�ݺ�ߣ

Politecnico di Milano, DEIB
Portoroz - 2017 - Riccardo Tommasini - @rictomm - Politecnico di Milano
Marco Balduini, Riccardo Tommasini, Emanuele Della Valle
A Cheaper, Faster yet more Accurate ?
Streaming Linked Data Framework
1

ESWC
RSP is Great!
2

EyE
ESWC
Why RSP?
3
- offers a generic overview over streams and static data
- enables query answering across heterogeneous sources
- consents to create/publish new streams or graphs

ESWC
The RSP Idea
4
in short

EyE
ESWC
CQL Model
5
Streams
Relations
Streams-to-Relations
Relations-to-Streams
Relationsto-Relations
The CQL continuous query language ?
- Arvind Arasu �� Shivnath Babu �� Jennifer Widom, 2006, VLDBJ

EyE
ESWC
RSP-QL Model
6
RDF Streams Solution
Mappings
S2R operators
R2S operators
R2R operators
The CQL continuous query language ?
- E.?Della?Valle, S. Ceri, D.?Barbieri, D. Braga, A.?Campi, 2008, FIS

EyE
ESWC
RSP in a Nutshell
7
RDF Stream-to-RDF
RDF-to-RDF (solution mappings)
RDF-to-RDF Stream
on RDF Streams

ESWC
RSP in Practice
8
With SLD

EyE
ESWC
An Social Media Example
9
How many micro-posts do occur over time?
How often does a hashtag appears in the micro-post
stream?
Two Information Needs

EyE
ESWC
Streaming Linked Data Server
Sources
Raw Stream
10
Adapter RDF Stream Bus Publisher
Visualizer
Recorder Re-player Analiser Decorator

EyE
ESWC
11

An Important
Optimisation
REGISTER STREAM sstr AS
CONSTRUCT {
?id sma:twitterCount ?tc }
FROM STREAM <social> [RANGE 1m STEP 1m]
WHERE {
SELECT (uuid() AS ?id) ?tc
WHERE {
SELECT
(COUNT (DISTINCT ?mp) AS ?tc)
WHERE {
?mp a sma:Tweet } } }
12

EyE
ESWC
13

EyE
ESWC
14

Using
C-SPARQL
REGISTER STREAM countT AS
CONSTRUCT {
?uid sma:twitterCount ?tot .}
FROM STREAM <sstr> [RANGE 15m
STEP 1m]
WHERE {
SELECT
(uuid() AS ?uid)
(SUM(?tc) AS ?tot)
WHERE {
?id sma:twitterCount ?tc }}
15

ESWC
Is RSP always great?
16

EyE
ESWC
Observations
17
It is flexible. :)
It forces RDF when query results are often relational :(
It is not optimal, i.e. RSP-QL vs SQL vs Path Queries
on SLD

ESWC
Revolutionising SLD
18

EyE
ESWC
A ��Lazy�� Processing Model
19
Stream operators can be applied on generic data items.
QL-specific operators requires a particular data type.
Postpone the data transformation as late as possible.
on streams

EyE
ESWC
Generic Programming
20
Generic programming is a style of computer
programming in which algorithms are written in terms of
types to-be-specified-later that are then instantiated
when needed for specific types provided as parameters.
an old idea

EyE
ESWC
A new Processing Model
21
Generic?
Streams<T>
Generic?
Instantaneous<T>
S2I<T>
I2S<T>
I2I<T>

EyE
ESWC
Lazy Transformation by Generic Programming
22
stream-to-instantaneous<T>
instantaneous-to-instantaneous<T>
instantaneous-to-stream<T>
on streams<T>

ESWC
RSP in Practice
23
with SLD Revolution

EyE
ESWC
Sources
Stream
Sink
Streaming Linked Data Revolution Server
24
Receiver Generic Stream Bus Translator
Stream
Recorder Re-player Processor Decorator

EyE
ESWC
25

ESWC
SLD vs SLD Revolution
26
Let��s be quantitative

EyE
ESWC
It is faster, cheaper yet more accurate than SLD.
27
R? = 0,96413R? = 0,99891
30
300
3000
1 10 100
Median Engine Memory (MB)
Median CPU Load (%)
SLD
SLD Revolution
Expon. (SLD)
Linear (SLD Revolution)

ESWC
Discussion & Conclusion
28

EyE
ESWC
Observations
29
It is faster, cheaper yet more accurate than SLD. :)
It requires to know EPL, SPARQL, JSON path queries. :(
It is optimised and, thus, not flexible. :(
on SLD Revolution

EyE
ESWC
Open Problems
30
RSP-QL is not always the best solution in terms of cost/performance
Can we identify an optimum?
Can we define a cost model for RSP-QL ?

Questions?
Email: riccardo.tommasini@polimi.it?
Twitter: @rictomm
31
Email: marco.balduini@polimi.it?
Twitter: @ balducci85
Pablo Picasso, Les Demoiselles d'Avignon, 1907. ?
Museum of Modern Art (MoMA), New York City, NY, US

�ݺ�ߣ

SLD Revolution: A Cheaper, Faster yet more Accurate Streaming Linked Data Framework

More Related Content

SLD Revolution: A Cheaper, Faster yet more Accurate Streaming Linked Data Framework