Abstract—The Hadoop Distributed File System (HDFS) is designed to store very large data sets reliably, and to stream those data sets at high bandwidth to user applications. In a large cluster, thousands of servers both host directly attached storage and execute user application tasks. By distributing storage and computation across many servers, the resource can grow with demand while remaining economical at every size. We describe the architecture of HDFS and report on experience using HDFS to manage 25 petabytes of enterprise data at Yahoo!.

In this paper we study how to efficiently perform set-similarity joins in parallel using the popular MapReduce framework. We propose a 3-stage approach for end-to-end setsimilarity joins. We take as input a set of records and output a set of joined records based on a set-similarity condition. We efficiently partition the data across nodes in order to balance the workload and minimize the need for replication. We study both self-join and R-S join cases, and show how to carefully control the amount of data kept in main memory on each node. We also propose solutions for the case where, even if we use the most fine-grained partitioning, the data still does not fit in the main memory of a node. We report results from extensive experiments on real datasets, synthetically increased in size, to evaluate the speedup and scaleup properties of the proposed algorithms using Hadoop. Categories and Subject Descriptors H.2.4 [Database Management]: Systems—query processing, parallel databases

Abstract. The MapReduce parallel computational model is of increasing importance. A number of High Level Query Languages (HLQLs) have been constructed on top of the Hadoop MapReduce realization, primarily Pig, Hive, and JAQL. This paper makes a systematic performance comparison of these three HLQLs, focusing on scale up, scale out and runtime metrics. We further make a language comparison of the HLQLs focusing on conciseness and computational power. The HLQL development communities are engaged in the study, which revealed technical bottlenecks and limitations described in this document, and it is impacting their development. 1

Large-scale data analysis lies in the core of modern enterprises and scientific research. With the emergence of cloud computing, the use of an analytical query processing infrastructure (e.g., Amazon EC2) can be directly mapped to monetary value. MapReduce has been a popular framework in the context of cloud computing, designed to serve long running queries (jobs) which can be processed in batch mode. Taking into account that different jobs often perform similar work, there are many opportunities for sharing. In principle, sharing similar work reduces the overall amount of work, which can lead to reducing monetary charges incurred while utilizing the processing infrastructure. In this paper we propose a sharing framework tailored to MapReduce. Our framework, MRShare, transforms a batch of queries into a new batch that will be executed more efficiently, by merging jobs into groups and evaluating each group as a single query. Based on our cost model for MapReduce, we define an optimization problem and we provide a solution that derives the optimal grouping of queries. Experiments in our prototype, built on top of Hadoop, demonstrate the overall effectiveness of our approach and substantial savings. 1.

Provenance is often used to validate data, by verifying its origin and explaining its derivation. When searching for “causes ” of tuples in the query results or in general observations, the analysis of lineage becomes an essential tool for providing such justifications. However, lineage can quickly grow very large, limiting its immediate use for providing intuitive explanations to the user. The formal notion of causality is a more refined concept that identifies causes for observations based on user-defined criteria, and that assigns to them gradual degrees of responsibility based on their respective contributions. In this paper, we initiate a discussion on causality in databases, give some simple definitions, and motivate this formalism through a number of example applications. 1

MapReduce-based data warehouse systems are playing important roles of supporting big data analytics to understand quickly the dynamics of user behavior trends and their needs in typical Web service providers and social network sites (e.g., Facebook). In such a system, the data placement structure is a critical factor that can affect the warehouse performance in a fundamental way. Based on our observations and analysis of Facebook production systems, we have characterized four requirements for the data placement structure: (1) fast data loading, (2) fast query processing, (3) highly efficient storage space utilization, and (4) strong adaptivity to highly dynamic workload patterns. We have examined three commonly accepted data placement structures in conventional databases, namely rowstores, column-stores, and hybrid-stores in the context of large data analysis using MapReduce. We show that they are not very suitable for big data processing in distributed systems. In this paper, we present a big data placement structure called RCFile (Record Columnar File) and its implementation in the Hadoop system. With intensive experiments, we show the effectiveness of RCFile in satisfying the four requirements. RCFile has been chosen in Facebook data warehouse system as the default option. It has also been adopted by Hive and Pig, the two most widely used data analysis systems developed in Facebook and Yahoo! I.

Prior work has shown dramatic acceleration for various database operations on GPUs, but only using primitives that are not part of conventional database languages such as SQL. This paper implements a subset of the SQLite command processor directly on the GPU. This dramatically reduces the effort required to achieve GPU acceleration by avoiding the need for database programmers to use new programming languages such as CUDA or modify their programs to use non-SQL libraries. This paper focuses on accelerating SELECT queries and describes the considerations in an efficient GPU implementation of the SQLite command processor. Results on an NVIDIA Tesla C1060 achieve speedups of 20-70X depending on the size of the result set.

This paper describes a workflow manager developed and deployed at Yahoo called Nova, which pushes continuallyarriving data through graphs of Pig programs executing on Hadoop clusters. (Pig is a structured dataflow language and runtime for the Hadoop map-reduce system.) Nova is like data stream managers in its support for stateful incremental processing, but unlike them in that it deals with data in large batches using disk-based processing. Batched incremental processing is a good fit for a large fraction of Yahoo’s data processing use-cases, which deal with continually-arriving data and benefit from incremental algorithms, but do not require ultra-low-latency processing.

We present an effective optimization framework for general SQLlike map-reduce queries, which is based on a novel query algebra and uses a small number of higher-order physical operators that are directly implementable on existing map-reduce systems, such as Hadoop. Although our framework is applicable to any SQL-like map-reduce query language, we focus on a powerful query language, called MRQL. Current map-reduce query languages, such as HiveQL and PigLatin, enable users to plug-in custom map-reduce scripts into queries for those jobs that cannot be declaratively coded in the query language, which may result to suboptimal, error-prone, and hard-to-maintain code. In contrast to these languages, MRQL is expressive enough to capture most of these computations in declarative form and at the same time is amenable to optimization. We describe an optimization framework that maps the algebraic forms derived from the MRQL queries to efficient workflows of mapreduce operations that consist of our physical plan operators. We also describe many algebraic optimizations, such as fusing cascading map-reduce jobs into one job and synthesizing a combine function from the reduce function of a map-reduce job. Finally, we report on a prototype system implementation and we show some performance results of evaluating MRQL queries on a small cluster of computers. 1.

MapReduce, especially the Hadoop open-source implementation, has recently emerged as a popular framework for large-scale data analytics. Given the explosion of unstructured data begotten by social media and other web-based applications, we take the position that any modern analytics platform must support operations on free-text fields as first-class citizens. Toward this end, this paper addresses one inefficient aspect of Hadoop-based processing: the need to perform a full scan of the entire dataset, even in cases where it is clearly not necessary to do so. We show that it is possible to leverage a full-text index to optimize selection operations on text fields within records. The idea is simple and intuitive: the full-text index informs the Hadoop execution engine which compressed data blocks contain query terms of interest, and only those data blocks are decompressed and scanned. Experiments with a proof of concept show moderate improvements in end-to-end query running times and substantial savings in terms of cumulative processing time at the worker nodes. We present an analytical model and discuss a number of interesting challenges: some operational, others research in nature.