Enclosure No. 2B - Appendix B Statement of Work.docx

Appendix B Argonne National Laboratory Statement of Work ALCF 400G Fabric September 23, 2025

1.0 Background

The Argonne Leadership Computing Facility (ALCF) requests proposals for the acquisition and installation of a 400 Gigabit Ethernet (IEEE 802.3bs, IEEE 802.3cu, & other 802.3 standards as appropriate) network fabric for interconnection of all computational resources at the ALCF. This fabric will provide access to a wide range of data storage resources, both collocated and remotely located, as well as data acquisition instruments across the entire National Laboratory complex. The fabric must seamlessly interconnect multiple generations of high-performance computing technologies in one grand unified Ethernet environment.

To maximize reliability, availability, and serviceability, the ALCF expects this fabric to be implemented using redundant pairs of chassis-based carrier-class modular systems. However, fabrics built out of individual scalable component units will also be given consideration. Throughout this Statement of Work, the ALCF will refer to the complete 400G network interconnect as "the fabric" and may refer to individual components of "the fabric" as "fabric devices" or "switches" interchangeably.

2.0 Key Words

The ALCF adheres to RFC-2119 as the governing document for the definitions of important words used in this Statement of Work. Proposals shall be evaluated using the following five definitions copied verbatim from RFC-2119. The ALCF does not adhere to RFC-8174 capitalization in this Statement of Work: the following five definitions apply regardless of the use of lowercase or uppercase.

· MUST: This word, or the terms "REQUIRED" or "SHALL", mean that the definition is an absolute requirement of the specification.

· MUST NOT: This phrase, or the phrase "SHALL NOT", mean that the definition is an absolute prohibition of the specification.

· SHOULD: This word, or the adjective "RECOMMENDED", mean that there may exist valid reasons in particular circumstances to ignore a particular item, but the full implications must be understood and carefully weighed before choosing a different course.

· SHOULD NOT: This phrase, or the phrase "NOT RECOMMENDED" mean that there may exist valid reasons in particular circumstances when the particular behavior is acceptable or even useful, but the full implications should be understood and the case carefully weighed before implementing any behavior described with this label.

· MAY: This word, or the adjective "OPTIONAL", mean that an item is truly optional. One vendor may choose to include the item because a particular marketplace requires it or because the vendor feels that it enhances the product while another vendor may omit the same item. An implementation which does not include a particular option MUST be prepared to interoperate with another implementation which does include the option, though perhaps with reduced functionality. In the same vein an implementation which does include a particular option MUST be prepared to interoperate with another implementation which does not include the option (except, of course, for the feature the option provides.)

3.0 Scope

The Offeror must provide a quote for the system and services specified in this Statement of Work. Line items must be sub-totaled into the following categories:

· Hardware

· Software

· Licensing

· Installation

· Initial Training & Education

· Ongoing Maintenance & Support for 5 Years from Date of Acceptance

3.1 Technical Requirements

The Offeror must provide an equipment offer that satisfies all of the following Technical Requirements For each bulleted Technical Requirement throughout, the Offeror must describe how each required technology will be made available in the product or the product roadmap

· This fabric will serve all Ethernet-based connectivity to a massive, heterogeneous, multi-vendor and multi-generational supercomputing facility

· This fabric must have forward-compatibility and extensibility to support new and emerging technologies, such as, but not limited to:

· 800 Gigabit Ethernet (IEEE 802.3df)

· 1.6 Terabit Ethernet (IEEE 802.3dj)

· Ultra EthernetTM Specification v1.0.1

· This fabric must have backward-compatibility to support existing and current state-of-the-art technologies, including, but not limited to:

· 100 Gigabit Ethernet (IEEE 802.3ba, 802.3bj, 802.3bm, and other 802.3 standards as appropriate)

· IEEE 802.3ad Link Aggregation and 802.1AX-2020 Link Aggregation with LACP

· NVM ExpressTM over Fabrics Revision 1.1a

· RoCE v2 (InfiniBand Architecture Specification 1.2.1 Annex A17 2014-09-02 or later)

· Direct GPU to Fabric Communication (NVIDIA GPUDirect, AMD ROCm-aware MPI over OpenUCX)

