SC12 logo  

8th Parallel Data Storage Workshop

held in conjunction with


Chairs: ,

General Chairs: John Bent, EMC and Robert Ross, Argonne National Laboratory

Monday, November 18, 2013
Room 503, Colorado Convention Center
Denver, CO

SC13 Workshop Web Page

abstract / agenda / announcement: probe presentations
committees / call for probe proposals

The proceedings of the 8th Pdsw are now online in the
ACM DIgital Library


Peta- and exascale computing infrastructures make unprecedented demands on storage capacity, performance, concurrency, reliability, availability, and manageability. This one-day workshop focuses on the data storage and management problems and emerging solutions found in peta- and exascale scientific computing environments, with special attention to issues in which community collaboration can be crucial for problem identification, workload capture, solution interoperability, standards with community buy-in, and shared tools. Addressing storage media ranging from tape, HDD, and SSD, to new media like NVRAM, the workshop seeks contributions on relevant topics, including but not limited to:

  • performance and benchmarking
  • failure tolerance problems and solutions
  • APIs for high performance features
  • parallel file systems
  • high bandwidth storage architectures
  • support for high velocity or complex data
  • metadata intensive workloads
  • autonomics for HPC storage
  • virtualization for storage systems
  • archival storage advances
  • resource management innovations
  • incorporation of emerging storage technologies


8:55am - 9:00am

Welcome – Dean Hildebrand, IBM and Karsten Schwan, Geogia Tech
9:00am – 10:00am Keynote Speaker -  Nisha Talagala, Fusion-io
The All-Flash Datacenter
Abstract & Speaker Bio | Slides
10:00am - 10:30am POSTER SESSION 1 - List of participants and links to posters
10:30am – 12:00pm SESSION 1: Dealing the Cards
Chair: Meghan McClelland, Xyratex
  Efficient Transactions for Parallel Data Movement
Jay Lofstead, Sandia National Laboratories*
Jai Dayal, Georgia Institute of Technology
Ivo Jemenez, University of California, Santa Cruz
Carlos Maltzahn, University of California, Santa Cruz
Paper | Slides
  Asynchronous Object Storage with QoS for Scientific and Commercial Big Data
Michael J. Brim, ORNL
David A. Dillow, ORNL
Sarp Oral, ORNL
Bradley W. Settlemyer, ORNL*
Feiyi Wang, ORNL
Paper | Slides
  Performance and Scalability Evaluation of the Ceph Parallel File System
Feiyi Wang, Oak Ridge National Laboratory*
Mark Nelson, Ink Tank Inc
Sarp Oral, Oak Ridge National Laboratory
Scott Atchely, Oak Ridge National Laboratory
Sage Weil, Ink Tank Inc.
Brad Settlemyer, Oak Ridge National Laboratory
Blake Caldwell, Oak Ridge National Laboratory
Jason Hill, Oak Ridge National Laboratory
Paper | Slides
12:00pm - 1:30pm Lunch (not provided)
1:30pm – 2:30pm SESSION 2: Shuffling the Deck
Chair: John Bent, EMC
  Structuring PLFS for Extensibility
Chuck Cranor, Carnegie Mellon University
Milo Polte, WibiData
Garth Gibson, Carnegie Mellon University*
Paper | Slides
  SDS: A Framework for Scientific Data Services
Bin Dong, Lawrence Berkeley National Laboratory, USA
Surendra Byna, Lawrence Berkeley National Laboratory, USA
Kesheng Wu, Lawrence Berkeley National Laboratory, USA
Paper | Slides
 2:30pm – 3:00pm Poster Presentations
3:00pm - 3:30pm POSTER SESSION 2 - List of participants and links to posters
3:30pm – 5:00pm SESSION 3: Playing with a Full Deck
Chair: Carlos Maltzahn, UCSC
  Predicting Intermediate Storage Performance for Workflow Applications
Lauro Beltrao Costa, University of British Columbia*
Samer Al-Kiswany, University of British Columbia
Abmar Barros, Universidade Federal de Campina Grande
Hao Yang, University of British Columbia
Matei Ripeanu, University of British Columbia
Paper | Slides
  Active Data: A Data-Centric Approach to Data Life-Cycle Management
Anthony Simonet, INRIA/University of Lyon*
Gilles Fedak, INRIA/University of Lyon
Matei Ripeanu, University of British Columbia
Samer Al-Kiswany, University of British Columbia
Paper | Slides
  Fourier-Assisted Machine Learning of Hard Disk Drive Access Time Models
Adam Crume, University of California, Santa Cruz*
Carlos Maltzahn, University of California, Santa Cruz
Lee Ward, Sandia National Laboratories
Thomas Kroeger, Sandia National Laboratories
Matthew Curry, Sandia National Laboratories
Ron Oldfield, Sandia National Laboratories
Paper | Slides
5:00pm - 5:30pm Short Announcements

special note - probe presentations

There will be PRObE presentations, demo and posters in the NMC booth, #1732, on the exhibitor floor at SC13 on Tuesday and Wednesday, 1:30 - 2:15 pm MST

program COMMITTEE:

