Science of High-Performance Cloud Systems (HPCloudSys)

The first Workshop on the Science of High-Performance Cloud Systems (HPCloudSys) is one of the workshops co-located with CloudCom 2016, Luxembourg, December 12, 2016.

Scope

The First Workshop on the Science of High-Performance Cloud Systems (HPCloudSys) seeks to bring together researchers and practitioners who work at the intersection of HPC and Cloud Computing to foster conversations about the systems aspects of HPC computations in the cloud. While the cloud promises cheap, efficient, and elastic compute cycles, there are a number of challenges at a systems level to running efficient HPC computations in the cloud. There has been a significant interest in recent years in science and engineering research computing in cloud environments. Work on these topics (some are listed below) is necessary in order to ensure effective and efficient computation in the cloud. There also arise many methodological challenges, in terms of the ways that environments are built, computations are orchestrated, and data are collected, archived, and shared. Our motivation in running this workshop is to foster more conversation about clouds for research from the perspective of cloud computing and the core recognized strengths of cloud systems, rather than from a perspective of “HPC taken to the cloud.”

Topics

Topics of interest include (but are not limited to):

• Systems-level aspects of computing in the cloud, including:
  • Low-latency and low-jitter virtualization in the cloud
  • Storage and network support for HPC applications in the cloud, including computation-aware resource provisioning
  • Technologies for capturing execution environments, such as containers
• Tools and methodologies for HPC in the cloud, including:
  • Orchestration of computation in the cloud
  • Dealing with variability of computations in the cloud
  • Collecting, processing, archiving, and cataloging of results
• Reproducibility and repeatability of science in the cloud

Discussions at the Workshop

The workshop will facilitate exchange of experience in the following areas:

• “HPC for the 99 Percent”, “Production Cloud”, “Deeply Programmable Cloud” – What do these slogans actually mean at this point in time? Our overall goal is to come up with a common understanding of the range of state-of-the-art technologies that allow to build systems that best satisfy researchers’ needs.
• Scale versus Functionality – Where are the discussed systems in this spectrum?
• Integration – How are these systems integrated with other campus and national cyberinfrastructures?
• Return on investment - why should the NSF invest in clouds
• Lessons Learned – What major design and technology transformations have occurred since these systems have been proposed and deployed?

The workshop will feature two keynote speakers (TBD): one from academia, and one from industry. There will also be brief presentations from cloud facilities that are available to researchers, to help acquaint participants with facilities and technologies that may benefit their work. The remainder of the time will be used for presentation of accepted papers, with ample time reserved for discussion. Attendance will be open to all (not just authors).

Important Dates

Submissions

Submissions should be in the IEEE CS format (no longer than 5 pages). Submissions must be original and should not have been published previously or be under consideration for publication while being evaluated for this workshop.

Accepted workshop papers will be published in a supplement of the Proceedings of IEEE CloudCom 2016, and submitted to IEEE Xplore (conditioned by the presentation at the workshop by the author).

Authors are invited to submit papers through the conference submission system: https://easychair.org/conferences/?conf=hpcloudsys16

Organizers

Program Committee