A Resilient, Scalable Architecture
1 An Infinit network is composed of 5 nodes with 3 redundant copies.
2 Node E crashes with one of the 3 copies.
3 Infinit ejects the failed node.
4 Infinit automatically replicates the missing copy onto another node.
5 Two new nodes join the network.
6 A copy is rebalanced to one of the new nodes.
Unlike most distributed systems relying on a master/slave model, Infinit benefits from a truly decentralized architecture (peer to peer).
This model protects Infinit from single points of failure and bottlenecks.
Fault tolerant and self-healing
Every data item is stored several times (configurable by the administrator) in order to ensure durability.
Should a failure occur, Infinit will automatically adapt the number of redundant copies to return to a healthy state.
Programmable For Your Applications
Easily activate and configure policies for your needs.
Choose between a replication system by giving a replication factor, erasure coding or none.
If activated, all the data items stored are encrypted (RSA/AES-256) and an access control mechanism is provided to manage keys.
Activate compression to reduce storage consumption depending on the nature of the data being stored.
Control the placement of the redundant pieces of data through several policies: rack-aware, region-aware (different data centers), etc.
Provide a policy to activate deduplication in order to avoid storing identical pieces of data.
Browse through the different versions of a file and rollback if necessary.
Adapt to your needs
Deploy low level virtual block device interfaces that can be formatted in the file system of your choice (ext4, NTFS, etc.) and used for non-competition flows such as databases.
Access a hierarchical organization of files through POSIX-compliant file system interfaces that provide advanced functionalities such as access control, versioning, etc.
Use an object storage interface (Amazon S3, OpenStack Swift) to access and store unstructured data items aka blobs.