· Data Center Bridging, including, but not limited to

· IEEE 802.1Qbb Priority-based Flow Control

· IEEE 802.1Qaz Enhanced Transmission Selection

· IEEE 802.1Qau Congestion Notification

· IEEE 802.1AB Link-Layer Discovery Protocol and Data Center Bridging Capabilities Exchange Protocol (DCBX)

· Multi-Chassis Link Aggregation (MLAG or MC-LAG)

3.1.1 Fabric Design Requirements

· The fabric must be built using currently-shipping, General Availability hardware and software. Pre-release or Beta hardware or software offers will be rejected

· The fabric must be a non-blocking Clos topology

· The fabric must be self-discovering, meaning that no manual configuration is required to add switches to grow the fabric: addition of a switch into the fabric must be done automatically when links are physically connected between existing and new switches

· The fabric must be self-healing, meaning that loss of any one switch component does not impact the integrity of the fabric as a whole and that replacement of a failed switch automatically triggers integration of that component into the fabric without further human interaction after the hardware replacement

· The fabric must be built with fully redundant components, meaning that all switches, cables, connections, management interfaces, power supplies, and any other equipment critical to functionality must be capable of failure and replacement without impacting fabric functionality

· The fabric must dynamically identify and configure all parallel links between two switches and aggregate those links for maximum performance

· The fabric, once accepted, must be able to be configured, monitored, and maintained by a single part-time engineer

· The fabric must be built in such a way that no programming or scripting is required for day-to-day operations for the life of the fabric. Day-to-day operations is defined as Move/Add/Change/Delete order processing for provision of Layer 2 services, as defined in the Open Systems Interconnect (OSI) model, to consumer devices such as server and storage Network Interface Cards (NICs). Layer 2 service delivery will be performed via Edge Ports (see section 4.1.3)

· Preference will be given to Offerors that supply a fabric meeting the following software design requirement:

· All software required to manage the fabric is integrated into the fabric switch devices themselves. In other words, any flow controller, subnet or fabric manager, or similar control plane is self-hosted within the network operating system on every device and no external equipment or connectivity is required.

· For systems that do not comply with the above requirement, the following requirements must be met:

· All external management hardware must be included in the Hardware quote

· All external management software must be included in the Software quote

· All external management hardware & software must be deployed on-premises at the ALCF facility

· All external management hardware & software must be fully capable of exposing all features with no limitations when fully isolated from public Internet connectivity

· On-site training and education must be provided for the entire ALCF Operations Team (estimate 50 students) on how to install, manage, maintain, and perform disaster recovery of these proprietary management systems

· Preference will be given to systems with no ongoing license maintenance requirements. In other words, any system where full functionality is achieved for the life of the equipment without any requirement to pay additional licensing fees or any requirement to maintain licensing files, keys, codes, or similar, will be scored highest

· For systems that require maintenance of licensing, the following shall apply:

· All licenses required for the full planned growth of the fabric (8192 Edge Ports) must be included with the initial installation

· All licenses must be perpetual: licenses must never expire and must never require renewal

· All license costs must be itemized in the detailed quote

3.1.2 Fabric Configuration Requirements

All configuration required to completely wipe and rebuild the fabric, across all fabric switches, from a factory default state, must be line-oriented human-readable plaintext capable of being copied directly into the fabric control plane. To meet this requirement, the fabric devices and fabric management infrastructure must provide the following configuration capabilities:

· An RS-232 and TIA-232-F (R 2012) compliant local console with full administrative functionality (Note: Embedded USB serial controllers do not meet this requirement!) At a minimum, the Transmitted Data, Received Data, and Common Ground circuits of the RS-232/TIA-232-F local console must be exposed to the exterior of every fabric device unit using a commonly available non-proprietary physical interface such as an 8P8C Modular Connector or a D-Subminiature connector

· An SSH Version 2 (RFC 4251 and many later improvements) compliant remote console with features identical to the features available on the local console

· The ability to use either of these interfaces to retrieve a complete copy of the entire fabric configuration expressed as a single line-oriented human-readable plaintext file

· The ability to use either of these interfaces to import the above required configuration file of the entire fabric and have said configuration applied successfully with no further human interaction

· The ability to transfer the above required configuration file to a configuration backup repository using the industry standard OpenSSH secure file copy utility (scp)

· Preference will be given to implementations that automatically synchronize fabric configuration to an external backup repository whenever a change is performed

3.1.3 Port Requirements

The ALCF has aggressive growth targets for this fabric, based upon the anticipated acquisitions of significant computational and storage assets over the next five years. To this end, the fabric must be built from a technology that will facilitate easy growth to meet these targets.

