Putting the Web first

Ruben Verborgh, Ghent University – imec

SemWeb.Pro, 21 November 2016

Putting the Web first

Ruben Verborgh

Ghent University – imec

It's 1999, you're into CS and have to decide whether you'll change the world through AI or Semantic Web. Choose well, there's 1 right choice

William Vambenepe (@vambenepe) September 23, 2016

@vambenepe One is about centralized intelligence for the happy few. The other is about decentralized intelligence for all. Tough choice ;-)

Ruben Verborgh (@RubenVerborgh) September 24, 2016

Web is the differentiating factor
for the Semantic Web.

If things don't work on the Web,
we have a serious problem.

Putting the Web first

Putting the Web first

Where are the Semantic Web
app developers?

My research focuses on bringing
the Semantic Web back to the Web.

Design things that work
like the rest of the Web.

Public SPARQL endpoints do not scale,
and never will.

The true potential of Linked Data
lies in connecting datasets.

Putting the Web first

Linked Data publishing so far
has been a story of two extremes.

the Linked Data Fragments axis

Possible Linked Data interfaces exist
in between those two extremes.

the Linked Data Fragments axis

Linked Data Fragments is a uniform view
on Linked Data interfaces.

the Linked Data Fragments axis

Every Linked Data interface
publishes specific fragments
of a Linked Data set.

We designed a new trade-off mix
with low cost and high availability.

the Linked Data Fragments axis

A Triple Pattern Fragments interface
is low-cost and enables clients to query.

the Linked Data Fragments axis with Triple Pattern Fragments indicated

This Triple Pattern Fragment shows
subjects born in Paris from DBpedia.

SPARQL queries are executed by clients,
by splitting them in supported fragments.

SELECT  ?person  ?name  WHERE {
    ?person  rdfs:label  ?name;
             rdf:type  dbpedia-owl:Artist;
             dbpedia-owl:birthPlace  ?city.
    ?city  rdfs:label  "Paris"@en.
}
LIMIT 100

Datasource: http://fragments.dbpedia.org/2016-04/en

This browser client evaluates
a complex query with fragments.

Evaluating queries over federations means
asking multiple servers for fragments.

Putting the Web first

@RubenVerborgh "1.000 results in 55.0s" That's why nobody wants federated search.

(((Jörg Prante))) (@xbib) August 19, 2016

.@xbib @claussni I’d argue it’s fantastic! Don’t tell me you could read those 1.000 results in just 55s ;-) And 1st 100 arrived much faster!

Ruben Verborgh (@RubenVerborgh) August 19, 2016

There's no silver bullet.
There's no single metric.

We compared Triple Pattern Fragments
against SPARQL endpoints.

We ran the Berlin SPARQL benchmark with:

The query throughput is lower,
but resilient to high client numbers.

throughput of different SPARQL query solutions

The server traffic is higher,
but individual requests are lighter.

server traffic of different SPARQL query solutions

Caching is significantly more effective,
as clients reuse fragments for queries.

cache usage of different SPARQL query solutions

The server requires much less CPU,
allowing higher availability at lower cost.

server CPU usage of different SPARQL query solutions

Federation is the killer use case
for Linked Data on the Web.

In federated scenarios, light interfaces
can achieve fast query times as well.

execution time of federated queries

Putting the Web first

Triple patterns are not the final answer.
No interface ever will be.

If we want to see intelligent clients,
we should stop building intelligent servers.

I challenge you to explore the axis
to find and measure other trade-offs.

the Linked Data Fragments axis with Triple Pattern Fragments indicated

Does it work
on the Web?

Putting the Web first

@RubenVerborgh

Ghent University – imec