Inactive

UPDATE 2: This will be the final update. The purpose of this update is to provide answers to questions received: Question 3 - We are working on a Dell solution to meet your specs but I had a few quest... UPDATE 2: This will be the final update. The purpose of this update is to provide answers to questions received: Question 3 - We are working on a Dell solution to meet your specs but I had a few questions. Do you need the Chassis and Blade unit to meet the 8u space exactly or is there room for a slightly larger unit. How important is staying within the spaced requested? Answer 3 - We have space restrictions and will therefore need to keep the space requirements listed in the specs, including the 8U requirement for the Compute Cluster blade chassis. Question 4 - Since the Mellanox network adapter is extremely constrained, would you take a requirement to that? Answer 4 – Due to interconnections with existing hardware, we don't have flexibility on the network setup. We have flexibility with the vendor, but we do require EDR InfiniBand 100Gb/s QSFP28 cards with the same number of ports as listed in the specs. Question 5 - Do you have an idea how soon this would be awarded after the RFQ closes? Answer 5 - TBD – tentatively within 60 days after RFQ closing Question 6 - Lastly, what is the requested lead time after award? Answer 6 - Per the RFQ, 60 days after award Question 7 - Would NRL consider NVIDIA T4 GPUs over V100s? T4 advantages: Newer, so longer time until it hits end-of-life About 10% higher boost clock speed: 1515 MHz vs 1380 MHz Significantly lower power consumption: 75 Watt vs 250 Watt 5.7x more memory clock speed: 10000 MHz vs 1752 MHz Answer 7 - Per the RFQ, Offerors can propose comparable items along with product literature for the items being offered so that compliance with the specifications can be evaluated. Question 8 - Should each node be equipped in the following way then? 2 x AMD EPYC 7302 CPUs, each with a min clock speed of 3.00 GHz and 16 cores (128MB L3 Cache, 155W) 128 GB DDR4 ram (8 x 16GB) running at a min speed of 3200MHz (DDR4 ECC RDIMM) 1 x 1TB NVMe SSDs (3D TLC, PCIe 3.0 x4 NVMe, 0.2 DWPD) 1 x 2.0TB SATA Hard Drives (2.5 inch, 6.0Gb/s, 7200RPM) 1 x 2-Port RJ45 Ethernet Controllers 1 x Mellanox ConnectX-5 InfiniBand/VPI Network Adapters, Single-Port EDR (100Gbps) and 100GbE QSFP28, supporting PCIe 3.0 x16 Answer 8 - For the "Compute Server", NRL requests a 2U server chassis that houses 4 nodes, where each node contains items as outlined in the list above. The total number of processors, hard drives, IB cards, amount of RAM etc. across all 4 nodes should match what is listed in Attachment 1 - Specifications. -------------------- UPDATE 1: The purpose of this update is to provide answers to questions received: Question 1: Are blade servers a firm requirement? Answer 1: Yes, blade servers are required for the "Compute Cluster" Question 2: Is the customer really asking for a 2U server chassis with four nodes instead? For example: 2 x AMD EPYC 7302 processors 512 GB RAM (1 or 2) x 1 TB NVMe drive (1 or 2) x 2.0 TB SATA drive 1 x 2-port 10Gb RJ45 Ethernet (1 or 2) x ConnectX-5 IB card Or just one 2U server with the appropriate quantities of components, i.e. 2 instead of 8 processors, etc. Answer 2: For the "Compute Server", NRL requests a 2U server chassis that houses 4 nodes, where each node contains items as outlined in the list above. The total number of processors, hard drives, IB cards, amount of RAM etc. across all 4 nodes should match what is listed in Attachment 1 - Specifications. -------------------- The Naval Research Laboratory (NRL), Information Technology Division, Code 5500 has a requirement for the ability to model networks at Internet scale in order to perform the most realistic evaluations possible, to obtain realistic and statistically significant results, and to make meaningful conclusions regarding the effectiveness of our communication technologies. Internet-scale simulations allow us to improve our understanding of how R&D algorithms and protocols perform on operational networks at scale before incurring the costs and risks associated with a full deployment on real Navy communication networks. In order to meet this need, this action will procure: 1. GPU Server – Quantity 2 2. Storage Server – Quantity 1 3. Compute Cluster – Blade Enclosure (Quantity 1) and Blade Modules (Quantity 20) 4. Compute Server 5. Various System Components / Spares For additional details on the submission of quotes and specifications, please see the attached Combined Synopsis/Solicitation and Attachment 1 - Specifications. The specifications are intended to be descriptive in nature for the types of items that would meet NRL's minimum requirements, and not limit to any particular manufacturer.