Hasso-Plattner-Institut
  
Hasso-Plattner-Institut
Prof. Dr. Felix Naumann
  
 

Stratosphere

Stratosphere is a joint DFG project conducted by the Technische Universität Berlin, Humboldt Universität Berlin, and the Hasso-Plattner-Institut. It explores how the elasticity of clouds can be exploited for processing analytic queries massively in parallel. Unlike most traditional DBMS, Stratosphere inherently supports text-based and semi-structured data.

Official Project Site

The sub-projects at HPI focus on data quality improvements of linked open data, efficient and scalable data profiling, and knowledge discoevry.

Data Cleansing

We defined the declarative data cleansing language Meteor, implement the underlying basic operations, and develop cost estimations for the operations. Furthermore, we provide test data sets and example queries to evaluate the efficiency and effectivity of the data cleansing process.

Data Profiling

Detecting dependencies in the evergrowing amounts of data has a high computational complexity. One way to cope with this complexity is to distribute the computational work among multiple interconnected computers. However, most existing data profiling algorithms are not designed for parallel execution on computer clusters but rather to run on a single machine. Therefore, we research distributed modifications of existing algorithms as well as new algorithms that can be efficiently executed on computer clusters and that scale out on the number of the cluster nodes.

Knowledge Discovery

Driven by applications such as social media analytics, Web search, advertising, recommendation, mobile sensoring, genomic sequencing, astronomical observations, etc., the need for scalable learning, mining, and knowledge discovery methods is steadily growing. Often the challenge is to automatically process and analyze TBs of evolving data. Extracting value (e.g., understanding the underlying structure and making predictions) from such data, before it is outdated, is a major concern. Therefore, the goal is to enable the scalability of such applications based on Stratosphere.

Please contact Felix Naumann, Toni Grütze (Knowledge Discovery on Stratosphere), or Sebastian Kruse (Data Profiling on Stratosphere) for further questions.

Former members

Publications

SOFA: An Extensible Logical Optimizer for UDF-heavy Data Flows

Astrid Rheinländer, Arvid Heise, Fabian Hueske, Ulf Leser, Felix Naumann
Information Systems, vol. 52:96-125 2015

Abstract:

Recent years have seen an increased interest in large-scale analytical data flows on non-relational data. These data flows are compiled into execution graphs scheduled on large compute clusters. In many novel application areas the predominant building blocks of such data flows are user-defined predicates or functions (UDFs). However, the heavy use of UDFs is not well taken into account for data flow optimization in current systems. SOFA is a novel and extensible optimizer for UDF-heavy data flows. It builds on a concise set of properties for describing the semantics of Map/Reduce-style UDFs and a small set of rewrite rules, which use these properties to find a much larger number of semantically equivalent plan rewrites than possible with traditional techniques. A salient feature of our approach is extensibility: We arrange user-defined operators and their properties into a subsumption hierarchy, which considerably eases integration and optimization of new operators. We evaluate SOFA on a selection of UDF-heavy data flows from different domains and compare its performance to three other algorithms for data flow optimization. Our experiments reveal that SOFA finds efficient plans, outperforming the best plans found by its competitors by a factor of up to six.

BibTeX file

@article{Rheinlaender15,
author = { Astrid Rheinländer, Arvid Heise, Fabian Hueske, Ulf Leser, Felix Naumann },
title = { SOFA: An Extensible Logical Optimizer for UDF-heavy Data Flows },
journal = { Information Systems },
year = { 2015 },
volume = { 52 },
number = { 0 },
pages = { 96-125 },
month = { 0 },
abstract = { Recent years have seen an increased interest in large-scale analytical data flows on non-relational data. These data flows are compiled into execution graphs scheduled on large compute clusters. In many novel application areas the predominant building blocks of such data flows are user-defined predicates or functions (UDFs). However, the heavy use of UDFs is not well taken into account for data flow optimization in current systems. SOFA is a novel and extensible optimizer for UDF-heavy data flows. It builds on a concise set of properties for describing the semantics of Map/Reduce-style UDFs and a small set of rewrite rules, which use these properties to find a much larger number of semantically equivalent plan rewrites than possible with traditional techniques. A salient feature of our approach is extensibility: We arrange user-defined operators and their properties into a subsumption hierarchy, which considerably eases integration and optimization of new operators. We evaluate SOFA on a selection of UDF-heavy data flows from different domains and compare its performance to three other algorithms for data flow optimization. Our experiments reveal that SOFA finds efficient plans, outperforming the best plans found by its competitors by a factor of up to six. },
priority = { 0 }
}

Copyright Notice

last change: Wed, 27 May 2015 16:11:48 +0200