Differences

This shows you the differences between two versions of the page.

--- teaching:mfe:is [2017/04/19 11:01]
svsummer
+++ teaching:mfe:is [2018/04/23 09:54]
svsummer [Master Thesis in Collaboration with Euranova]
@@ Line 24: / Line 24: @@
   * Contact : [[ezimanyi@ulb.ac.be|Esteban Zimanyi]]
+** Dynamic Query Processing on GPU Accelerators
+   This master thesis is put forward in the context of the DFAQ
+   Research Project:  "Dyanmic Processing of Frequently Asked
+   Queries", funded by the Wiener-Anspach foundation.
+   Within this project, our lab is hence developing novel ways for
+   processing "fast Big Data", i.e., processing of analytical queries
+   where the underlying data is constantly being updated.  The
+   analytics problems envisioned cover wide areas of computer science
+   and include database aggregate queries, probabilistic inference,
+   matrix chain computation, and building statistical models.
+   The objective of this master thesis is to build upon the novel
+   dynamic processing algorithms being developed in the lab, and
+   complement these algorithms by proposing dynamic evaluation
+   algorithms that execute on modern GPU architectures, thereby
+   exploiting their massive parallel processing capabilities.
+   Since our current development is done in the Scala programming
+   language, prospective students should either know Scala, or being
+   willing to learn it within the context of the master thesis.
+   *Validation of the approach* Validation of master thesis' work
+   should be done on two levels:
+  - a theoretical level; by proposing and discussing alternative ways
+    to do incremental computation on GPU architectures, and comparing
+    these from a theoretical complexity viewpoint
+  - an experimental level; by proposing a benchmark collection of CEP
+    queries that can be used to test the obtained versions of the
+    interpreter/compiler, and report on the experimentally observed
+    performance on this benchmark.
+  *Deliverables* of the master thesis project
+   - An overview of query processing on GPUs
+   - A definition of the analytics queries under consideration
+   - A description of different possible dynamic evaluation algorithms
+     for the analytical queries on GPU architectures.
+   - A theoretical comparison of these possibilities
+   - The implementaiton of the evaluation algorithm(s) (as an interpreter/compiler)
+   - A benchmark set of queries and associated data sets for
+     the experimental validation
+   - An experimental validation of the compiler, and analysis of the results.
+   *Interested?*
+   - Contact :  [[svsummer@ulb.ac.be][Stijn Vansummeren]]
+   *Status*: available
+** Complex Event Processing in Apache Spark and Apache Storm
+  The master thesis is put forward in the context of the SPICES
+  "Scalable Processing and mIning of Complex Events for
+  Security-analytics" research project, funded by Innoviris.
+  Within this project, our lab is developping a declarative language
+  for Complex Event Processing (CEP for short). The goal in Complex
+  Event Processing is to derive pre-defined patterns in a stream of
+  raw events. Raw events are typically sensor readings (such as
+  "password incorrect for user X trying to log in on machine Y" or
+  "file transfer from machine X to machine Y"). The goal of CEP is
+  then to correlate these events into complex events. For example,
+  repeated failed login attempts by X to Y should trigger a complex
+  event "password cracking warning" that refers to all failed login
+  attempts.
+  The objective of this master thesis is to build an
+  interpreter/compiler for this declarative CEP language that targets
+  the distributed computing frameworks Apache Spark and/or Apache
+  Storm as backends.  Getting aquaintend with these technologies is
+  part of the master thesis objective.
+  *Validation of the approach* Validation of the proposed
+  interpreter/compiler should be done on two levels:
+  - a theoretical level; by comparing the generated Spark/Storm
+    processors to a processor based on "Incremental computation" that
+    is being developped at the lab
+  - an experimental level; by proposing a benchmark
+    collection of CEP queries that can be used to test the obtained
+    interpreter/compiler, and report on the experimentally observed
+    performance on this benchmark.
