How does write-behind work ?

[httaotao]

In this page write-behind, we can get this setence.

On a translator tree with write-behind, control flow is like this:
     ....
     5. write-behind adds the write buffer to its internal queue and does a STACK_UNWIND() towards the fuse-bridge.

...  Write-behind doesn't cache the write buffer, unless option flush-behind on is specified in volume specification file.

I am confused what happen? it seems a paradox and I coludn't understand. thanks.

[pranithk]

In this page write-behind, we can get this setence.

On a translator tree with write-behind, control flow is like this:
     ....
     5. write-behind adds the write buffer to its internal queue and does a STACK_UNWIND() towards the fuse-bridge.

...  Write-behind doesn't cache the write buffer, unless option flush-behind on is specified in volume specification file.

I am confused what happen? it seems a paradox and I coludn't understand. thanks.

Without write-behind every write that happens from fuse goes over network. Write-behind lies about completing this write and returns to fuse immediately saying write is successful and tries to send these writes in the background. If the write-requests are small, it also merges the contiguous writes to form bigger buffer so that it can send all of the data in one shot. So the purpose of write-behind is to reduce number of writes over network in fuse's write path.

I am yet to do a dev-session on this. Will cover the details when I do it.

[httaotao]

I get it , thanks