Scalability of Persistent Queries

1
Scalability of Persistent Queries
(Contact Daniela Rus, rus_at_cs.dartmouth.edu)
Mobile Agent Persistent Query System
Persistent Query Module
Inside each Query
A
Soldier in the field
A
Star Clustering Algorithm
Serval Architecture

• Estimate degrees of nodes by sampling
  similarities
• Find and mark star cluster centers
• Insert documents into appropriate clusters

Scalability of Persistent Queries
Lessons Learned
1) sampled star is effective and scalable for
persistent queries 2) agents can do local
processing on the persistent-query results before
sending the documents back 3) avoid information
overload 4) current implementation can sustain
1,600 persistent queries 5) sampled star
algorithm solves the memory problems
observed with the star implementation 6) future
direction for modeling and scalability
experiments for the Mobility TIE