· Preference will be given for modular systems with little or no cabling infrastructure required for this incremental growth

· Modular systems should be as large as possible and delivered with as many open slots as possible for future expansion

· Preference will be given to solutions including at least two fully redundant modular units

3.1.3.1 Edge Port Definition

Edge Ports are defined as the locations where the myriad ALCF-operated servers, storage, and existing network systems will connect to the fabric. Edge port growth targets serve as a guide for the scaling requirements of the fabric.

3.1.3.2 Fabric Port Definition

Fabric Ports are defined as the links between fabric devices (switches) and are never used for customer connections.

3.1.3.3 Edge Port Quantities

· At the time of acceptance, the fabric must provide 512 non-blocking Edge Ports of 400 Gigabit Ethernet

· The fabric offered must be capable of growing to a minimum of 8192 non-blocking Edge Ports of 400 Gigabit Ethernet

· Based on the modularity of the hardware offered, the fabric must include sufficient supporting hardware infrastructure (defined as mechanical, electrical, cooling, and related infrastructure such as chassis, slots, rails, power supplies, fans, and all other non-Ethernet-Port devices necessary for operation of the Fabric) to support incremental growth to this final fabric size of 8192 ports. For example:

· If the fabric is built using large chassis systems with slot-based switches, sufficient slots must be included at the time of acceptance to support, at a minimum, the final fabric size of 8192 ports

· If the fabric is built using discrete components, an itemized list of all components required to facilitate each growth increment is required

· Offeror must describe the fabric growth mechanism. Required details that must be included in the growth description include:

· Minimum Edge Port growth increment (for example, standard port quantity per blade)

· Minimum Fabric Port growth increment

· Edge Port quantities that will trigger a requirement for additional Fabric Ports

· Offeror must identify Edge Port growth increments that will result in oversubscription of the fabric and describe a corresponding growth increment for the Fabric Ports to eliminate the oversubscription

3.1.3.4 Edge Port Physical Interface

· The ALCF strongly prefers Quad Small Form-factor Pluggable Double-Density Multi-Source Agreement (QSFP-DD MSA) Edge Ports for maximum interoperability with existing equipment, existing cabling, and in-progress system acquisitions

· Octal Small Form Factor Pluggable Multi-Source Agreement (OSFP) connectors are acceptable but will be scored lower. These connectors are incompatible with the existing ALCF network deployment and will impose additional integration costs outside the scope of this Statement of Work

· Edge Port physical layer connectivity must be delivered via industry-standard duplex 1.25mm ferrule Type LC IEC 61754-20 PC (Physical Contact) terminations supporting ISO/IEC 24702-compliant OS2 Single-Mode Fiber

· Edge Ports must not be terminated with Multi-fiber Push On (MPO) connectors. Fabrics where Edge Port transceivers are only available with MPO technology must include all ancillary hardware to expose duplex Type LC IEC 61754-20 PC terminations

· Edge Ports must support installation of 100G Lambda Multi-Source Agreement 400GBASE-FR4 transceivers for connections to 400G Edge Devices

· Edge Ports must support installation of CWDM4 Multi-Source Agreement 100GBASE-CWDM4 transceivers for connections to 100G Edge Devices

· The offered fabric must support the use of generally available Multi-Source Agreement (MSA)-compliant transceivers from any manufacturer with no degradation in performance or functionality

3.1.3.5 Edge Port Logical Interface

3.1.3.5.1 Link Aggregation Requirement

· All ALCF systems connecting to Edge Ports will use IEEE 802.3ad LACP-based Link Aggregation Groups. As such, the Edge Port Logical Interface must supply the following:

· At a minimum, every connected edge device will require two 400GBASE-FR4 connections in an active-active aggregation group

· The fabric must support 4096 simultaneous Link Aggregation Groups

· Preference will be given to fabrics capable of supporting 8192 or more simultaneous Link Aggregation Groups

