<div>Brendan,</div>
<div> </div>
<div>I'm a day late but maybe not a dollar short :-) When I read the original question, I was going to ask, "do the compute (render) nodes push their results when ready, or does the head (view) node pull?" and from the subsequent discussion and clarifications it seems to be the former. And yeah what Mark said.
</div>
<div> </div>
<div>So if it were me, each compute node would send the (short) message, "I'm ready"; the head node would maintain a list of ready nodes, and pull from them sequentially ("Ok node number 7, upload now please"). That way the only collision is word-sized and not image-sized, and the overhead is trivial. But that would be me using FTP and tcsh :-) Dunno what you'd do specifically with the software you have.
</div>
<div> </div>
<div>Peter</div>
<div><br><br> </div>
<div class="gmail_quote">
<blockquote class="gmail_quote" style="PADDING-LEFT: 1ex; MARGIN: 0px 0px 0px 0.8ex; BORDER-LEFT: #ccc 1px solid">
<div class="Ih2E3d">(Hahn)<br></div>it's the _timing_ of the data. using bcast, you attempt to cause the<br>render nodes to, as simultaneously as possible, saturate their own<br>links, and therefore (N-1)-times oversaturate the viewer link.
<br></blockquote></div>