One RPC protocol to rule them all

jeffpc / nx01

testing

MIT License

1 stars 0 forks source link

One RPC protocol to rule them all #85

Open jeffpc opened 8 years ago

jeffpc commented 8 years ago

Instead of having several different RPC protocols for the different components communicating, we should probably have just one and the different daemons can return -ENOTSUP or -ENOSYS for the commands they don't like.

The four different modes of communication:

admin util <-> client daemon (to setup caches and connection info)
admin util <-> server daemon (to setup storage)
fs fuse module <-> client daemon (file access)
client daemon <-> server daemon (synchronization)

Thynix commented 8 years ago

Do we expect any overlap between endpoints' support?

jeffpc commented 8 years ago

At the very least, NOP should (IMO) be in all of them. Aside from that, the RPCs for managing storage will likely be used for both client and server daemons. (After all, a cache is a storage device that's just used differently - but the setup is the same.)