1
This talk will…
|Explain how the underlying structure of different
P2P systems affects querying
|Compare the relative advantages and
disadvantages of each
|Summarise issues in query processing
|Show current directions for research in P2P
search
Related: [3]
CAP Theorem Related: [3][4]
You can only have two of these three properties
The nature of the system is determined by this choice
CAP Theorem: Sacrifice Partitions
Classic Distributed Systems style
Examples: single site & clustered databases
CAP Theorem: Sacrifice Availability Pessimistic locking
Examples: distributed databases
CAP Theorem: Sacrifice Consistency
Optimistic, conflict resolution, expirations Examples: DNS, Web Caching, PIER
Querying the Network
Querying is limited by constraints on bandwidth and processing
Addressed by
Minimizing data sent over network
Applying economic cost models
On an internet scale this is not sufficient for distributed databases
Consistency is often sacrificed (CAP Theorem)
From [19]
29
PIER: Example 2
{Takes, {st-id, m-id}}{Module, 1002}{Module, 2001}{Takes, {1, 1002}}{Takes, {4, 1002}}{Takes, {2, 2001}}{Takes, {3, 2001}}
PIER: Example 2 (Problems)
name from
(student innerjoin takes on student.st_id = takes.st_id )where mid =“1002”
A measurement study of Napster and
. In Biennial Conference on Innovative Data Systems Research . ACM Symposium on Operating Systems Principles
A game theoretic framework for incentives