· The fabric must support 16 Edge Ports in a single Link Aggregation Group

· Preference will be given to fabrics capable of supporting 64 or more Edge Ports in a single Link Aggregation Group

3.1.3.5.2 Agnostic Ethernet Network Requirement

· The fabric must not require any drivers, software, daemons, or other services on edge systems: all systems must be able to take equal and full advantage of all Fabric functionality with only the layer 2 LAG serving as the demarcation

· The fabric must not impose any hardware requirements, such as specific or proprietary Host Channel Adapters or Network Interface Cards, on the edge systems: equal and full performance of all Fabric functionality must be achievable regardless of the 400G Ethernet PHY used by the ALCF for each interface

3.1.3.5.3 Buffer and Latency Requirements

· The ALCF will be connecting both low and high latency end user systems to this fabric. This convergence of Local Area Network (LAN) and Wide Area Network (WAN) must be seamlessly supported by the fabric

· For latency-sensitive end user systems, the fabric should provide a maximum Edge Port to Edge Port Layer 2 latency, on the longest switching path, of under 32 microseconds

· Preference will be given to fabrics capable of providing an edge-to-edge longest-path L2 latency of under 16 microseconds

· For WAN-enabled end user systems, the fabric must provide sufficiently deep buffers to support multiple simultaneous intercontinental wire-speed data transfers

· To be acceptable, the fabric must provide a minimum of 32GB of packet buffer available to any WAN-enabled Edge Port, based on a target of 8 parallel line-rate streams with an 80ms Round Trip Time (RTT) or 4 parallel line-rate streams with a 160ms RTT

· If the Offeror’s fabric contains a mixture of hardware with different latencies or different buffer sizes, Offeror must explain in detail what methods are necessary to allocate either type and what hardware or software restrictions are in place that will limit the ability to assign ports to devices in the future

· Preference will be given to systems that can achieve both low latency for local traffic while also providing deep buffers for global traffic while doing so completely transparently. In other words, fabrics that can supply both low latency and deep buffers on every Edge Port will be most highly ranked for this requirement

3.1.3.6 Fabric Port Quantities & Interfaces

· The quantity of Fabric Ports will be driven by the Offeror’s product and capacity requirements. Any Fabric Port technology will be acceptable if it meets the non-blocking Clos requirement

· The ALCF will consider all available transceiver and optical fiber technologies as valid for Fabric Ports, with a preference for solutions requiring a minimal number of cables or no cables at all (i.e., an internal chassis backplane would be highest ranked above any cabled connections)

· The Offeror must describe the fabric port interconnect, both in terms of internal hard-wired midplane/backplane equipment as well as external cabled connections

3.1.4 Fabric Monitoring Requirements

· The fabric must be able to be completely monitored using Simple Network Management Protocol (SNMP) Version 2c. "Completely monitored" means that all available port counters for all Edge Ports and Fabric Ports are visible using SNMP queries. In addition, all hardware environmental status (such as but not limited to power redundancy, power consumption, internal device temperatures, and fan status) must be visible using SNMP queries

· Offeror must provide details of any hardware counters explicitly unavailable using SNMP and explain how to retrieve these counters in the absence of SNMP support

· The Offeror should include advanced monitoring software, including all licenses and subscriptions required, with detailed documentation of what advantages are gained when using the proprietary monitoring software

· Preference may be shown for well-articulated responses demonstrating advanced monitoring systems and techniques that go above and beyond the features available in industry-standard SNMP monitoring systems

· The ALCF is particularly interested in solutions that provide automated Root Cause Analysis of outages

· Fabrics which cannot remain fully functional and completely independent when disconnected from their monitoring software will be considered externally managed fabrics and must meet all the hardware and software requirements of sections 4.1.1 and 4.1.2 to be considered acceptable. Advanced monitoring software will only increase a response’s ranking when it is not an explicit requirement to enable all fabric functionality

3.1.5 Fabric Cabling Requirements

· All cables must be labeled with identical information at both ends of the cable

· The ALCF must approve the format and placement of all labels prior to delivery

· Cables must be installed in a professional and skillful manner and in compliance with all applicable regulatory codes

· Preference will be given to installers trained and certified to industry standard cabling techniques, such as NECA 1-2023 "Standard for Good Workmanship in Electrical Construction" and NECA/FOA 301-2009 "Standard for Installing and Testing Fiber Optics (ANSI)”

