I have implemented a distributed renderer using the PovRay engine. I have
only gotten a 67% (best) (%60 worst so far 16 min vs. 40 min. The ideal max
is 80%) speed improvement on 5 nodes. One of the major wastes of time is
that each node parses the data and the parsing time of the image is
non-trivial. Also the image is split into more chunks than nodes for load
balancing reasons (So if there are 100 chunks time parse time which is
static, you see the problem). Also I have access to a lot more nodes and I
would like to use them but the benefit of them is going to decrease unless
I can get rid of the parse time (as chunk size goes down (aka more
chunks/nodes) parse time has more of effect).
So what I need is a way to parse once (by the master node) and render from
the pre-parsed file.
A) Has any one created a way to extract/read pre-parsed data?
B) Know how to do it with no or about 2 days of coding?
Any help is appreciated!!
-James
Post a reply to this message
|