Data Access Models in Location Dependent Information Services

1
Data Access Models in Location Dependent
Information Services

• Yu Meng
• May 1, 2004

2
Outline

• Introduction
• Related concepts
• Location models
• Query types
• Valid scopes
• Access models
• On-demand Access
• Broadcasting
• Summary

3
Introduction

• What is LDIS
• What are the challenges
• Cellular Architecture for LDIS

4
Introduction

• Provide local or nonlocal news, weather, traffic
  reports, navigation maps and directory services
  in wireless environment.

5
Introducton

• Mobile environment constraints,
• Spatial data,
• User movement.

6
Introduction
7
Related Concepts

• Location Models
• Query types
• Valid scope

8
Related Concepts -Location Models

• Geometric model.
• Latitude-longitude pair returned by GPS.
• Advantage good for heterogeneous system,
• Disadvantage costly in terms of data volume
• Symbolic model.
• Real-world entities.
• Logical entities
• Advantage easy to manage data with well
  organized structures.
• Disadvantage hard to convert among heterogeneous
  systems.(good topic for RFC)

9
Related Concepts -Query types

• Local vs. non-local queries.
• Tell the local weather,
• Find the weather in New York City.
• Simple vs. general queries.
• Download the local traffic report,
• List the hotels within 30 miles,
• List the hotels with a room rate below 100.

10
Related Concepts -Valid Scope

• The area or areas within which the query result
  is valid.
• Data object returned (query, result, vs)
• (nearest-hotel, A, vs),
• (nearby-restaurant, A,B, 1,2).

11
Related Concepts Valid Scope Example
12
Data Access Models

• On-demand access
• Broadcasting
• Hybrid of the two.

13
On-demand Access

• Data placement,
• Data replication,
• Query scheduling,
• Indexing.

14
On-demand Access-Data Placement
15
On-demand Access-Data replication

• The system creates certain copies of the data and
  places them at different locations in the
  network.
• Work done are based on network topology and
  access patterns.
• Problem Access patterns may be time dependent
  periodically or temporally. Is EMM a solution?

16
On-demand Access-Query scheduling

• Query scheduling determines query processing
  order.
• Work has been seen in improving average queuing
  delay.
• What happens if client moves?
• Is prediction a solution?

17
On-demand Access-Query Scheduling
18
On-demand Access-Indexing

• Disk indexing
• Geometric location model MBR based indexing. May
  be inefficient caused by overlapping.
• Symbolic location model mapping to valid data
  object is needed.
• Several R tree based algorithms are proposed but
  none works superior to others in all cases.

19
On-demand Access-Indexing
20
On-demand Access-Indexing
21
On-demand Access-Indexing
22
Broadcast

• Broadcast lets an arbitrary number of users
  simultaneously access data.
• Good for simple queries.
• Hard for general queries.

23
Broadcast-Air indexing

• Client can download a indexing info to predict
  availability of queried data.
• Indexing size and latency.
• Broadcasting strategy how to divide bandwidth?
  Based on the statistics.
• Not adaptive!

24
Data Caching

• Data may be cached at the mobile clients for
  better performance.
• Data consistency
• Location dependent cache invalidation.
• Time dependent cache invalidation.

25
Data Caching-Data Replacement

• LRU
• P/X
• Distance based algorithm
• Valid scope

26
Data Caching-Data Prefetching

• Feasible for simple queries.
• May be hard for general queries.
• Not much work on this issue.

27
Summary

• LDIS is a developing technology.
• Many research opportunities remains.
• SPOT (Smart Personal Objects Technology )
  announced by Microsoft in 2003