Path: utzoo!yunexus!geac!syntron!jtsv16!uunet!lll-winken!lll-lcc!ames!mailrus!uflorida!gatech!hubcap!frazier
From: frazier@CS.UCLA.EDU
Newsgroups: comp.parallel
Subject: Wormhole Routing
Keywords: wormhole, virtual cut-through
Message-ID: <3384@hubcap.UUCP>
Date: 31 Oct 88 13:16:54 GMT
Article-I.D.: hubcap.3384
Sender: fpst@hubcap.UUCP
Lines: 22
Approved: parallel@hubcap.clemson.edu


Somebody asked what wormhole routing (Chuck Seitz) was, and what
makes it different from virtual cut through (Kleinrock).  Well,
there is no difference.  Virtual cut through is a scheme
whereby one begines to forward a packet as soon as it is routed
and the necessary output port is free, i.e. before the entire 
packet has arrived.  This is basically an optimization to
current packet switching practice, and has been implemented on
the Post Office (Stevens, Robinson, Davis), along with several
other current communications chips.  Wormhole routing, proposed
by Seitz and Dally on the Torus Chip, is simply virtual cut
through where there isn't room on the node for the entire packet
to be stored, so that one is forced to do the cut through.  The
advantage of not supporting "regular" packet switching is that
one needs only minimal buffer space at each node.

Greg
--------------------------------------------------------------

Greg Frazier	    o	Internet: frazier@CS.UCLA.EDU
CS dept., UCLA	   /\	UUCP: ...!{ucbvax,rutgers}!ucla-cs!frazier
	       ----^/----
		   /