A2_02838_Cluster_SOW.pdf

STATEMENT OF WORK

TITLE: High performance computer expansion

LAB REQUESTING SERVICE: Communications Technology Laboratory (CTL), Radio Frequency Technology (RFT) Division (672)

I. BACKGROUND INFORMATION

The Radio Frequency Technology (RFT) Division of the Communications Technology Laboratory (CTL) utilizes electronic design automation (EDA) simulations and high frequency finite element (FE) simulations to predict the properties of circuits and to compare with the results of experiments. The RFT division has a state of the art research laboratory to support the most sensitive research done at the National Institute of Standards and Technology (NIST).

Included in this building is a High-Performance Computation (HPC) Cluster for finite element simulations of electromagnetic systems, including but not limited to millimeter-wave circuits, transistors, amplifiers, and filters.

II. SCOPE OF WORK

The HPC cluster requires additional compute nodes for a wide range of applications including:

simulations of patterned circuit boards for microwave circuits, dielectric resonators, waveguides, silicon, and metal sheet structures. In addition, a log-in visualization node is also required to permit users to visualize simulation results. The system is required for a variety of new programs that are fundamental to NIST mission critical work. One of the key aspects of this tool is its ability to run high frequency structure simulation software and interface with the existing hardware. It is critical that the HPC cluster have as little down time as possible during the expansion and integration of the new compute nodes. NIST requires that the integration of the new compute nodes minimally impact ongoing work, which requires seamless integration and experiential knowledge that the high frequency structure simulation software operate on the new compute nodes and integrate with existing hardware. Failure to integrate will mean that it is impossible to accurately model the microwave circuit characteristics or achieve repeatable circuit boards that are required for NIST mission critical work.

This purchase consists of a log-in visualization node and computational node(s), a high-performance interconnection interface, associated cabling, operating system, scheduling program, service and on-site installation contracts, 5-year hardware warrantees, and 5-year service contracts. A noteworthy feature of the system is that the hardware must run FE simulation software (specifically, High Frequency Structure Simulation by ANSYS) through remote simulation managers over the NIST network and integrate with existing hardware.

III. SPECIFICATIONS

The Contractor shall provide one (1) rack-mount log-in visualization node and one (1) rack-mount compute node, and associated signal cabling, as described below.

Log-in and visualization node (1 required):

• The node must include a 2U Dual Socket capable of supporting a GPU.

• The node must have a CPU or CPUs totaling 8 cores or more, capable of running at least 16 simultaneous execution threads, clocked at least at 2.1 GHz or faster, with Intel Xeon Silver 4110 architecture.

• The node must have at least 512 GB (at least 16 GB/core) of registered ECC RAM, at a speed of DDR4-2666 MHz or faster.

• The node must be enclosed in a rack-mount chassis, 19-inch width, and the rail kits must be included.

• The chassis must also include redundant power supply for the motherboard, CPUs, and RAM, capable of operating on single-phase power.

• The node must have a motherboard capable of driving the CPUs and RAM at their full rated speeds.

• The node must have one (1) 2.5” RAID 1 total Volume 1024 GB capacity SDD (or twice the total RAM in each node), SATA drive, with a rated transfer speed of 6 GB/s or faster. SATA controllers must be included.

• The node must include a Nvidia Titan XP graphics card or better.

• The node must have an InfiniBand host channel adapter, 4×QDR 40 Gb/s or faster equivalent, QSFP connector or equivalent.

• The node must have two (2) gigabit or faster ethernet ports. These ports may be provided by the motherboard or by an accessory card.

• The node must have one dedicated ethernet baseboard management controller for the intelligent platform management interface. These ports may be provided by the motherboard or by an accessory card.

• The node must have all power cords and communication cables included.

Compute node (at least 8 required):

• The node must have a CPU or CPUs totaling 12 cores or more, capable of running at least 24 simultaneous execution threads, clocked at least at 2.1 GHz or faster, with Intel Xeon Silver 4110 architecture.

• The node must have at least 512 GB (at least 16 GB/core) of registered ECC RAM, at a speed of DDR4-2133 or faster.

• The node must be enclosed in a rack-mount chassis, 19-inch width, and the rail kits must be included.

• The chassis must also include redundant power supply for the motherboard, CPUs, and RAM, capable of operating on single-phase power.

• The node must have a motherboard capable of driving the CPUs and RAM at their full rated speeds.

• The node must have one (1) 2.5” RAID 1 total Volume 1024 GB capacity SDD (or twice the total RAM in each node), SATA drive, with a rated transfer speed of 6 GB/s or faster. SATA controllers must be included.

