Re: [MirageOS-devel] papers on immutable storage

[Thomas Gazagnaire <thomas@xxxxxxxxxxxxxx>]

cc/ the Mirage list as the list of pointers can be interested to other.

Some talks to understand weak consistency models: 
- consistency models vs. baseball games: 
http://cacm.acm.org/magazines/2013/12/169945-replicated-data-consistency-explained-through-baseball/abstract
- basho, immutability changes everything: http://vimeo.com/52831373

(a very short list) of related papers to branch consistency: 
- bayou: http://zoo.cs.yale.edu/classes/cs422/2013/bib/terry95managing.pdf
- cloud type: http://research.microsoft.com/apps/pubs/default.aspx?id=163842 
(and other papers of Burckhard around the fork/join model)

To understand what values are created when a new entry is added: look at 
radix/prefix/suffix trie (or tries) papers
- http://www.comp.nus.edu.sg/~stevenha/visualization/suffixtree.html#
- http://en.wikipedia.org/wiki/Radix_tree
Basically every block in a normal heap is translated to a new file (in the Git 
backend at least).

Best,
Thomas

> Thomas,
> 
> Can you recommend any reading on Irmin like database? I want to understand 
> better the use-cases and how they are handled. I want specifically to look at 
> two opposite cases, one where the database is distributed between many nodes 
> and many clients write to it so concurrency and consistency are a challenge 
> and another where there the database is on one node and only one client.
> 
> Also what happens in Irmin backend when a new key/value is added? Is a new 
> file created every time (if the value is different every time)? Is there 
> support for encryption?
> 
> Thanks,
> 
> Gregory


_______________________________________________
MirageOS-devel mailing list
MirageOS-devel@xxxxxxxxxxxxxxxxxxxx
http://lists.xenproject.org/cgi-bin/mailman/listinfo/mirageos-devel