Brand_Name_JUSTIFICATION_Redacted.pdf

v1.0_APRIL2016

PARTIAL BRAND NAME SPECIFICATION

ASECC Lab Cluster Upgrade

PR 1300626160

BACKGROUND:

Naval Undersea Warfare Center Division, Newport (NUWCDIVNPT) Code 1041 has a requirement to upgrade aging equipment and to expand the capabilities in the unclassified High

Performance Computing Cluster (HPCC) to allow the Division to model new projects that currently exceed existing processing limitations. In Phase I of the Advanced Scientific and

Engineering Computation Center (ASECC) lab cluster upgrade, a computational node with greater than 6 TB RAM, 40 or more processing cores and 3 times the RAM in local disk space was procured for the Classified HPCC. The ASECC user community requires a computational node with the equivalent capacity in the Unclassified HPCC, to design and test model computations that can be qualified before the model is taken to the classified network. If the

HPCC is not upgraded, future modeling projects which support multiple customers, sponsors, and future systems to be tested as very large models using real classified input will be at risk.

The models needing this capability span multiple NUWCDIVNPT departments and programs.

The models the systems represent are too big to build physical tests and the simulation runs take the place of physical testing. The models of these systems are extremely complex and detailed.

Precision and accuracy are critical and imperative for a given model to represent the real world hardware that is built for an actual vessel. These models are used to evaluate the subsystem's acoustic profile, acoustic response, responses to shocks and Electromagnetic Interference (EMI).

The simulation results constitute a go/no-go on the Navy's evaluation of these systems. Current

ASECC resources, with modifications, can handle parts of these models measured at a single frequency. However, performing the tests in batches this small will take much longer than any of the ASECC client projects can afford. The loss of fidelity in running small batches may constitute a failure of the models being built.

This procurement supports the unclassified ASECC cluster of ASECC Lab 0044 on the NUWC

Unclassified RDTE Core network, eMASS 1761 and adds enhancements to the components added to the ASECC Unclassified Cluster updated in FY12 under NUWCDIVNPT Contract

Number N66604-12-C-2148, to make the unclassified cluster compatible to the capabilities of the ASECC Classified Cluster updated in FY16 under NUWCDIVNPT Contract Number

N66604-16-P-1986. The estimated value of this procurement is $196,218.33 of which the Brand

Name Specified components are estimated at $56,128.00

Period of Performance for Maintenance is three years after delivery, which is 30 days after date of contract.

REQUIREMENT:

The following table lists the Government’s required components for this procurement. Note the

Government’s customized system configurations are contained in an attachment to this document entitled “Government System Configuration Specifications.”

COMPONENT DESCRIPTION QTY.

Hyper Node 4U Rackmount Hyper Compute Node 1

Storage Node 4U Rackmount Storage System 1

Compute Nodes 2U Rackmount Compute Node 2

40GBs (40/10 GBE

QSFR/SFP+)

Ethernet Switch

A NIAP or DISA APLITS approved 40Gig Ethernet

QSFP/SFP+ switch

1U Tray Console Keyboard, mouse, monitor mountable in an 19 inch rack mountable 1U pull out tray

System Footprint

All items must fit within 14U of 19 inch rack space, except the analysis workstation

N/A

Maintenance and

Warranty

- All software must include three (3) years of maintenance from time of delivery, WHICH IS WITHIN 30 DAYS

AFTER DATE OF CONTRACT.

- All hardware must include a three (3) year warranty from the original equipment manufacturer from time of delivery and include the following:

phone support;

next business day parts; and failed disk replacement program allowing the

Government to retain the failed drive for destruction and only provide a photocopy of the label for proof of failure.

-No on-site support representative is required.

N/A

BRAND NAME SPECIFICATION:

Items that are brand name components with no substitutions allowed are as follows:

HYPER NODE PART NUMBER QTY

Intel Xeon E7-8890 v3 2.5GHz

Eighteen-Core Processors

Intel Xeon Processor

E7-8890 v3 (45M

Cache, 2.50GHz)

Crucial MX300 1 TB SATA III

2.5 inch SSD configured as

RAID 6

CT1050MX300SSD1 22

Crucial MX300 500 GB SATA

III 2.5 inch Solid State Mirrored

Boot Drive

CT525MX300SSD1 2

Red Hat Enterprise Linux 6

(RHEL6) Server, Standard, (4

Sockets - Up to 1 Guest), installed w/ 3 years support

Red Hat Enterprise

Linux Server, 6, Standard (Physical or

Virtual Nodes)

