Decoupling Datacenter Storage Studies from Access to Large-Scale Applications

Christina Delimitrou; Sriram Sankar; Kushagra Vaid; Christos Kozyrakis

doi:10.1109/L-CA.2011.37

Decoupling Datacenter Storage Studies from Access to Large-Scale Applications

Christina Delimitrou Stanford

Sriram Sankar Microsoft

Kushagra Vaid Microsoft

Christos Kozyrakis Stanford

IEEE Computer Architecture Letters, 2012

DOI: 10.1109/L-CA.2011.37

Abstract

Suboptimal storage design has significant cost and power impact in large-scale datacenters. This article presents a modeling and generation framework for storage applications that captures spatial and temporal locality and I/O intensity, recreates application I/O loads, and enables storage-system studies without access to application code or full deployments. The methodology is validated with production traces and applied to SSD caching and defragmentation.

Materials

Bibtex

@inproceedings{2012decouplingdatacenterstoragestudiesfromaccesstolargescaleapplications,
  title = {Decoupling Datacenter Storage Studies from Access to Large-Scale Applications},
  author = {Christina Delimitrou and Sriram Sankar and Kushagra Vaid and Christos Kozyrakis},
  booktitle = {IEEE Computer Architecture Letters},
  year = {2012},
  doi = {10.1109/L-CA.2011.37},
}

Hi, we are the Stanford MAST team

Decoupling Datacenter Storage Studies from Access to Large-Scale Applications

Abstract

Materials

Bibtex