[Pvfs2-developers] Re: noncontig-test

[Sam Lang]

On Aug 9, 2007, at 11:24 AM, Rob Ross wrote:

> we've talked about this off-and-on for quite some time, providing a  
> mechanism for sending an unexpected message that sets the server up  
> to receive a large incoming request. is it time to implement this?

What about putting the looping in the state machine for IO?  From the  
server's perspective, it looks like multiple IO operations, but from  
the client its just adjusting the next state action after the IO  
completes.  This only fixes the problem at hand, but its a lot easier  
to implement.

We might also try to reduce the size of each segment's encoding.  If  
the type is indexed, the optimal encoding size should be 16 bytes (5x  
smaller).

-sam

>
> rob
>
> Robert Latham wrote:
>
>> - a way to announce to PVFS "hey, i'm about to send you a giant
>>   request".  I'm 100% ok with a rendezvous approach if those
>>   additional messages mean a single call to PVFS_sys_io
>