3.2 Installation

The Offeror must provide an on-premises installation of the fabric in pre-existing modern four-post square-hole air-cooled cabinets (42U, 45U, or 48U depending on location) that will be provided by the ALCF. Detailed rack and power specifications of the proposed installation location will be provided in APPENDIX 1 Argonne is extremely safety conscious, and the Offeror must adhere to all Argonne safety regulations while working on-site. A Job Safety Analysis (JSA) will be jointly prepared by Argonne safety staff and the Offeror prior to arrival for installation. This document will contain all safety requirements applicable to the installation.

Upon completion of installation, the fabric must be fully functional and ready to receive custom configuration from ALCF personnel to integrate fully with the existing Layer 2 and Layer 3 networks already in place within the ALCF. Functional & performance acceptance testing of the equipment will be conducted to verify manageability and configurability conforms to the SOW requirements. Functional & performance acceptance testing details will be provided in APPENDIX 2

3.3 Maintenance and Support

· Five years of maintenance and support must be quoted from the date of acceptance. Pricing must reflect two options:

· Purchase of all five years at the time of acceptance

· Purchase of one year at the time of acceptance with four additional annual renewals scheduled and priced up front

· A product lifecycle timeline must be included showing a detailed plan for continued hardware and software support for the offered products for a minimum of 10 years from the date of acceptance. Put another way, the ALCF requires a written commitment to renewal of support for an additional 5 years at the 5th anniversary of acceptance

· Preference will be given to any offers including a written commitment to 15 years of hardware and software support from the date of acceptance

· For any field-replaceable hardware component that is not considered fully redundant, Offeror must include a minimum of 2 spare copies of said component upon delivery, to be kept as on-site spares. This requirement only applies to customer field-replaceable components. Factory-built permanent equipment, such as metal chassis, bolted backplane boards, or other non-field-replaceable parts are exempt

· The Offeror must provide 24-hour Support Engineer availability and next-business-day advance hardware replacement.

· "Support Engineer" is defined as any support personnel capable of evaluating and interpreting debug diagnostic output from the Network Operating System and Fabric Devices without further escalation

· The ALCF Business Day is 0700-1900 Central Time, Monday through Friday, year-round

· The Offeror must provide an enterprise-grade support experience with direct access to a Support Engineer after ticket creation. Support systems that require interactive ticket triage before reaching knowledgeable personnel is undesired

· The Offeror should provide a Support Engineer response to all tickets, regardless of severity, within 1 hour of ticket creation

· The Offeror should provide a maximum delay of 24 hours from ticket creation to shipment of Advance Hardware Replacement components if a hardware RMA is requested by ALCF personnel. The ALCF will not perform in-situ hardware debugging on the production fabric and expects immediate replacement of any component exhibiting hardware error without delay

· The Offeror must describe in detail the ticket creation and escalation process and customer experience. The Offeror must describe and diagram their ticket process flow

· The Offeror must not include on-site support after the completion of acceptance: the support contract must permit the ALCF Operations Team to perform all hardware and software maintenance for the life of the equipment

4.0 Government-Owned Property

No government-owned property will be provided.

5.0 Security

Access to the Argonne National Laboratory campus is controlled and requires a pre-arranged Gate Pass. Individuals who are not US citizens are allowed on site but require a detailed security investigation and will need to be identified and documented a minimum of 45 days prior to installation.

Access to the ALCF Data Center is also controlled. An ALCF staff member will be assigned to work with the Offeror and enable access while the installation team is on site.

6.0 Cybersecurity

The Offeror shall describe its process for identifying and remediating security vulnerabilities in all hardware and software delivered by the Offeror, including vulnerabilities discovered by third parties, vulnerabilities in upstream components, and vulnerabilities discovered by the Offeror but not yet publicly known. In particular, the Offeror shall describe timelines and processes for obtaining and disseminating Common Vulnerability Enumeration (CVE) identifiers for vulnerabilities discovered within Offeror’s products.

