Difference between revisions of "TTT"

From MonetDB
Jump to navigationJump to search
(Replaced content with "{{DISPLAYTITLE:ThursdayThinkTank}} The Thursday Think Tank is a weekly talk given at the CWI as part of the Database Architectures group. The talk can be given by either...")
Line 1: Line 1:
 
{{DISPLAYTITLE:ThursdayThinkTank}}
 
{{DISPLAYTITLE:ThursdayThinkTank}}
  
The Thursday Think Tank is currently costed by Holger.
+
The Thursday Think Tank is a weekly talk given at the CWI as part of the Database Architectures group. The talk can be given by either one of the group members, or by an interested party. It is typically given in room L302. Below is the schedule for the weekly talks. Currently, [http://www.cwi.nl/people/2827 Mark Raasveldt] is in charge of managing the TTT schedule.  
  
 
= Agenda =
 
= Agenda =
 
+
<div id="mw-ttt-calendar"></div>
== December 15, 2011: Concurrency Control for Adaptive Indexing ==
 
;Speaker: Felix Halim
 
;Abstract
 
:Adaptive indexing constructs and optimizes indexes incrementally, as a side effect of query processing. The goal is to give all performance benefits from indexing while protecting workload performance from the costs. However, read-only queries that result in index optimization now resemble update transactions, in that termination of a read-only transaction implies lock contention, logging, and possibly rollback and recovery after system failures.
 
 
 
:The present paper considers adaptive indexing in the context of concurrency control. We apply and extend prior work to reduce concurrency control conflicts and overheads to practical—or even negligible— levels. In order to achieve this, the design and implementation of adaptive indexing must separate index structures from index contents, which relaxes the constraints and requirements during adaptive indexing compared to those of traditional index updates. Furthermore, we observe that even within the context of structural changes, index refinement is optional. Our experiments demonstrate that the performance overhead of concurrency control during structural updates is minimal and that adaptive early termination alleviates problems with concurrency control.
 
 
 
== January 12, 2012: SRBench: A Benchmark for Streaming RDF Storage Engines ==
 
; Speaker: Jennie
 
; Abstract
 
:Processing streaming RDF data is gaining increasing attention from the semantic web community.  Work has been done on building stream RDF processing engines, and several extensions to the SPARQL language have been proposed to query streaming RDF data.  The growing number of proposals calls for a benchmark for streaming RDF engines to make comparison among different systems possible.  However, existing benchmarks, both in the streaming data processing area and in the RDF data processing area, are not sufficient. The existing benchmarks on streaming data processing do not capture the properties of the RDF data; while the existing benchmarks on RDF data processing do not capture the properties of streaming data.  To advance the state-of-the-art in streaming RDF data processing, we propose SRBench, a benchmark especially designed for streaming RDF storage engines.  As both the SPARQL and the streaming SPARQL languages are still evolving, the first version of SRBench targets at examining the expressiveness of the language for querying streaming RDF data. This presentation gives an overview of the current status of SRBench.
 
 
 
== January 19, 2012: Subquery plan reuse based Query Optimization ==
 
; Speaker: Vamsi
 
; Abstract
 
In this presentation, we revisit the problem of query optimization in relational DBMS. We propose a scheme to reduce the search space of Dynamic Programming based on reuse of query plans among similar subqueries. The method generates the cover set of similar subgraphs present in the query graph and allows their corresponding subqueries to share query plans among themselves in the search space. Numerous variants to this scheme have been developed for enhanced memory efficiency. Our implementation and experimental study in PostgreSQL show that one of the schemes is better suited to improve the performance of (Iterative) Dynamic Programming.
 
 
 
== January 26, 2012: Stethoscope GUI ==
 
; Speaker: Mrunal Gawande
 
; Abstract
 
:Analyzing a textual trace is a difficult task. Would the job be relatively easy if the textual information is presented in a graphical format? This talk presents a GUI front end to visualize the textual trace of MonetDB MAL instructions produced by a tool titled "Stethoscope". The trace could be analysed in an offline and on-line manner. The talk presents a demo of the offline version of GUI along with architecture of the entire tool.
 
 
 
== February 2, 2012: The Scilens Cluster Machine ==
 
; Speaker: Arjen de Rijke
 
; Abstract
 
:Now that the Scilens Machine is up and running, it is time to present it in more details. We will discuss the design, the selection of the hardware, the setup of the software and the intended way the machine is supposed to be used. We will also discuss the selection of the second half of the hardware and ideas on how to use the machine.
 
 
 
== February 9, 2012: Introduction to MonetDB, Part 2 ==
 
; Speaker: Stefan
 
; Abstract
 
<center><strong style="color:red;">MISSING!</strong></center>
 
 
 
== February 16, 2012: Jacqueline ==
 
; Speaker: Fabian
 
; Abstract
 
<center><strong style="color:red;">MISSING!</strong></center>
 
 
 
== March 1, 2012: An introduction to KRIMP: mining patterns with the Minimum Description Length principle==
 
; Speaker: Thibault
 
; Abstract
 
:During this talk, I will introduce some of the work carried out by the Algorithmic Data Analysis group of Utrecht University. Pattern mining considers a database of transactions (e.g. supermarket sales records) and seeks to automatically detect its regularities (e.g. frequent item sets). A common issue is the explosion of results: finding “significant” regularities is a difficult problem. KRIMP addresses it using the Minimum Description Length (MDL) principle. In this context, the best item sets are are the ones that compress the data set best. The main algorithm will be presented, as well as some extensions for other tasks: clustering, comparing two databases, and detecting concept drifts in data streams.
 
 
 
:Sources:
 
:* Arno Siebes, Jilles Vreeken, and Matthijs van Leeuwen. “Item sets that compress”. In SDM , 2006
 
:* Jilles Vreeken, Matthijs van Leeuwen, and Arno Siebes. “Characterising the difference”. In KDD , 2007
 
:* Matthijs van Leeuwen, Jilles Vreeken, and Arno Siebes. “Identifying the components”. In ECML/PKDD (1) 2009
 
:* Matthijs van Leeuwen and Arno Siebes. “Streamkrimp: Detecting change in data streams”. In ECML/PKDD, 2008
 
 
 
== March 8, 2012: Strabon: a Semantic Geospatial Database System==
 
;Speakers: Kostis Kyzirakos and Babis Nikolaou
 
; Abstract
 
:We present the system Strabon, a fully-implemented semantic geospatial database system that can be used to store linked geospatial data expressed in RDF and query them using SPARQL, the standard query language for RDF data. Strabon supports spatial selections, spatial joins, a rich set of spatial functions similar to those offered by geospatial database systems, multiple Coordinate Reference Systems (CRS), and widely accepted serializations for geometric objects. Implementation of Strabon has been based on PostgreSQL and recently has been ported to MonetDB. We present some preliminary evaluation results for both systems and issues encountered during porting to MonetDB. We also present a prototypical implementation in MonetDB of a reasoner for the RCC8 calculus and corresponding performance issues, and discuss future work.
 

Revision as of 16:07, 18 January 2016


The Thursday Think Tank is a weekly talk given at the CWI as part of the Database Architectures group. The talk can be given by either one of the group members, or by an interested party. It is typically given in room L302. Below is the schedule for the weekly talks. Currently, Mark Raasveldt is in charge of managing the TTT schedule.

Agenda