In the Web of Data, in which Data are increasing in Volume, Variety, and Velocity, the Stream Reasoning and RDF Stream Processing research areas try to make sense of these continuous streams of schemaless data, by taking advantage of both the vast Stream Processing literature and Semantic Web technologies at the same time. State-of-the-art solutions proposed so far mostly proved that taming Variety and Velocity simultaneously is feasible. In this Thesis, we propose μCR, an infrastructural redesign based on micro-services of two existing approaches, i.e., Cascading Reasoning (CR) and Network of Stream Reasoners (NoSR). Our proposal, taking inspiration from the Unix philosophy, merges the two existing approaches, producing a network of simple reasoners, each with its reasoning expressiveness, that can be used as rewriting target for continuous queries. To enable this approach, we present a Stack and propose our solutions for its foundational layers. Finally, to produce a Proof of Concept of a μCR enabled stream reasoner, we focus on rewriting directly to Raw Stream Processing systems based on the Dataflow execution model. In fact, accordingly Stonebraker's principles, we recognize as a necessity the ability to perform little expressive reasoning on fast changing data in an horizontally scalable way.Therefore, we present Metamorphosis, i.e., a proof-of-concept implementation for RDF Stream Processing over Kafka Streams using the Dual Streaming model. We then prove the feasibility of our approach using SRBench benchmark for query expressiveness.
Nel Web of Data di oggi, per il quale gli aspetti di "Volume", "Variety" e "Velocity" sono sempre più di crescente importanza, Stream Reasoning e RDF Stream Processing si propongono come campi di ricerca volti al dare un significato a questo continuo stream di dati, utilizzando al contempo gli strumenti dello Stream processing e le tecnologie del Semantic Web. Le soluzioni proposte finora si sono concentrate principalmente sulla fattibilità di gestire contemporeamente "Variety" e "Velocity" dei dati. In questa tesi proponiamo μCR, una reinterpretazione infrastrutturale basata su micro servizi di due approcci già proposti in letteratura, "Cascading Reasoning" (CR) e "Network of Stream Reasoners" (NoSR). La nostra proposta, ispirata alla filosofia Unix, unisce i due approcci sopra citati, producendo un network di "reasoners" semplici, ognuno caratterizzato da una specifica espressività di ragionamento, che possono essere utilizzati come obbiettivo per la riscrittura di query continue. Al fine di attuare questo approccio, abbiamo proposto uno "Stack" e delle soluzioni per i suoi livelli fondamentali. Infine, per produrre un esempio di Stream Reasoner abilitato ad interagire all'interno della rete μCR, ci siamo concentrati sulla riscrittura verso sistem di "Raw Stream Processing" basati sul modello di esecuzione Dataflow. Infatti, seguendo i principi proposti da Stonebraker, riconosciamo come una necessità l'essere capaci di eseguire ragionamenti poco espressivi su dati in rapido cambiamento scalando orizzontalmente, per poter gestire anche la componente "Volume" nei dati. Per questi motivi, presentiamo Metamorphosis, un esempio di implementazione di RDF Stream Processing utilizzando "Kafka Streams" e il "Dual Streaming model". Quindi, proviamo la fattibilità del nostro approccio testando l'espressività delle query implementabili grazie al benchmark SRBench.
An infrastructural view of cascading stream reasoning using micro-services
SCORSOLINI, PHILIPPE
2018/2019
Abstract
In the Web of Data, in which Data are increasing in Volume, Variety, and Velocity, the Stream Reasoning and RDF Stream Processing research areas try to make sense of these continuous streams of schemaless data, by taking advantage of both the vast Stream Processing literature and Semantic Web technologies at the same time. State-of-the-art solutions proposed so far mostly proved that taming Variety and Velocity simultaneously is feasible. In this Thesis, we propose μCR, an infrastructural redesign based on micro-services of two existing approaches, i.e., Cascading Reasoning (CR) and Network of Stream Reasoners (NoSR). Our proposal, taking inspiration from the Unix philosophy, merges the two existing approaches, producing a network of simple reasoners, each with its reasoning expressiveness, that can be used as rewriting target for continuous queries. To enable this approach, we present a Stack and propose our solutions for its foundational layers. Finally, to produce a Proof of Concept of a μCR enabled stream reasoner, we focus on rewriting directly to Raw Stream Processing systems based on the Dataflow execution model. In fact, accordingly Stonebraker's principles, we recognize as a necessity the ability to perform little expressive reasoning on fast changing data in an horizontally scalable way.Therefore, we present Metamorphosis, i.e., a proof-of-concept implementation for RDF Stream Processing over Kafka Streams using the Dual Streaming model. We then prove the feasibility of our approach using SRBench benchmark for query expressiveness.| File | Dimensione | Formato | |
|---|---|---|---|
|
Philippe_Thesis_final.pdf
Open Access dal 05/04/2020
Descrizione: Thesis text
Dimensione
6.38 MB
Formato
Adobe PDF
|
6.38 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/147402