Storage nodes and trackers for distributed file systems

1 , storage nodes

In a distributed file system, all files are stored in a storage node ( Storage Node ). A storage node typically corresponds to an actual directory of the disk in the house. We create multiple storage nodes with multiple servers, and the files are automatically replicated between these storage nodes according to the rules. In addition, we can create multiple storage nodes on the same physical machine, point to different disk space, achieve a certain degree of redundant backup, and a bit similar to the RAID mechanism.

2 , tracking device

The tracker is responsible for bridging the role between the client and the storage node. For the complex distributed file system, the complex storage node internal operation is transparent to the client, and the client will use the tracker to complete the Distributed File System .

3 , trackers, and storage nodes for a simple example diagram

the tracker (Tracker) is provided to the client for use by the tracker to find the appropriate storage node to provide to the client, and the client will read and write the file directly with the storage node server, a design that reduces the bottleneck of the tracker. And our LVS DR mode load balancer is a bit similar. :)

4 , tracker, and storage node internal interaction work

The tracker does a lot of work, such as:

Maintain storage node information and know which files are stored on which nodes;

Control file copy and distribution policies;

Implement load balancing and scheduling of storage nodes;

Detection of availability of storage nodes;

Implement failover of storage nodes;

...

Storage nodes and trackers for distributed file systems