Flud backup
From Wikipedia, the free encyclopedia
 |
This article or section needs to be wikified to meet Wikipedia's quality standards. Please help improve this article with relevant internal links. (June 2007) |
Flŭd backup is a completely decentralized system for creating and maintaining online backup of data. Its unique architecture allows it to provide nearly infinite backup resources for free, in a way that is virtually immune from failure.
[edit] Architecture
flŭd is a completely decentralized peer-to-peer backup system. There is no central server or authority, no controlling company, organization, or individual. All participants in the flŭd network share resources using an enforced fairness mechanism; in order to gain storage resources, a participant must provide resources. These sharing relationships are symmetrical and cheat-resilient, and are sometimes referred to as smart contracts.
In order to tolerate a massive failure of participating nodes in the flŭd network, erasure coding techniques are used to protect data. As of January 2007, these techniques allow virtually all data to be recoverable from the network even when nearly half of the nodes have failed. While it is unlikely that half of all nodes in a fully-grown flŭd network (which could number in the millions) would fail at the same time, it is possible that some unlucky node could have a significant number of its trading partners fail in a correlated manor. The generous erasure coding techniques employed by each flŭd node should allow even such unlucky participants to recover all data in this unlikely case.
Individual nodes in the flŭd network are allowed to use any methods they choose to make decisions about where to store data and with who to enter into resource-sharing relationships. This agency allows for diversity in storage strategy. The flŭd backup prototype maintains localized trust records, which serve as a history of action of other known agents. These trust metrics are used to decide which nodes are most reliable, and each node tries to maximize its interactions with highly reliable peers while minimizing those with less reliable peers. This implies that users wishing to participate in the flŭd network should be willing to leave their computers on all the time; although flŭd's trust metrics are tolerant of intermittent and occasional failures, nodes which are not usually available will be seen as unreliable.
