1. PROSE CS 218 Fall 2017

2. Another MANET Routing Approach
Using prefix labels instead of more ordinary addresses
And distributed hash tables to translate from the node’s permanent name to the prefix labels
Prefix labels describe how to get to a node’s current location
From a known root
So knowing the label tells you how to route
Prefix labels change as nodes move
The DHT allows those who know the node’s name to find its prefix label

3. PROSE Prefix Labels
Build a directed acyclic graph from the MANET nodes
First elect a root node
Then use a breadth first search to find nearby nodes to the root
Label those nodes, then continue to search
Each label specifies how to get from its owner to the root node

4. An Example Graph The root’s children The root
Why isn’t E a child of the root?
Because PROSE builds an acyclic graph

5. Distributed Hash Tables
Store data associated with names at many different nodes
Hash each name to find a matching key
The key specifies which node stores the data associated with the name
There are complexities . . .

6. How Does PROSE Use Its DHT?
Nodes use it to find the prefix labels of other nodes
Destinations they want to send messages to, typically
Each destination hashes its unique node ID (e.g., its MAC) to a prefix label
That prefix label (or nearest match) is the destination’s anchor node
Destination node then sends its unique ID and its own prefix label to its anchor
Any node sending to that destination hashes its ID
Getting the same anchor identity
Send request to anchor

7. What Does An Anchor Node Do?
Either reply to request with the destination’s prefix label
Or forwards request to destination
Which replies to the source
In either case, source learns the destination’s prefix label
And can send directly to it without consulting the anchor again

8. An Example Node N wants to send to node K
N hashes K’s identity to find its anchor, which is C C N K

9. Continuing the Example
N asks C for the prefix label of K
N already has C’s prefix label
More precisely, can route toward it based on the hash
C sends N K’s prefix label, which is 0000 C N
0000 K

10. Now N Can Send Directly To K
Using the prefix label
How to decide the route?
Using simple rules based on how labeling is done C N K

11. That’s All Well and Good, But . . .
What happens when someone moves?
Different cases:
The moving node is an ordinary node
The moving node is an anchor
The moving node is the root node
Each case handled differently

12. Moving an Ordinary Node
Perform a leave operation when move is started
To free up the old label
Which is geographically based
Perform a join operation when move is completed
To be spliced into the tree in a new location
Simple for leaf nodes
More complex for internal nodes
If necessary, elect a new root
Handles the case of the root moving

13. Moving an Anchor Anchor identity is based on candidate node prefix
So if an anchor node moves, its prefix might change
Which might cause it to stop being an anchor for other nodes
In that case, they find a secondary anchor and get translation information to them
It will be the node with the largest prefix matching the old anchor
So on the path to the old anchor
So in the physical vicinity of the old anchor

14. Performance Issues
Moves cause fewer control messages than in AODV and OLSR
Paths tend to be longer than shortest path
Generally 2x or more
Not likely to perform well in situations of high node mobility
Especially if all moving
If root and anchors tend not to move, fewer issues