+  *Deliverables* of the master thesis project
+   - An overview of the processing models of Spark and Storm
+   - A definition of the declarative CEP language under consideration
+   - A description of the interpretation/compilation algorithm
+   - A theoretical comparison of this algorithm wrt an incremental
+     evaluation algorithm.
+   - The interpreter/compiler itself (software artifact)
+   - A benchmark set of CEP queries and associated data sets for
+     the experimental validation
+   - An experimental validation of the compiler, and analysis of the results.
+   *Interested?*
+   - Contact :  [[svsummer@ulb.ac.be][Stijn Vansummeren]]
+   *Status*: available
+** Graph Indexing  for Fast Subgraph Isomorphism Testing
+   There is an increasing amount of scientific data, mostly from the
+   bio-medical sciences, that can be represented as collections of
+   graphs (chemical molecules, gene interaction networks, ...). A
+   crucial operation when searching in this data is that of subgraph
+   isomorphism testing: given a pattern P that one is interested in
+   (also a graph) in and a collection D of graphs (e.g., chemical
+   molecules), find all graphs in G that have P as a
+   subgraph. Unfortunately, the subgraph isomorphism problem is
+   computationally intractable. In ongoing research, to enable
+   tractable processing of this problem, we aim to reduce the number
+   of candidate graphs in D to which a subgraph isomorphism test needs
+   to be executed. Specifically, we index the graphs in the collection
+   D by means of decomposing them into graphs for which subgraph
+   isomorphism *is* tractable. An associated algorithm that filters
+   graphs that certainly cannot match P can then formulated based on
+   ideas from information retrieval.
+   In this master thesis project, the student will emperically
+   validate on real-world datasets the extent to which graphs can be
+   decomposed into graphs for which subgraph isomorphism is
+   tractable, and run experiments to validate the effectiveness of
+   the proposed method in terms of filtering power.
+   *Interested?*
+   - Contact :  [[svsummer@ulb.ac.be][Stijn Vansummeren]]
+   *Status*: available
 ===== Complex Event Processing in Apache Spark and Apache Storm =====
@@ Line 130: / Line 260: @@
 **Status**: available
-=====Assessing Existing Communication Protocols In The Context Of DaaS =====
-Data-as-a-Service (DaaS) is an emerging cloud model. The main offering of DaaS is to allow data producers/owners to publish data services on the cloud. The idea of publishing data via a service interface is not new. SOA protocols have enabled this long ago. Yet, these protocols were not developed with the cloud and the big data in mind. This is probably why the term DaaS has emerged. It marks the need for protocols and tools that enable big data exchange.
-DaaS services need to exchange large amounts of data. Large here refers to large message size, large message count, or a combination of both. RESTful services, for instance, communicate over HTTP, which is not a good choice for communicating large messages/files. SOAP services are not bound to HTTP, but they introduce another overhead of requiring messages to be strictly formatted in XML. This is why researchers started to reconsider older protocols like the BitTorrent, and suggesting extension to existing protocols like the SOAP with Attachments.
-The topic of this thesis is to perform a comprehensive survey on the protocols data exchange, and assess their suitability for DaaS. A quantitative comparison of protocols need to be done, considering at least these two dimensions: (1) the protocol: SOAP, REST, BitTorrent, etc, and (2) the message: short inline, long inline, file. The assessment should be in terms of reliability, performance, and security.
-**Deliverables** of the master thesis project
-  * A report that reviews the state of art communication protocols.
-  * Propose a tool for DaaS developers to choose the best protocol/s based on their application needs. Such a tool might also provide means of automatically switching between protocols on certain thresholds.
-  * Experiments to assess the suitability of protocols for DaaS, and to compare between them. These experiments need to be repeatable, so that others can use them on their own datasets and configurations.
-**Interested?**
-  * Contact : [[ezimanyi@ulb.ac.be|Esteban Zimanyi]]
-**Status**: available