Authors: Rachana Ananthakrishnan (University of Chicago), Vas Vasiliadis (University of Chicago)
Abstract: The storage needs of investigators at research institutions are growing in scale and complexity, driving the rapid evolution of storage environments in advanced computing centers, from "scratch + project space" to a diverse collection of systems spanning multiple storage tiers and deployment options. These changes are placing unprecedented demands on advanced computing system administrators who are forced to make difficult tradeoffs between capacity, access modalities, economics, and operational efficiency. This session will bring together research computing professionals and storage providers, to engage in a dialogue on the challenges they face and the strategies they are adopting to address them.
Long Description: Our goals in this session are to (a) identify and describe the most critical challenges facing research computing professionals in supporting the storage needs of their constituents, and (b) facilitate the sharing of ideas for addressing these challenges.
The topic of storage management has become a top priority in most advanced computing environments. In ad hoc discussions with research computing administrators and leadership over the past 2-3 years it has become clear that meeting the storage needs of investigators requires constant juggling of demand and supply, and makes it difficult to achieve stable and cost-efficient operating environments. The Supercomputing conference brings together many professionals from this community with a wide ranging set of perspectives on addressing this topic; we anticipate that they will welcome an opportunity to hear from their peers and exchange ideas.
We have attended past BoF sessions focused on specific storage-related topics, e.g., approaches to managing archival storage, that focus on tactics for solving narrowly defined problems. We believe that there is interest in approaching these topics in a more holistic manner, and from a more strategic and long-term perspective.
A diverse group of institutions who recognize the relevance of this topic have already expressed interest in sharing their insights. Presenters will include representatives from the following:
- City University of New York (Invited speaker: Alexander Tzanov, Director, CUNY HPC Center)
- Penn State University (Invited speaker: Gary Skouson, Senior HPC Engineer)
- Princeton University (Invited speaker: Curtis Hillegas, Associate CIO, Research Computing)
- Purdue University (Invited speaker: Preston Smith, Executive Director, Research Computing)
- University of California, Irvine (Invited speaker: Philip Papadopoulos, Director, Research Cyber Infrastructure Center)
- University of Florida (Invited speaker: Erik Deumens, Research Computing Director)
- University of Minnesota (Invited speaker: James Wilgenbusch, Director of Research Computing)
Many of the invited speakers are widely recognized thought leaders within the research computing community, which we believe will help attract attendees to this session. To further enrich the dialogue, we also plan to invite representatives from storage solution providers to attend the session and provide perspectives "from the supply side".
We hope to explore how storage management strategies take into account economics/cost efficiency, access modalities, data migration needs, support for hybrid (on-premise/public cloud) environments, and networking considerations. The discussion will be guided by the following questions, among others:
- Which use cases are placing the greatest demands on your current research storage infrastructure?
- How prepared is your institution to address these demands?
- What does your future state look like for research storage?
- What strategies are you considering/adopting to achieve the future state?
- What hurdles do you anticipate as you evolve your storage environment?
- How are current storage offerings aligned (or not) with your future plans?
We expect that attendees will broaden their understanding of issues in managing research storage and identify with some of the strategies discussed as potential options for improving their own environments. At minimum, we hope to surface a core set of topics that can be discussed in greater depth at subsequent events.
URL:
Back to Birds of a Feather Archive Listing