HomeViews: PeertoPeer Middleware for Personal Data Sharing Applications

1
HomeViews Peer-to-Peer Middleware forPersonal
Data Sharing Applications

• Roxana Geambasu, Magdalena Balazinska, Steven D.
  Gribble, and Henry M. Levy
• Department of Computer Science and Engineering
  University of Washington, Seattle, WA
• SIGMOD'07

2
The IDEA

• selective sharing technique through exchange of
  capabilities in a peer-to-peer environment.

3
Outline

• Introduction
• Motivation for the features of HomeViews
• Architecture and Implementation
• Applications on top of HomeViews
• Evaluation
• Relatedwork
• Conclusion

4
1. INTRODUCTION

• Challenges
• 1) need to organize files into directories
• 2) search objects of interest through large
  volumes of personal data
• 3) manually share their data.

5

• Therefore, come forth tools like desktop search
  tools, peer-to-peer and Web-based file-sharing
  systems
• Shortcoming about these tools
• 1) not integrated with each other
• 2) manual uploads and downloads
• 3) not deal with dynamically.

6

• Thus,present the HomeViews,it supports
• 1) The the creation of database-style views over
  an user's file repository,
• 2) a lightweight protection mechanism,
• 3) seamless sharing and integration of local and
  remote data and views,
• 4) peer-to-peer communication between HomeViews
  instances on peer computers.

7
2. MOTIVATION

• personal data management state( three key
  requirements)
• Need powerful but simple tools to organize their
  files
• selective sharing data with friends, family,and
  colleagues across the Internet
• Want to seamlessly integrate shared files in a
  distributed data-sharing environment.

8

• Current states are
• users must manually copy or download shared data
  onto their home machines in order to include them
  in their local organizational structures. Its
  the cumbersome, inefficient, and static.

9

• To resolve these problems
• Presented a new generation of personal data
  management and data sharing applications. Figure
  1 assumed a scenario shows how to manage personal
  data

10
(No Transcript)
11

• The goal is to implement a personal-data-sharing
  applications with several equirements
• data is organized into views populated by
  queries
• views must be directly sharable with trusted
  parties, without the need to register with a
  service
• it should be possible to integrate remote data
  shared by others with local data
• views are dynamic.

12
3. ARCHITECTURE AND IMPLEMENTATION

• 3.1 Architectural Overview of HomeViews
• Figure 2 shows the HomeViews system architecture.

13
parses requests into internal data structures
provides a lightweight access control scheme for
views based on capabilities
the query is on a local view
the query is on a remote view
validate remote queries
provides indexing and keyword or attribute-based
search functions
14
3.2 Capability-based Access Control

• capability consists of a uniquely identifies a
  single object in the Internet, and a set of
  access rights for object. HomeViews capabilities
  protect views and enable view sharing.
• capability is like a ticket or door key

15
Naming and Access Control in HomeViews
whether a local view.

• Capability has three parts (Figure 3).

H(MAC Adrdess)locally unique-forall- time view
ID
Random data
Whether pair
matches with CapTables (acess control)
16
Relationship(capability versus view) 11 or n1
Don't store in capability, eliminate the need for
encryption
Whether pair
matches with capabilitys (acess control)
17
3.2.3 Sharing

• Figure 5 illustrates the use of capabilities for
  the dynamic album-sharing scenario

18
(No Transcript)
19
3.3 Query Language

• HomeViews' the query language is called SQLCapa,
  is a modified version of SQL with integrated
  capability-based access control
• The Table 1 summarizes SQLCapa's modifications to
  SQL.

20
(No Transcript)
21

• E.g.

22
keyword queries
23

• 3.3.2 Creating Views

24

• 3.3.3 Capability Restriction

SELECT (read), DROP (delete the view), ALTER
(modify the view definition), REVOKE (revoke
capabilities defined for the view), CATALOG
LOOKUP (look up the view definition in the
catalog)
25

• 3.3.4 Capability Revocation
• Revocation statement
• The REVOKE C'IT USING CIT

26

• 3.3.5 Catalog Information Lookup

capability
Return the attributes of capability X in view
table
27

• 3.4 Query Processing
• Query Execution Algorithms
• Recursive evaluation
• If capabilities do not have the CATALOG
  LOOKUP((look up the view definition in the
  catalog) right.
• Query rewrite
• If capabilities include the CATALOG LOOKUP right

28
local query
If V is defined on capabilities, recursive
query
Remote query
29

• 4. APPLICATIONS
• Two Data-Sharing Applications
• ViewGallery and ViewFS
• ViewGallery is a modified version of the
  well-known centralized photo-sharing application
  Gallery v.1

30

• Application Requirements

31
5. EVALUATION

• Evaluation goal
• (1) the impact of various system components on
  overall performance,
• (2) whether HomeViews is sufficiently fast to
  organize and share data in practice.

32
opensource query engine
33

• Simple queries are one level deep they involve
  a single view,itself defined directly over a base
  view.
• Complex queries involve views whose definitions
  include multiple other views composed in various
  ways.
• Depthselection on top of another(remote) view
• Breathapplying union, set difference, or
  intersection on top of other (remote) views

34
Complex queries
35
Evaluation of Simple Queries
local query engine
36
Evaluation of Complex Queries
37
Increase 80 from 1 to 5 depth
38
Scalability
using a large number of concurrent requests
(keep 100), Beagle query engine
39
Discussion

• Our results demonstrate that our prototype is
  sufficiently fast to be practical in medium-scale
  environments(hundreds)
• Using a Spotlight,could significantly improve
  performance and scalability
• query engine account for the majority of the
  execution time.
• On fast networks, the depth and breadth of views
  have little influence on recursive query
  evaluation times
• Caching technology

40
RELATED WORK

• Another selective sharing technique is to encrypt
  data with multiple keys and distribute different
  keys to different users
• HomeViews integrates concepts and mechanisms from
  capability systems into database views in a
  distributed peer-to-peer system.

41
CONCLUSION

• HomeViews, a new peer-topeer middleware system
• We implemented two applications on top of
  HomeViews,
• Finally, our measurements demonstrate the
  negligible cost of our protection mechanism and
  the practicality of our approach for medium-scale
  environments.

42

• THANK YOU