Partner with vendors to provide technology for NCAR science

CISL maintains close contacts with High Performance Computing (HPC), storage, networking, technology developers, and related vendors as a way to track existing and emerging technologies and to share information with them regarding the computational requirements of the Earth System Sciences. To this end, CISL maintains up-to-date nondisclosure agreements (NDAs) with HPC-related vendors that allow CISL to keep abreast of the latest technical developments and plans in this fast-moving industrial space. CISL also receives loaner equipment and software from vendors and occasionally makes sole-source purchases for evaluating and testing new technologies under these agreements.

CISL’s HPC Futures Lab is an important facility that houses much of this equipment and enables CISL to efficiently manage the complex task of keeping abreast of technological developments and innovations. Many of these relationships with vendors go beyond NDA information sharing and develop into partnerships with shared research and development goals and benefits.

Note that additional details about FY2017 research and technical collaborations accomplishments appear throughout this annual report.

High Performance Computing

CISL/HSS continued participating in the benchmarking and operational testing of various software and hardware technologies as a form of acceptance testing and production readiness. This was done both through loaner programs and via remote access to these new and emerging technologies.

In FY2017, CISL/HSS continued their Joint Center of Excellence collaboration with Hewlett Packard Enterprise (HPE) through engagement on their memory-driven computing architecture and by providing feedback on their cluster management software. CISL also reviewed HPE’s proposed architecture of Gen-Z chip-to-chip protocol for tighter integration of CPU, GPU, memory, and NVMe, then evaluated whether the proposed technologies can improve our application and workflow performance and lower our power footprint.

CISL/HSS also collaborated with Altair to improve the use and functionality of PBS Pro. CISL developers met with PBS Pro developers to share issues and concerns and to collaborate on how the NCAR HPC environment leverages scheduling and resource management technology. CISL provided and demonstrated request-for-engineering (RFE) needs that will help multiple supercomputing sites specifically focusing on the improved use of c-groups, containers, and syslog.

CISL/HSS provided Tulip Infiniband Plugins to the Laboratoire Bordelais de Recherche en Informatique (LaBRI). These plugins create a seamless import of Infiniband network topologies and related data into Tulip. Tulip is a graph visualization tool developed by LaBRI. LaBRI provided some compilation guidance code changes and links to the project from their site.

High Performance Storage

CISL maintains a number of strategic vendor engagements in support of High Performance Storage Solutions. These include NCAR’s centralized high-performance storage environment (GLADE) and its advanced, highly scalable and flexible mass storage and archival resource, the High Performance Storage System (HPSS).

Vendor Relationships

Vendor relationships ensure that we are well positioned to understand emerging technologies in software stack, network, system, storage, and archival systems. The following vendor engagements and non-disclosure agreements are currently in place:

Vendor

Technologies

Data Direct Networks, Inc.

High Performance Storage, Burst Buffers, Cloud Solutions, Storage Servers

Dell

Servers, Disk Subsystems, User Interface Software

Hewlett Packard Enterprise

High Performance Computing Systems

IBM

Servers, Disk Subsystems, Software (HPSS, Spectrum Scale)

Intel

Microprocessors, Compilers, Libraries, Lustre, NVMEoF

Juniper

Ethernet Technologies

Mellanox

Ethernet Technologies, InfiniBand Technologies

NetApp

Disk Subsystems

Oracle

Tape Libraries, Tape Drives, Library Management Hardware and Software

Seagate

Disk Subsystems

Silicon Mechanics

Servers and Storage

Spectra Logic

Archival Storage Systems, NAS Storage Systems, Tape Libraries, Tape Drives, Library Management Hardware and Software, User Interface Software

Super Micro

Servers, Disk Subsystems

Red Hat

Operating System Software, Virtualization Software, Cloud Software

Amazon

Commercial Cloud and Cloud Software

Google

Commercial Cloud and Cloud Software

Microsoft

Commercial Cloud and Cloud Software

Penguin Comp

Commercial Cloud and Cloud Software, Servers, Disk Subsystems, User Interface Software

NVIDIA

Graphics Cards, Graphics Processing Technologies, Deep Learning

Altair

PBS Professional Scheduling Software

SchedMD

SLURM Workload Manager Scheduling Software

AMD

Processor Technology

Cavium

Processor Technology

Cray

High Performance Computing Systems

VMWare

Virtualization, Cloud-based Software

SUSE

Operating System


Funding

These activities are supported with NSF Core funds.