We approach the problem of handling storage congestion at store-and-forward (DTN) nodes by migrating stored data to neighbors. The proposed solution includes a set of algorithms to determine which messages should be migrated to which neighbors and when. It also includes an extension to the DTN custody transfer mechanism enabling a "pull" form of custody transfer where a custodian may request custody of a message from another custodian. This approach allows us to decouple the problem of storage allocation among a relatively proximal group of storage nodes from the overall problem of path selection across a larger network. Doing so admits the possibility of localized routing loops for some messages which has been shown to be desirable for avoiding some head-of-line blocking problems. We select eligible storage neighbors using a function of available storage and incident link characteristics. Using simulation, we evaluate this approach and show how migrating custodian storage in this fashion can improve message completion rate by as much as 48% for some storage-constrained DTN networks.

The Delay Tolerant Networking (DTN) architecture approaches the problem of reliable message delivery in intermittent networks using a store‐and‐forward approach where messages may remain stored for relatively long periods of time in persistent storage at DTN routers. Forwarded messages are removed from persistent storage only when transfer acknowledgment to another router or final recipient is received. Congestion in such networks takes the form of persistent storage exhaustion. Several solutions exist including slowing sources, using alternative routes, discarding traffic, or migrating messages to alternative storage locations. We propose storage routing (SR), a congestion management solution of the last form. SR employs nearby nodes with available storage to store data that would otherwise be lost given uncontrollable data sources (such as sensors). SR determines a collection of messages and neighbors to migrate them to using a set of locally scoped distributed algorithms, possibly incorporating loops that are known to be optimal for some DTN routing scenarios and decouples storage management from global DTN route selection. Simulations show up to a 500 per cent performance improvement using SR as compared with a comparable scenario lacking SR. Furthermore, we show a desirable parameter insensitivity to node storage capacity, neighborhood search radius, and message lifetime. Copyright © 2007 John Wiley & Sons, Ltd.

No abstract

No abstract

No abstract