Is FDMI asynchronous or synchronous?

[johnbent]

Can FDMI only operate asynchronously on data after it has been stored or can FDMI insert itself into the write path? I am asking because I was looking at the demo that the EU R&D team once did: https://github.com/Seagate/cortx-motr/tree/main/doc/fdmi_demo/demo-fdmi

[johnbent]

@huanghua78 answered this question in a different place. Pasting here:

Let me explain what this demo did.
(yes, "did", because some container operations are not supported now.)

Can FDMI only operate asynchronously on data after it has been stored or can FDMI insert itself into the write path?
The simple answer is: FDMI only operate asynchronously on data after it has been stored.

FDMI is a kind of subscribe/notification system. When some object or index operations are done in Motr, the FDMI source dock sends out notification messages (filtered by user-defined filters) to FDMI plugin dock. When the FDMI plugin gets the messages, it can inspect the data, retrieve the parameters of the original operations, and do something on the object/index.

The demo did:
The clovis app creates some object, write 4KB or 4MB data to this object.
The Motr FDMI source dock generates FDMI notifications and send them to plugin.

The plugin/app has two processes:
One process receives the FDMI notification, and classify the WRITE to be BIG or SMALL. Then send a user-defined message to another process with PIPE.

Another process reads the user-defined message from PIPE. According to this message, it inserts a KEY/VAL pair into the BIG container or SMALL container. Here, KEY/VAL is: FID_STRING/Dummy_STRING.

The clovis app completes write several operations. Then, it list the contents from the BIG container and SMALL container. The list contains BIG objects FID from BIG container, and SMALL objects FID from SMALL container.

So, you can see, this demo does not insert FDMI into the write path. It inspects the size of the object size from FDMI notifications, and inserts some KEY/VAL into a pre-defined container.