marcus366 / simplified-gfs Goto Github PK

When client asks a new chunk for writing, Master Server should generate a new chunk for the current file, and then pick an active chunk server to store this chunk.

For the current version of GFS, we always use the first chunk server to store chunks, which can't be a good way to solve the problem.

What we want is the same as title.

We appreciate you implementing a generic hash table named gfs_hash_table_t.

It is just used for storing uuid of chunk to avoid duplication of uuid now, but the hash table is so powerful that it may be useful for any other reasons later.

on_clnt_newchk is a function being called while a client read/write a chunk and the chunk is full.

The Client should use a LRU cache to store the meta data, which can reduce the burden of Master Server and speed up the read/write.