Any products provided by the Offeror must have patches and fixes released in a timely manner and in full compatibility with the existing Network Operating System or Systems as provided in the offer. It is expected that disclosure and remediation practices will follow industry standard best practices as well as government directives. Examples of applicable government directives that will apply to this procurement include, but are not limited to:

· Cybersecurity & Infrastructure Security Agency (CISA) Binding Operational Directive (BOD) 23-02: Mitigating the Risk from Internet-Exposed Management Interfaces

· CISA BOD 23-01: Implementation Guidance for Improving Asset Visibility and Vulnerability Detection on Federal Networks

· CISA BOD 22-01: Reducing the Significant Risk of Known Exploited Vulnerabilities

· CISA BOD 20-01: Develop and Publish a Vulnerability Disclosure Policy

· CISA BOD 19-02: Vulnerability Remediation Requirements for Internet-Accessible Systems The Offeror must describe their practices in sufficient fashion to document compliance with these and any other applicable CISA BOD.

The Offeror must provide the ability, documentation, and automation required to rotate every secret in all components in the system to a unique, non-shared, site-assigned and Offeror-unknown credential without requiring physical interaction with the hardware. This includes, but is not limited to passwords, tokens, SSL certificates and keys, SSH certificates and keys, code signing keys, out-of-band management passwords, and all secured secrets used for system operation whether exposed in the device configuration or hidden in the Network Operating System.

The Offeror must provide the ability and documentation to disable all unencrypted communication channels, such as but not limited to telnet and HTTP, with no impact to system reliability, availability, or serviceability.

The Offeror must describe in detail how all default accounts and credentials are to be disclosed to the ALCF upon system delivery.

The Offeror must commit in writing to disclosure of any future instantiation of credentials after delivery, for example during system patching or updates, using detailed release notes.

7.0 Place of Performance

The fabric will be installed in the Building 240 Data Center on the Argonne National Laboratory campus. The Offeror must install the equipment into cabinets prepared by the ALCF on site. The Offeror is welcome to pre-assemble and test the system off site, but it must be transported to the ANL campus disassembled and reassembled on site.

8.0 Period of Performance

The fabric should be delivered as soon as possible based on component availability. Installation and acceptance testing should begin as soon as is practical after system delivery, but not more than 90 days after delivery.

Appendix 1: Installation & Physical Location Details The ALCF is housed in a single large data center in Building 240, the Theory and Computing Science Building, on the Argonne National Laboratory campus. The data center is a single contiguous raised floor facility with a combination of air-cooled, indirect-water-cooled (air to water heat exchanger), and direct-water-cooled systems. Pertinent details of the facility are itemized herein for the convenience of the Offeror.

As stated in section 2.3, cabinets will be pre-existing and should not be included in the offer. Technical details of the facility relevant to equipment delivery and installation are as follows:

· Standard floor tile grid of 24x24 inches

· Height from concrete floor to top of tile: 48 inches

· Floor tile rolling load limit: 1250lbs

· Floor tile concentrated load limit: 1500lbs

· Overall facility dimensions: 128ft x 322ft

· Air-cooled facility dimensions: 128ft x 208ft Three locations in the air-cooled section have been identified for deployment of fabric hardware, indicated in Orange. The upstream WAN provider demarcation is indicated in Yellow. Approximate cable paths are also marked in Orange (exact cable paths will be identified during installation planning).

Figure 1: Installation Locations The primary 400G location, at the upper right of Figure 1, will be equipped with the following cabinet and power. It is expected that the first 512 Edge Ports be in this location and occupy a single 48U rack:

· 1 Hewlett-Packard Enterprise (HPE) 48U 1200mm G2 Enterprise Shock Rack

· 4 HPE Cray Supercomputer 60A 208V 3-Phase Master 12 CX series PDU (2 UPS, 2 Non-UPS)

· Maximum continuous power consumption & heat rejection to air: 28kW The existing 100G & 400G ALCF Core is located near the top center of Figure 1, near the upstream WAN demarc. To connect with the existing equipment, as well as facilitate growth and deployment of additional equipment in the empty space at the center of the facility, a rack has been identified in this location and will be made available for installation of fabric hardware. The Offeror may choose to use this as a location for components (for example, a spine chassis) in addition to the delivery of 512 Edge Ports at the Primary location. The rack in the network core is as follows:

· Chatsworth Products International (CPI) N-Series Teraframe Gen 2 Cabinet, 800x1200mm