Red Hat Scalable File System

Add-On, Standard, installed w/ 3 years support

Red Hat Scalable File

System Add-On, Standard

STORAGE NODE PART NUMBER QTY

Intel Xeon E5-2643 V4 3.4GHz

Six-Core Processor

Intel Xeon Processor

E5-2643 v4 (20M

Cache, 3.40GHz)

Crucial MX300 1 TB SATA III

2.5 inch SSD for mirrored system disks

CT1050MX300SSD1 2

Red Hat Enterprise Linux 6

(RHEL6) Server, (1-2 sockets, up to 1 guest), installed w/ 3 years support

Red Hat Enterprise

Linux Server, 6, Standard (Physical or

Virtual Nodes)

Red Hat Scalable File System

Add-On, Standard, installed w/ 3 years support

Red Hat Scalable File

System Add-On, Standard

Compute Node PART NUMBER QTY

Intel Xeon E5-2667 v4 3.2GHz

Eight-Core Processor

Intel Xeon Processor

E5-2667 v4 (20M

Cache, 3.20GHz)

Crucial MX300 1 TB SATA III

2.5 inch SSD system disk

CT1050MX300SSD1 4

Crucial MX300 275 GB SATA

III 2.5 inch SSD system mirror disks

CT275MX300SSD1 2

Red Hat Enterprise Linux 6

(RHEL6) Server, (1-2 sockets, up to 1 guest), installed w/ 3 years support

Red Hat Enterprise

Linux Server, 6, Standard (Physical or

Virtual Nodes)

This procurement is for HPCC enhancements within the unclassified NUWCDIVNPT ASECC cluster. This requirement addresses the need for the capability of processing exceptionally large models where data space in RAM and local storage exceed all capacity available at

NUWCDIVNPT and systems accessible to NUWCDIVNPT on the DoD HPCC program. The enhancements consist of a hyper compute node, a storage node, two compute nodes, and increased network throughput on the ASECC unclassified cluster backend to support multiple projects that require a minimum of 6 Terabytes of shared RAM memory of active data space.

The enhancements will address the need to expand current productivity levels when servicing

ASECC users’ mathematical computation requirements. By increasing the processing capability, jobs will finish faster freeing up solver licenses for other jobs thereby increasing overall output.

Procurement of the components listed in the attached component system configuration, some of which are specified brand name components, are required to update obsolete scientific computational equipment that runs physics-based modelling methods of design. Larger models are being designed that require state of the art technology.

The specified brand name components included in this procurement, and as shown in the table above, are vital to ensure NUWCDIVNPT, Code 1041, maintains a homogeneous environment for ASECC Lab 0044 to handle the larger physics-based models being built at the Command.

Further discussion of each specified brand name component is as follows:

1. INTEL PROCESSORS

Intel Processors are required for compatibility with existing systems at NUWCDIVNPT.

The Intel Xeon E7-8890 v3 2.5 GHz 18-Core Processor, the Intel Xeon E5-2643 V4 3.4 GHz

6-Core Processor and, the Intel Xeon E5-2667 v4 3.2 GHz 8-Core Processor, are 4 th

Generation Intel Xeon v3/v4 processors of the Haswell family of Intel processor microarchitecture, with improved graphics, battery life, and security over earlier architectures.

The Intel Haswell family’s instruction set includes the Advanced Vector Extensions (AVX)

256-bit instruction set with wider vectors, new extensible syntax and rich functionality with improved performance for transcendental mathematics. These capabilities are incorporated into the code of the primary mathematical solvers currently in use in the ABAQUS and LS-

DYNA software application suites running on Unclassified and Classified Clusters in the

Advanced Scientific and Engineering Computation Center (ASECC). The ASECC

Unclassified Cluster runs on the NUWCDIVNPT RDT&E Classified Network, Lab 902, eMASS ID 1761. The ASECC Classified Cluster runs on the NUWCDIVNPT RDT&E

Classified Network, Lab 903, eMASS ID S-221.

The ASECC Cluster workload is expanding and the ASECC’s user requirements have increased. Since five to sixteen software licenses are required per ABAQUS job, for example, the number of jobs that can be run in parallel is limited by the total number of software licenses. In order to maintain current productivity levels when servicing ASECC users’ computation requirements, the increased processing capability of the Haswell AVX instruction set is needed to use the code incorporated into the code of the mathematical solvers run on the ASECC Classified Cluster. By increasing execution rates jobs finish faster, freeing up solver licenses for other jobs to utilize, thereby increasing overall output, compensating for the lower number of software licenses.

Software and hardware running in the ASECC computational cluster currently run on Intel

Processors. The three different nodes included in this requirement (the hyper node, the storage node, and compute nodes) execute the same instruction set and share the same architecture. Therefore, they all require Intel processors for compatibility with the existing

ASECC application software and cluster hardware and are the only components that meet the

Government’s minimum requirements.

