Conferência

Decreasing Memory Footprints for Better Enterprise Java Application Performance

23rd International Conference on Database and Expert Systems Applications, DEXA 2012

Stoyan Yordanov ; João Cachopo2012Springer Berlin / Heidelberg

Informações chave

Autores:

Stoyan Yordanov (Stoyan Yordanov Garbatov); João Cachopo (João Manuel Pinheiro Cachopo)

Publicado em

Setembro 2012

Resumo

In this paper, we present a work for reducing the memory footprint of enterprise Java applications. The work relies on the predictions provided by stochastic models of the applications’ data-access patterns. The models, built during the execution of the application, are used both at compile-time, to control the in-memory representation of data, and, at run-time, to decide which portions of the data to load. The combined effect of these two approaches allows for an effective reduction in the memory used by the application, leading to a significant performance improvement. We evaluate the newly developed approaches on the TPC-W benchmark, with different database sizes, and show that our solution increases the benchmark throughput by 10.78% on average, with a maximum of 35.43% when operating over larger databases.

Detalhes da publicação

Editora

Springer Berlin / Heidelberg

Título do contentor da publicação

23rd International Conference on Database and Expert Systems Applications, DEXA 2012

Local da conferência

Vienna, Austria

Primeira página ou número de artigo

430

Última página

437

Domínio Científico (FOS)

computer-and-information-sciences - Ciências da Computação e da Informação

Palavras-chave

  • heap management
  • in-memory object representation
  • persistence

Idioma da publicação (código ISO)

eng - Inglês

Acesso à publicação:

Acesso apenas a metadados