krl3_questions.docx

Here are our questions for this project:

- Section 3.1 "HPC system shall interface with facility and existing hardware and the existing power supply"

- define existing hardware 3 phase 60 amp receptacles 460R9W

- describe how the new system will interface with facility and existing hardware (expected connections, configurations requirements etc)

- what is the existing power supply (description, make/model) Power and network interfaces (copper)

- RHEL licensing - will NSWC provide, or do they want it included in the bid price?

Building wide UPS and Generators are the back up power sources

- What is the available power type & connector type and how many whips - 208v 60A 3 phase IEC320 0U or 208v 30A single phase L6-30 1U?

Need to have the rack PDUs to be - 208v 60A 3 phase IEC320 0U

- How many tons of available cooling does the data center have to specifically support this cluster?

Unknown, however it should be more than adequate for the HPC requirements. Awaiting specific information from the Site Manager and will follow up when we have that information.

- How many kW per rack can be supported by power and cooling?

Unknown, however it should be more than adequate for the HPC requirements. Awaiting specific information from the Site Manager and will follow up when we have that information.

- Is there adequate height from the dock to the data center for delivery of the HPC as a turnkey solution? (estimated height of 84") Awaiting specific information from the Site Manager and will follow up when we have that information.

- 42U rack will be 750mm wide x 1200mm deep in order to support the length of the systems & PDUs - verify that is acceptable.

Acceptable

- Can the racks be side by side with side panels removed to allow for between rack cabling? This would reduce the costs for cabling.

Yes

- Sections 3.1.3 requested specifications for Cluster Management Server are overly robust, is a less robust system that is specifically designed for our cluster management software acceptable?

Yes

- Section 3.1.2 refers to a 4 port KVM, is this expected to be included in the proposal, and if so, what make/model is required?

Needs to be included in proposal. No make/model requirements. Needs to integrate into the rack and be able connect to Management Server, Login Node, Scratch Storage, and Cold Storage servers. Must also be able to plug into compute nodes (if needed).

- Section 3.1.6 "hardware RAID controller" - this is not possible due to the NVMe drives. Verify that this is acceptable and not required.

Acceptable

- Section 3.1.7 The requested read/write throughput of 6GB/s is not possible with the requested quantity of drives. Verify that this is acceptable and not required.

Typo, that should have read 6Gb/s as a minimum requirement. Acceptable.

- Section 3.1.8 "Tightly integrated Cluster Expansion and Infrastructure Interconnect reliability, availability, and serviceability (RAS)" - What is the expected solution that would meet this requirement?

Capability to manage operating system and management system remotely via web GUI and/or command line via a network connection.

- Section 3.2.1 "Verify and test the integration of the thermal and shutdown process with the on-site APC is" - Appears to be an incomplete statement. What is the complete requirement for this item? Are there existing thermal sensors in place with the current APC setup? What is expected to be integrated with the existing APC?

Requirement: “Verify and test the integration of the thermal and shutdown process with the on-site APC is working properly with the system administrator.”

This requirement is no longer needed for the facility we are building this HPC in. The installation site has backup power needed, but if cooling fails, we need to ensure the system can safely shutdown if thermal thresholds are met. Thermal monitoring solution will need to be provided.

- Section 3.2.1 "Verify and test required software with code engineers for functionality of installed software" - What are the software packages and what tools are required to support them? Will NSWC install the software?

All software required to run a cluster (including, but limited to, RHEL OS, RHEL repos, Cluster Management System Software, Cluster software, scheduler, etc.). Testing of additional government propriety software is not required.

- Section 3.2.2 Clarify the expected procedure for support/troubleshooting and warranty service. Conflicting statements regarding support:

"All troubleshooting and repair work shall commence within 48 hours of the reported problem and is to be conducted onsite at NSWCCD" "Technical Support The Contractor shall perform technical assistance for operation and troubleshooting of the systems identified in the PWS via phone/email" Business hours level of support (8x5) with next day on-site service option if needed for serious cluster issues. Most issues can be addressed using phone/online technical support.

- Section 3.2.2 "At no time shall hard drives & other non-volatile parts be returned, including for warranty or maintenance purposes." & Section 5.0 - "With the exception of hard drives and other non-volatile parts, the Government may return and release custody of defective parts of a system" - Many devices (motherboards/power cables/controllers/backplanes/power supplies etc.) have small amounts of non-volatile storage ex: BIOS, BMC, flash etc.. Exactly what components can and cannot be returned for warranty?

Specifically, components that may store classified information within its non-volatile memory. (This is usually just limited to hard drives)

- Section 8.0 What are the security requirements and what is the procedure to obtain clearance?

Will need to be answered by NSWCCD Security Office 301-227-1408

- Request due date be extended to allow for a second round of questions to address any additional questions that arise from the answers during the initial question period. See amended solicitation

- Request due date extension to allow for time to address the addendum, one business day is not enough. amended solicitation

Regarding the requirement that work must be completed within five consecutive business days:

We are concerned that this timeframe is too short to ensure the quality and thoroughness of the work. Based on our experience, we believe a minimum of six weeks is necessary to complete the project effectively.

Acceptable.

Regarding the requirement for system installation within a Top-Secret Facility:

As our team does not currently hold the necessary clearances, we would like to understand how access to the facility will be managed to allow us to perform the required installation.

The TPOC can and will escort. Please contact NSWCCD Security Office 301-227-1408 for more specific information.

Regarding the completion of system installation by August 31, 2024, with a performance period not exceeding five days:

This timeline appears very tight, assuming that all equipment would need to be delivered by August 20, 2024, to meet the installation deadline. Given that equipment delivery is dependent on the vendor and may not all be available by then, we are concerned that the schedule may be too aggressive. Could this timeline be reconsidered to account for potential delivery delays?

Yes, we should be able to push the date out into FY25.