EECS 600 Lecture Note
Dan Liu  02/16/2005

Topic: Rumor Routing Algorithm For Sensor Networks

1.Recall last two lectures on routing: 
   Dynamic source routing  vs. Greedy Perimeter Stateless Routing
Q: Why is the later called Stateless?
A: Only need to state the information of neighbours but not that of long-distant-hops.


2.What is Rumor Routing? What is its properties?
A routing protocol to deliver packets to events by limiting the scale of query propagation.
Low cost but long delay as well as high probability of success.


3.Requirement of applications in data-centric Network?
a.Data is more important so no need of routing if no data available
b.Focuses on data routing instead of naming based schemes


4.Previous approaches for communication in data-centric Network?
Flooding: (short latency)
  propagate package to the next hop recursively until hit the target and return the answers.
a.Query flooding: 
  Flooding the entire network with a query packet. (N transmissions per query. Q*N total where Q=number of queries)
  May cause higher overhead.Expensive for high query/event ratio, but otherwise good for reverse path setup.  
b.Event Flooding: 
  Flooding the network whenever a node witness an event.All other nodes can form gradients toward the event, based on the number of hops to sink.(N transmissions per query. E*N total where E = number of events)

Q:What is gradients prapogation?
A:Instead of prapogating randomly, just send the packets along the direction with increasing hop numbers.

  
5.Rumor Routing vs Query & Event Flooding?
Rumor Routing: a logical compromise between flooding queries and flooding event notifications with moderate 
Query/event ratios.

Q:How to obtain the Query/event ratios?
A:Though plenty of testing and evaluation. So still not very efficient method by now.


6.Action of agents in rumor routing
Informs nodes it encounters of any events it has witnessed along its route: carries a list of all the events it has encountered, along with the number of hops to that event. When it arrives at node A from its neighbor B, it will synchronize its list with the node’s list.TTL is decremented, and if it is greater than zero, the agent is forwarded.

7.Queries Processing
a.If a node has a route (event path state) toward the target event, it forwards the query along the route.
b.If it does not, it forwards the query to a random neighbor, assuming the query has not exceeded its TTL.

Q:How about multiple queries? 



7.