My favorites | Sign in
Project Home Downloads Wiki Issues Source
Project Information
Members
Featured
Wiki pages
Links

FineFS is a simple replicated filesystem, which aims to create data clusters. It is designed for web applications, but can be used for any other usage which needs data replication between machines.

  • Data consistency. Writings are atomic. You never get files that are in-modification state.
  • Robustness. Using failure detection and retry strategies, temporary shutdowns are managed without data loss. The fully decentralized design avoid any single point of failure.
  • Synchron/asynchron process. Files information are synchronously distributed across the cluster, while binary data are asynchronously replicated.
  • High-performance. Data are eventually replicated on local hard disks, for better optimization of basic operations (data read and write).
  • Easy to deploy and maintain. Based on well-known technics, FineFS is easy to set-up and doesn't need some extended maintenance.


Featured documentation


Why use FineFS?

For example, you have many frontal web servers. You need to share the same data over all machines.
You can set up a file server. But it is a single point of failure. If the file server fall down, you loose all data.
You may prefer to use a Storage Area Network. They usually provide some fail-safe mechanisms. But they are also very expensive, and you must set a high-speed link between the SAN and your servers because data are always accessed through the network.
You can use an external service like Amazon S3. But it may be a shame to have some empty disk space on your servers, and data access may be really slow.
It is possible to set up a distributed filesystem, like Coda, Ceph, POHMELFS, GlusterFS, GfarmFS, Tahoe, Moose FS, Chirp, GFS, ... But these systems may have some technical restrictions or different design goals (like computing grid), and they are hard to install and maintain.
FineFS is an easy and convenient mean to create a cluster. Data are accessed locally most of the time, or fetched over the network if necessary.

Powered by Google Project Hosting