The 1st Workshop on Transient Data Management (TDM)

To be held in Conjunction with HPDC 2019, Phoenix, Arizona

With the rise of burst buffers, deep memory hierarchies, and tiered storage, managing data across these various locations has become more complicated. Additionally, emerging application classes, such as AI and in situ analytics, require extremely efficient access to relevant data in order to glean rapid insights as the data is being produced. Applications now try to use faster, but smaller tiers for performance, but must address capacity issues by migrating or evicting data.

Hierarchical storage management (HSM) has traditionally addressed these needs, but with some of the tiers having a byte-addressable interface, rather than a file system (POSIX) interface, HSM is no longer sufficient.

This workshop solicits novel work that explores issues related to managing data as it moves across the memory/storage hierarchy from on package high bandwidth memory, node local NVMe, remote fast memory/storage devices, centralized scratch space, data lakes, and long term archiving solutions. Requested papers will address storing, streaming, accessing, migrating, quality of service, consistency models, searching, and other concerns.

This workshop contributes by exploring the various techniques for data management and movement as well as hardware management and access techniques such as quality of service and API/interfaces for supporting such operations.

Topics of interest include:
• System software/OS features to enable data management operations
• Out of core computation data management techniques
• Tools and techniques to accelerate data searching and selection
• Ensemble run data management techniques and challenges
• In Situ analytics data support
• Data management for large scale streaming applications
• Use of staging areas, such as other nodes, burst buffers or other private or shared acceleration tiers, or even centralized scratch space for managing intermediate data between computation tasks
• Supporting data loading and migration for Big Data or AI workflow systems like Spark, Hadoop, and Tensorflow
• State of the practice papers on the above topics

Papers must be unpublished work in the ACM format no longer than 5 pages (not including references). All accepted papers will be including in the HPDC Workshop Proceedings.

Workshop page: https://sites.google.com/view/tdm2019/home

Submission site: https://easychair.org/conferences/?conf=tdm-2019

Important Dates:
• Submission Deadline: April 4, 2019 (AoE -- firm)
• Responses to Authors: May 7, 2019
• Camera Ready due: May 14, 2019
• Workshop: June 25, 2019

Program Co-Chairs:
• Jay Lofstead (Sandia National Labs) (gflofst AT sandia DOT gov)
• Jai Dayal (Intel) (jai DOT dayal AT intel DOT com)

Program Committee
• Suren Byna (LBL)
• Phil Carns (ANL)
• Elsa Gonsiorowski (LLNL)
• Anthony Kougkas (IIT)
• Johann Lombardi (Intel)
• Kathryn Mohror (LLNL)
• Kimmy Mu (HDF Group)
• Mike Sevilla (Tidalscale)
• Norbert Podhorzski (ORNL)
• Galen Shipman (LANL)
• Min Si (ANL)
• Osamu Tatebe (University of Tsukuba)
• Noah Watkins (Red Hat)
• Weikuan Yu (FSU)