[OpenAFS-devel] RX Performance

[Ken Hornstein]

>i dont believe tcp is the answer.  in the vafs work, rx was replaced by
>aal5 for data transactions.  aal5 is more like udp than tcp.   the aal5
>datagrams didnt consist of a number of standard rx datagrams.  it was
>simply a biggish packet (or pdu) with an rx header.  very easy to handle.
>there are some other limits in the fileserver that probably 'held back'
>filserver performance.  if i remember, there is some strange 45k limit
>in the fileserver.

The problem with this that I see is that all of the serious thinking about
how to make a good-performing stream protocol in the last 10 years has been
done with TCP, and _not_ done with RX.  So we could duplicate that in
RX, which would be hard, or just use TCP, which would be easy.

>further, use of tcp would tend to limit scalability/number of clients.

I had this discussion with Love in the AFS Workshop; he pointed out that
most modern OS's have solved the problem with select()ing on large numbers
of descriptors (Solaris has /dev/poll, the BSDs have kqueue, and Linux
has something which I forget right now).  That seems to be the big one.

--Ken