Ahmed Amer, Santa Clara University
John Bent, EMC (General Chair)
Randal Burns, Johns Hopkins University
Andreas Dilger, Intel
Fred Douglis, EMC
Garth Gibson, Carnegie Mellon University and Panasas Inc.
Dean Hildebrand, IBM (PC Chair)
Peter Honeyman, University of Michigan
Song Jiang, Wayne State University
Sanjay Kumar, Intel
Carlos Maltzahn, University of California, Santa Cruz
Meghan Wingate McClelland, Xyratex
Ron Oldfield, Sandia National Laboratories
Narasimha Reddy, Texas A&M University
Robert Ross, Argonne National Laboratory (General Chair)
Karsten Schwan, Georgia Tech (PC Chair)
Keith A. Smith, NetApp
Yuan Tian, Oak Ridge National Laboratory


John Bent, EMC
Scott Brandt, University of California, Santa Cruz
Evan J. Felix, Pacific Northwest National Laboratory
Garth A. Gibson, Carnegie Mellon University and Panasas Inc.
Gary Grider, Los Alamos National Laboratory
Peter Honeyman, University of Michigan
Bill Kramer, National Center for Supercomputing Applications
   University of Illinois Urbana-Champaign
Darrell Long, University of California, Santa Cruz
Carlos Maltzahn, University of California, Santa Cruz
Rob Ross, Argonne National Laboratory
Philip C. Roth, Oak Ridge National Laboratory
John Shalf, National Energy Research Scientific Computing Center,
   Lawrence Berkeley National Laboratory
Lee Ward, Sandia National Laboratories

call for Proposals: Availability of 1000 Nodes for Systems Research Experiments

NSF's PRObE ( operates four clusters to support systems research at scale.  The largest is Kodiak (, which is 1000 nodes (two core x86, 8GB DRAM, two 1TB disks, 1GE and 8Gbps IB) donated by Los Alamos National Laboratory.

Today Kodiak is hosting researchers from Georgia Tech, Carnegie Mellon and Los Alamos.  Princeton researchers have published results from Kodiak at the most recent NSDI (Wyatt Lloyd, "Stronger Semantics for Low-Latency Geo-Replicated Storage", NSDI 2013).  On PRObE staging clusters are researchers from U Central Florida, UT Austin, Georgia Tech and Carnegie Mellon.

PRObE resources are intended for (infrastructure) systems researchers committed to public release of their research results, typically publishing in distributed systems (eg. OSDI or SOSP), cloud computing (e.g. SOCC), supercomputing (e.g. SC or HPDC), storage (e.g. FAST), or networking (e.g. NSDI).

PRObE resources are managed by Emulab ( a cluster manager for allocating physical nodes that has been in use for systems research for over a decade (Brian White, "An Experimental Environment for Distributed Systems and Networks," OSDI 2002).  Users start by porting and demonstrating their code on a 100-node staging cluster such as Denali built from the same equipment donation from Los Alamos.  With demonstrated success on a staging cluster, and a compelling research goal, Kodiak can be requested and allocated, possibly exclusively, for hours to days.

To start using PRObE resources:
- visit to learn about the resources
- visit to request a PRObE-specific Emulab account
- have a research leader or faculty member get an account and define a project on
- use Portal to get onto Denali, to allocate a single node experiment, login into that node to customize and resave the OS image for your project, then launch a multi-node experiment to demonstrate your system at <100 scale
- use to request a large allocation on Kodiak (this is a HotCRP paper review web site, where your paper is a short justification for your research, your preparedness for using Kodiak, and your credientials and appropriateness for using NSF resources)
- PRObE managers will review, approve and schedule your use of large allocations of Kodiak time

In a matter of weeks another style of large PRObE resource will come online.  Susitna is 34 nodes of 64 core x86 processors, for a total of more than 2000 x86 cores.  Susitna also has NVidia donated K20 GPU coprocessors with 2496 cuda cores each, for a total of 84,864 cuda cores.  With 128 GB DRAM, a hard disk and an SSD each, Susitna nodes are interconnected by 40Gbps ethernet, 40 Gbps infiniband and 1Gbps ethernet.

NSF PRObE resources will be available for at least the next two years.

All uses of PRObE resources are obligated to publish their results, either in conferences or one their web sites, and acknowledge NSF PRObE resources used in these publications.

See also our PRObE introduction article in the June 2013 USENIX ;login: vol 38, no 3, 2013 (