Advertisement
Published in:
2014 | OriginalPaper | Chapter
Separating Data and Control: Asynchronous BFT Storage with 2t + 1 Data Replicas
Authors : Christian Cachin, Dan Dobre, Marko Vukolić
Published in: Stabilization, Safety, and Security of Distributed Systems
Publisher: Springer International Publishing
Activate our intelligent search to find suitable subject content or patents.
Select sections of text to find matching patents with Artificial Intelligence. powered by
Select sections of text to find additional relevant content using AI-assisted search. powered by
The overhead of Byzantine fault tolerant (BFT) storage is a primary concern that prevents its adoption in practice. The cost stems from the need to maintain at least 3
t
+ 1 copies of the data at different storage replicas in the asynchronous model, so that
t
Byzantine replica faults can be tolerated. This paper presents
MDStore
, the first fully asynchronous BFT storage protocol that reduces the number of replicas that store the payload data to as few as 2
t
+ 1 and maintains metadata at 3
t
+ 1 replicas on (possibly) different servers. At the heart of
MDStore
lies a metadata service built upon a new abstraction called “timestamped storage.” Timestamped storage allows for conditional writes (facilitating the implementation of the metadata service) and has consensus number one (making it implementable with wait-free semantics in an asynchronous system despite faults). In addition to its low replication overhead,
MDStore
offers strong guarantees by emulating a multi-writer multi-reader atomic register, providing wait-free termination, and tolerating any number of Byzantine readers and crash-faulty writers.