> -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED]]On Behalf Of Sukru Tikves > Sent: Thursday, September 27, 2001 3:26 PM > To: [EMAIL PROTECTED] > Subject: [freenet-tech] One more thing... (Avarage Network Load) > > > One may argue that searching a rare key is a "rare" > operation and one can easiliy wait for "one hour" for > data, anyway. > > But do you know how many 404 pages are served in a > day? A failed query is a very good example of a query > that will traverse the entire network even in the best > case. > > So let's make simple calculations... > > Let's assume that Joe User makes 10 bad queries in a > day. (This is reasonable). And let's assume a VERY > small packet size (200 bytes) per query. Since every > node must process the query and a REPLY, the avarage > load is > > L = N * 2000 bytes/day > L = N * 0.185185... bits/sec > ( where N is the network size ) > > So for a N node network, every node must have at least > N * 0.0925 bps baud rate in order to survive. > > Let's put osme numbers > > 1,000 nodes = 185.185 bps > 100,000 nodes = 18,518 bps > 1,000,000 nodes = 185 Kbps > 10,000,000 nodes = 1.851 Mbps > > so we hope that Joe User will be able to get a 2 Mbit > line before freenet reaches 10,000,000 nodes for > "minimal" operation. > > (This numbers are ver OPTIMISTIC.) > > If we solved the rotung problem, the situation > changes. The avarage query will travel only logn > nodes. So > > L = logN * 2000 bytes/day > L = logN * 0.182185... bits/sec > > For 10,000,000 nodes with avarage of 11 connections > per node: > > L = 1.29 bits/sec (less tan two bits/sec). > > (This is BAD query load. This does not include good > queries and data transfers). > > We HAVE to fix routing!!!! :) > > I am not quite sure how you think routing is going to help with bad queries? If you are wishing to guarrenty that the document is returned if it is anywhere within freenet then you much search all of the nodes anyway, regardless of routing. If that is not the case then nodes can just discard messages as they reach their bandwidth limits (I seem to remember reading that this is the case now), and you get a best effort result. I am wondering however how the hops-to-live is affected when the search via the "best" node fails, and the node restarts the query on the second best. Is it the same as for the original search, decremented by 1, or what? I have read everything I could find and not found this. > ===== > Sukru Tikves, > Hacettepe University > Computer Engineering > Department, Ankara > > http://piskare.cjb.net > > __________________________________________________ > Do You Yahoo!? > Listen to your Yahoo! Mail messages from any phone. > http://phone.yahoo.com > > _______________________________________________ > freenet-tech mailing list > [EMAIL PROTECTED] > http://lists.freenetproject.org/mailman/listinfo/tech > _______________________________________________ freenet-tech mailing list [EMAIL PROTECTED] http://lists.freenetproject.org/mailman/listinfo/tech
