Download Folie 1

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Document related concepts
no text concepts found
Transcript
Sebastian Obermeier,
Stefan Böttcher
University of Paderborn
Germany
ICEC 2008, Innsbruck, Austria
Agenda:
Scenario
Query Shipping vs. Data Shipping
Solution
Experiments
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
2/19
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
3/19
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
4/19
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
5/19
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
6/19
Query Q
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
7/19

Intermediate node N checks whether it can answer Q
 Only Q's result is transferred
 Test can be complex and time consuming
 Small missing parts of information lead to cache-misses:
 Qcache = //restaurant[./@areaID<50]//description
 Q = //restaurant[./@areaID<35]//description
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
8/19
Query Q: {1,3,4}
{1,2}
{2,3}
{1,2,3}
{7}
{1,3,4,7}
{1,4,7}
Scenario
ICEC 2008
{1,2}
Query Shipping vs. Data Shipping
Solution
Experiments
9/19

Request parts of the document
 Combination of cached content can answer Q
 Tests are fast
 Huge amount of overhead if read-set is large, e.g. if Q uses count()
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
10/19

No arbitrary queries
 Query templates predefined

Mostly point and range queries including filters

Database can track queries

Focus on content, e.g. text, pictures, and videos
 Database updates are rare

Egoistic node behavior
 do not spend much energy
to other node’s queries
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
11/19


Split XML document into disjoint
fragments according to a
Split Schema Graph (SSG)
S6
S1
Querying node determines by SSG
necessary fragments to answer query Q
S2
S3
S6
S1

XML
XML
S2
S3
Q is executed locally on the read-set of Q
(=merged segments)
S3
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
12/19

XML document split
into disjoint parts
Segment 1 /1/2/2/1
<restaurants>
<restaurant id = "25" areaID="15">
<name>Forester`s House</name>
<description>Traditional… </…>
<style>German</style>
</restaurant>
<restaurant id = "35" areaID="17">
<name>Garden of Sun</name>
<description>Large beer garden…</…>
<style>Austrian</style>
</restaurant>
...
</restaurants>
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
13/19

//restaurant[@areaID>13][@areaID<19]/name
Required Segments
1 /*/*/2/*/
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
14/19




1600 devices, logical clock
24MB Information Repository
 Max. distance 5 hops
Individual query profiles
 Each with 164 XPath queries
 80% request
hotspot data (5MB)
Hotspot changes
during evaluation
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
15/19
Scenario
ICEC 2008
XPath Query
Shipping
XPath Query Shipping
500kB Cache
XPath Query
Shipping
XPath Query Shipping
500kB Cache
Query Shipping vs. Data Shipping
Solution
Experiments
16/19
XPath Query
Shipping
XPath Query
Shipping (GZIP)
Scenario
ICEC 2008
XPath Query Shipping
500kB Cache
XPath Query Shipping
(GZIP) , 500kB Cache
Query Shipping vs. Data Shipping
Solution
Experiments
17/19
500kb Cache
1000kb Cache
2000kb Cache
XPath Query
XPath Query
Shipping (GZIP) Shipping (GZIP)
500kB Cache
Scenario
ICEC 2008
XPath Query
Shipping (GZIP)
1000kB Cache
Query Shipping vs. Data Shipping
XPath Query
Shipping (GZIP)
2000kB Cache
Solution
Experiments
18/19

Querying and caching mechanism
that allows clients to execute queries locally

Application based fragmentation schema

Simple cache contribution tests by IDs

S6
S1
XML
S2
S3
2 /4/*/1 == 2 /4/2/1
Coupes with egoistic node behavior

Reduces network traffic up to 88%

Improves query response time up to factor 5

Reduces bottlenecks

Can be individually used for each query type
Scenario
ICEC 2008
Query Shipping vs. Data Shipping
Solution
Experiments
19/19
Related documents