Built in fault tolerance
CPFS uses multiple servers to provide redundancy. Failover is immediate upon server failure. Upon recovery, data automatically resynchronises with other servers.
Data are stored in a manner similar to RAID-5 disk arrays, with only 25% redundancy. High speed is maintained.
Run fully in user space
This feature makes the file system nearly independent of the kernel version. All that is required is a Linux kernel supporting FUSE (Filesystem in USErspace.
Easy to understand
The design of the system is simple and fully documented, allowing it to be modified as needed.
Easy to build
Get a docker image, slightly modify the configuration and your package is ready.
A FUSE-based client is used to view a single tree of files, with full metadata consistency and close-to-open data consistency for application compatibility.
I/O bandwidth and storage capacity can be expanded as needed.
Support two permission models
Filesystem access permissions may be checked in either metadata servers or in clients.
The system has been extensively tested with workloads for different applications and failure cases.
Prebuilt for many distributions
We have prebuilt and tested CPFS on CentOS, Fedora, Ubuntu, and Debian, allowing easy deployment.