon [section 6] and PNUTS in Yahoo! [section 5].
- personal comments on such topics from blogs, non-research articles, and you.