RE: [dtn-interest] on ditching fragmentation

Hi all,

Susan F. Symington wrote:

> One problem with it, however, might be that it isn't related to the
> capabilities of the network. If a link cannot routinetly stay up long
> enough to transmit data of that length, then there's no way to use
> reactive fragmentation to move data along that link. If a link routinely
> stays up for much longer than it takes to transmit data of that length,
> then there is a lot of wasted bandwidth carrying all those embedded
> hashes. If segments are too large, then lots of data is lost in the
> "interrupted" segment when a connection is lost and that data needs to be
> resent. If segments are too small, lots of bits are wasted with the many
> signed hash values that are interspersed throughout the payload.

I'm really having trouble following the rational for fragmentation (please excuse me if I don't use the correct terminology), so I thought I'd detail an instance in which it would be useful as I see the argument:

• On the "source" node, there is an application that creates a 10 MB bundle addressed to the "sink" node, and submits it to the local bundling agent.
• The local agent has links to *many* different nodes that come into contact intermittently and randomly (unpredictably). There's never enough time to transfer more than 1 MB of data during a single contact.
• Because the local agent doesn't know when a particular node will be in contact again (or even *whether* that node will be in contact again) it needs to fragment the original bundle into 10 or more fragments, transferred independently, in order to transfer it to the outside world successfully.
• It transfers each of these fragments to the nodes it comes into contact with in order of contact: ABACDABCAE. Now A has 4 fragments, B & C have 2 each and D & E have 1 each.
• A, B, C, D and E propagate the fragments through the network, and eventually they come to the "sink" node, which waits until it has all 10 (or however many) fragments, collates them, and forwards the original 10 MB bundle.

In this scenario, the original bundle will not be transmitted across the network at all without fragmentation.

Please correct me if I'm wrong in this interpretation. :)

Peter Brett

Do you need help?
In addition to our free technical library, Pantek provides professional IT services to companies who utilize Open Source software: Learn more about Pantek's Technical Support offerings Place an order for 24/7/365 Technical Support here Learn more about Pantek, who we are, and what we do Chat with a Live Operator Related mailman.dtnrg.org Links dtn-interest Request more information about Pantek

P.S. What happens if node D dies a horrible death sometime inbetween collecting a fragment from the source and delivering it to the sink?

-- 
E-mail:       peter@peter-b.co.uk
Website:      
http://www.peter-b.co.uk

_______________________________________________
dtn-interest mailing list
dtn-interest@mailman.dtnrg.org
http://mailman.dtnrg.org/mailman/listinfo/dtn-interest