Scholars Central

Conference abstractPublished 2017

Memory management in modern data analytics platforms

Authors: Mayuresh Kunjir

Duke University, USA

Ask about this article

Abstract

Select any words in the title or abstract, or press this and choose a sentence.

The field of data processing is seeing an explosion of computation platforms such as Hadoop, Spark, HBase, Cassandra, Flink, Storm, and others. Some key observations of these platforms are: They are JVM-based and therefore developer friendly, stable, and secure; they run in multi-tenant environments where resources are isolated using containers, e.g. Yarn, Mesos, and Docker and; they use memory extensively both for data storage and computations. The central premise of this work is that data analytics will increasingly be done in memory on shared-nothing clusters running JVM inside containers. Given this, it is high time for a better understanding of memory management in these modern data analytics platforms. This work uses Thoth, a data-driven platform for multi-system cluster management developed at Duke University, to build a deep understanding of different interplays of memory-management decisions made at multiple levels such as the resource manager (e.g. Yarn), the application framework (e.g. Spark), and the JVM. An empirical study is carried out over a suite of industrystandard analytical workloads to show the impact of various memory management options on response time, efficiency, and reliability. A set of guidelines follows to ensure a safe and efficient execution of the workloads.

Ask about this article

Answers come only from this article's title, abstract and keywords, each sentence pointing to where it comes from. Your question is sent to OpenAI to write the answer. Scholars Central does not store it.

Try:

Similar by topic

More from this journal

The newest in Journal of Computer Engineering & Information Technology.

All its articles

Conferences in this field

Upcoming conferences in this field.

Follow

New articles from what you follow go into your email digest, once you agree to alert emails.

  • JournalJournal of Computer Engineering & Information Technology
  • SubjectComputer Science & IT
  • SubjectEngineering
  • BylineArticles whose byline prints “Mayuresh Kunjir”

For authors and organisers

Publish, present and keep up with your field

Find the right journal for your paper and submit your manuscript.

Photo: Pexels