2. CRUCIAL SOLID STATE DRIVES

Crucial Solid State Drives (SSD) are required for compatibility with existing systems in the

ASECC unclassified lab. To keep the lab’s high performance computing systems functioning at the highest efficiency levels and maximum availability, it is necessary to provide high speed storage to supplement the large memory footprint of these systems;

support the largest Abaqus models that are run on the system; and to ensure uninterrupted runtime during long scientific calculations. The storage devices must be capable of managing the system storage disk space at least at memory speed levels; otherwise the system storage disk space will be unable to keep up very large calculations being performed and will prematurely terminate these calculations.

The ASECC staff, working with NUWCDIVNPT users, measured the necessary performance for large striped (data across 2 or more disks) storage devices. Using the fastest available rotational disk drive technology (15000 RPB disks) did not meet the throughput needs of the extremely large Abaqus models being run on the ASECC clusters. To meet these needs, the

Crucial SSDs are needed. The size of the SSDs for the storage disk is a function of

NUWCDIVNPT’s largest model currently being executed on the ASECC clusters. At 2.7

Tera Bytes (TB) this model exceeds all current ASECC resources. Currently

NUWCDIVNPT users cannot run models to produce the expected results for their sponsors.

The Crucial MX300 SSD family incorporates Redundant Array of Independent NAND

(RAIN) technology from Micron, its parent company, to provide a 1.5 million hours Mean-

Time-Between-Failure (MTBF) rating, Power Loss Protection, and Data Path Protection; it employs a thermal throttling mechanism to protect against overheating, and Dynamic Write

Acceleration caching technology built into the NAND for power efficiency. Since these devices will be used as a SSD RAID to provide disk storage for the model, the drive speed across the entire disk space is the most important factor. Other manufacturer products specify faster transfer rates, but those rates are for a limited (less than 10%) amount of the total drive's capacity. Since NUWCDIVNPT’s users work on models that exceed the total capacity on each of these drives and require striping across all of them for speed and capacity, fast cache on each drive will be overrun long before it can be emptied into the much slower primary NAND storage on each drive. The Crucial implementation provides a consistent transfer rate across the entire individual MX300 drives capacity.

The ASECC Unclassified lab, in meeting the utilization requirements of NUWCDIVNPT’s users, requires SSDs that are tested to perform at the highest currently available transfer rates across the entire storage capacity and have longevity, measured in total TBs transferred in the life of the drive, that are above 300 TBs transferred without any of the drives showing degradation. This requirement is necessary since these drives will be used in situations where massive amounts of data will be written and deleted continuously for large portions of the calendar year.

The costs of including an additional manufacturer is the increased risk of failure while in operation. This will drive down the maintainability numbers for sustained compute node operations and the ability to bring user models to completion. Additionally, any loss in functionality forces users to wait in queues longer and those running will consume available limited software tokens longer. Put a different way, if drives with lower MTBFs were included instead, more instances of compute node down time due to drive failures are expected.

The Crucial SSDs drives are self-encrypting, thus strengthening NUWCDIVNPT’s security posture on this system.

In addition to the requirements for compatibility with existing systems in the ASECC lab, and faster performance required for the mathematical computations that may run for several hours and even days, it is critical that the SSD drives used do not malfunction during multiple daily runs. The Crucial SSDs are the only components that meet the Government’s minimum requirements.

3. RED HAT ENTERPRISE LINUX 6 SERVER

The ASECC cluster currently runs on Red Hat Enterprise Linux 6.0 (RHEL6) (approved under Department of Navy Application and Database Management Systems (DADMS) ID

79260) Operating System. For interoperability with existing cluster hardware and software, system compatibility, management, and for Information Assurance (IA) compliance with the

Defense Information Systems Agency (DISA) Security Technical Implementation Guides

(STIGs) and with National Information Assurance Partnership (NIAP) requirements, the

RHEL6 must be installed on all new nodes included in this procurement. The RHEL6 is the only software license that meets the Government’s minimum requirements.

4. RED HAT SCALABLE FILE SYSTEM

With RHEL6 as the operating system and over 14TB of disk storage on the Hyper Node and the Storage Node, Red Hat requires this add-on to make all of the disk space available as one volume. The Red Hat Enterprise Linux Add-on is the only file system add-on compatible with the RHEL6, thereby making it the only add-on that meets the Government’s minimum requirements.

RESELLERS:

As part of Code 1041’s initial market research, the aforementioned list was presented to several potential sources. This requirement will be synopsized/solicited on the Federal Business

Opportunities (FEDBIZOPS) website and all vendors capable of meeting the requirements will be considered for award.