sciencekmfk.blogg.se - october 2023

The servers communicate via internal RPCs when necessary.

Metadata server: Maintains database of users, namespaces, and of course SFJ.

No knowledge of users/files/how those blocks fit together.

Block data server: Maintains a key-value store of hash to encrypted contents.

There are two types of servers relevant to this discussion:

Journal ID (JID): Monotonically increasing within a namespace.

It is an append-only record where each row represents a particular version of a file.

This is our big metadata database which represents our file system! Note that it doesn’t contain file contents, just blocklists.