• The node must have an InfiniBand host channel adapter, 4×QDR 40 Gb/s or faster equivalent, QSFP connector or equivalent.

• The node must have two (2) gigabit or faster ethernet ports. These ports may be provided by the motherboard or by an accessory card.

• The node must have one dedicated ethernet baseboard management controller for the intelligent platform management interface. These ports may be provided by the motherboard or by an accessory card.

• The node must have all power cords and communication cables included.

• A total of minimum 128 total cores are required. For example, 4 compute nodes with 32 cores would yield 128 total cores.

Software (required):

• Vendor must install the operating system and perform associated benchmarking. Appropriate operating systems include CentOS 7 or higher, or RedHat. Compatibility with listed applications is required.

• Vendor must install FE High Frequency Structure Simulator (HFSS) simulation software by ANSYS. License provided by NIST.

• Vendor must install MATLAB R2018a. License provided by NIST.

• Vendor must install C, C++, FORTRAN 2008 compilers.

• Software installation can be on site or at the vendor’s facility.

• Vendor must install head node cluster control software and scheduler compatible with the software.

• Vendor must supply cables to connect the nodes ethernet ports to the ethernet switches.

Support (required):

• Vendor must include 5-Year 24x7, 4-Hour, on-site hardware support.

• Vendor must be able to provide as needed on-site software support.

• Vendor support must include guaranteed forward compatibility and availability of replacement parts for 5 years.

• Vendor support must include guaranteed integration with existing hardware.

Testing (required):

• Vendor must complete burn-in testing on all components, and verify device functionality with the operating system and software.

o Benchmarking for cluster computing: LAPACK or equivalent.

• Vendor must perform benchmark simulations on the head node and compute nodes o Benchmarking for HFSS: “PCB Differential Via” or equivalent.

Warranty:

The Contractor shall include at least a five (5) year warranty on all hardware.

IV. PERIOD OF PERFORMANCE

The period of performance shall be five (5) years from award of this contract.

V. PLACE OF PERFORMANCE

All work shall be completed at the Contractor’s facility. `1 Installation of the instrument and training shall be accomplished at NIST, Boulder, CO. Normal duty hours are Monday through Friday, 8:30 a.m. to 5:00 p.m. except Federal holidays.

VI. DELIVERABLES

Description Quantity Due Date Log in node One (1) 45 days after award of this requirement.

Compute node Eight (8) 45 days after award of this requirement Cabling All necessary required 45 days after award of this requirement.

Burn-in testing Once Prior to shipping from vendor facility.

Software benchmarking Once Prior to shipping from vendor facility or during on-site installation.

VII. PERFORMANCE REQUIREMENT SUMMARY

The Contractor shall demonstrate that all performance standards in this statement of work have been met.

VIII. RISK ASSESSMENT

This acquisition is not for services or construction, so risk is not formally assessed. The vendor will not require access to the NIST site, nor will they be granted access to NIST computer accounts or networks.

IX. GENERAL INFORMATION

Safety:

The Contractor employee shall be responsible for knowing and complying with NIST installation safety prevention regulations (http://www-i.nist.gov/mml/safety/policies/index.htm.) Such regulations include, but are not limited to, general safety, fire prevention, and waste disposal.

Patent Rights:

The Government retains a Government use license to all inventions arising from this work.

Security:

NIST is a restricted campus. An identification badge is required for access for entry into buildings and is shown to the armed Security Police when entering the campus.

Identification Badges:

Contractor employees shall comply with NIST identification and access requirements. The Contractor employee is responsible for absences due to expired identification and access documents. If on site, each Contractor employee shall wear a visible identification badge http://www-i.nist.gov/mml/safety/policies/index.htm provided by the NIST Security Office. The badge must show the full name, title, and if required by NIST, the words “Contractor” in front. The Contractor employee shall turn in the NIST identification badge and vehicle pass to the TPOC, COR, or Contracting Officer (CO) upon termination of their services under this contract.

Vehicle Registration:

All Contractor employees must register their vehicles with the NIST Security Office to gain access to the campus. A valid driver’s license, Government-furnished civilian ID, proof of insurance and current registration must be presented to the NIST Security Office, at which time a NIST vehicle pass will be issued. The pass shall be displayed on the vehicle’s rear-view mirror in accordance with instructions. The Contractor employee shall follow NIST procedures for removal and turn-in of the vehicle pass upon termination of services under this contract.

Media Inquiries:

The Contractor employee shall not respond to any media inquiries. Any inquiries from the media shall be immediately relayed to the TPOC, COR, and/or CO. There shall be no interviews, comments, or any other response without the knowledge and approval of the NIST Director.