· 2 Server Technology 60A 208V C3W36RL-DQJE2MT2 Switched PDU

· Maximum continuous power consumption & heat rejection to air: 14kW

· Cable tray distance between Primary location and Core location: approximately 120 feet The third location identified in Figure 1, at the lower left, will be the first major fabric expansion and is anticipated to be exercised approximately 1 year from the date of acceptance (see Section 4.1.3.3). Cabinets and power requirements for this new location have yet to be determined. The location information is included to empower the Offeror to provide a robust and expandable architecture facilitating easy expansion across the great distances in this unique supercomputing data center. The approximate cable tray distance between the Primary location and the Expansion location is 320 feet.

Appendix 2: Functional & Performance Acceptance Testing Details To verify that the delivered fabric conforms to all SOW requirements, a series of functional and performance acceptance tests will be conducted. A finalized Acceptance Test Plan (ATP) will be negotiated between the winning Offeror and the ALCF no less than 30 days prior to fabric delivery. The following are examples of tests that may be included in the ATP.

Test Overview The Acceptance Test Period is planned to start no later than 10 business days after the Offeror notifies the customer Technical Representative that the fabric is fully installed and ready to be tested and the Technical Representative has acknowledged that notification.

The Acceptance Test will consist of component tests, integration tests, and reliability tests.

Test Period Prerequisites

· All fabric hardware installed and functioning (Offeror)

· Management network connectivity verified (Offeror and ALCF)

· All required management hardware and software installed and functioning (Offeror and ALCF)

· All administrative access credentials provided to ALCF Failure Resolution Any defect that prevents an Acceptance Test stage from progressing will be considered a failure of acceptance testing. At the end of the Acceptance Test Period, for each failure, the Offeror will provide: an explanation for failure, a root cause (if known), or a documented plan to determine the root cause if not yet known. The ALCF will classify each failure using a Severity Level for the defect. Table 1 defines the Severity Levels and the expected responses.

Examples of Tests

· Validate management console connectivity using SSH per the requirements of Section 4.1.2

· Validate local console connectivity per the requirements of Section 4.1.2

· Execute a variety of management console commands, as appropriate to the fabric management platform, to document fabric device functionality. Examples of typical configuration commands may include, but are not limited to:

· “show tech-support”*

· “show interfaces”*

· “show transceiver”*

· “shutdown”* and “no shutdown”* of interfaces and optical media

· Instantiation and deletion of interface configuration

· Instantiation and deletion of Layer 2 and Layer 3 configuration, such as VLAN creation, router configuration, and similar

* Or equivalent command relevant to the management platform provided

· Validate optical transceiver installation, detection, and interoperability using various transceivers supplied by the customer, per the requirements of Section 4.1.3.4

· Validate fabric component failure, removal, re-installation, redundancy, and dynamic component deletion and discovery by inserting and removing various components as appropriate (such as, but not limited to switching cards, network modules, optical modules, management modules, power supplies, and similar) to verify the fabric meets the Redundancy, Reliability, Availability, and Serviceability requirements of Section 4

· Validate interoperability by interconnecting the fabric with the existing ALCF 100Gigabit Ethernet and 400Gigabit Ethernet equipment using industry-standard components

· Validate the network topology, using “show” commands or other utilities as appropriate, to confirm the fabric is nonblocking

· Measure network latency using Verbs RoCE latency tests between ALCF-provided equipment connected to Edge Ports

· Measure network bandwidth Verbs RoCE throughput tests between ALCF-provided equipment connected to Edge Ports

· Validate available Deep Buffers, using “show” commands or other utilities as appropriate, to demonstrate that memory buffers are allocated and freed as traffic loads vary between Edge Ports Collaboration and Revision The tests to be selected, and the terms of success or failure, are subject to mutually agreed-upon revision to ensure that they remain up-to-date and relevant. Tests, benchmarks, and acceptance criteria can be modified as needed with mutual agreement from the customer and Offeror.

Deliverable and Data Sharing Customer and Offeror will co-author a final report of test results to inform the final Acceptability determination by the ALCF. All tests should be concluded. Customer and Offeror agree to share all performance data and system outputs for the tests in the agreed-upon ATP. The final report will be delivered within 10 business days of the completion of all tests.

image1.png