Step 5: Storage Configuration & Mechanical Installation

The disk subsystem of the industrial-grade DePIN node is based on the principle of strict separation of roles. We categorically reject the idea of storing everything on a single disk. To optimize costs, protect the server’s logic, and guarantee maximum uptime (SLA), our architecture uses two physically separate disks. At this stage, we are focused exclusively on selecting the hardware, configuring the network connections, and ensuring proper physical installation within the chassis.

1. System Configuration: SATA SSD

Storage Type: A reliable, time-tested 500 GB solid-state drive (SATA SSD).
Its purpose: To provide a purely hardware-based deployment environment for the kernel of the future Ubuntu Server operating system, the Docker environment, and the node management scripts. The speed of a standard SSD is more than sufficient for running the Linux kernel and management utilities. It is not economically feasible to allocate expensive server storage resources for this purpose.
Connection: This drive connects directly to any available SATA port on our Supermicro X11-DAI-N motherboard using its original, standard SATA cable. For power, connect it using a high-quality cable from the main power supply.
Installation: The SSD itself must be securely fastened with screws in the mounting bay or on the slides of a 4U chassis. Do not allow the drive to “dangle” from the cables: the constant airflow from powerful chassis fans creates microvibrations that can damage the connector pins over time.

2. Computing Circuit: U.2 NVMe SSD

Drive type: Strictly enterprise/server-grade U.2 NVMe drive (Intel/Solidigm, Samsung enterprise series) with a capacity of 2 TB. The use of standard consumer-grade M.2 NVMe drives under 24/7 DePIN workloads is not permitted due to their tendency to overheat quickly and their low total bytes written (TBW).
Its purpose: Purely computational workload. This storage device will be used by the network exclusively for the rapid exchange of large data sets during the server rental process (loading and processing neural network weights, the io.net AI computation cache, and rendering textures). It has tremendous endurance (high DWPD rating) and is equipped with PLP (Power Loss Protection) hardware protection technology—built-in capacitors will safeguard the customer’s data in the event of a power outage.

Block diagram of U.2 NVMe connectivity via a PCIe x8 slot:

Since all four full-size PCIe x16 slots on our Supermicro X11-DAI-N motherboard are occupied by graphics cards, we’re using an available PCIe x8 slot to connect a high-performance U.2 drive.

Adapter: A specialized expansion card (PCIe to U.2 / Mini-SAS HD SFF-8643 Adapter) is installed in the PCIe x8 slot. When purchasing, choose high-quality shielded cards to prevent interference from neighboring graphics cards.
High-speed cable: A specialized shielded Mini-SAS HD data cable (SFF-8643 to SFF-8639) runs directly from the adapter board to the drive. The cable must not have any sharp kinks and must not come into contact with the hot backplates of the graphics cards.
Power Supply: Caution! Power to the SFF-8639 connector on the cable that plugs into the U.2 drive itself must be supplied via a separate, thick wire directly from the breakout board of the master power supply. This ensures a single ground path with the motherboard. Remember the strict rule from Step 4: no power draw from the motherboard and no thin adapters from SATA cables!
Installation in the Case: The U.2 NVMe drive itself has a solid metal housing that acts as a heat sink, as it is designed for through-air cooling. Position the drive strictly within the direct line of sight of the chassis’ front fans to prevent the drive from throttling during peak read/write loads. Secure it firmly to the mounting frame of the 4U chassis. Vibration from powerful server fans is the main enemy of the longevity of contacts and connectors; any play must be completely eliminated here.
Choosing a U.2 NVMe Storage Capacity: Entry-Level vs. Top-Tier
In our configuration, we use a highly reliable industrial-grade U.2 NVMe drive connected via a dedicated expansion slot. Using a single high-capacity server drive instead of multiple consumer-grade drives ensures that your server will never go offline due to overheating or memory degradation under constant load.
We offer two configuration options to choose from, depending on your initial budget and long-term goals with DePIN.
Option 1: Basic Standard (Recommended for ROI)
Disk capacity: 1.92 TB (listed as 2 TB in the specifications).
Estimated cost: approximately $250.
Who it’s for: Ideal for the “Quick Start” strategy and configurations aimed at achieving the fastest possible return on investment (7–8 months). This capacity is sufficient for running basic cloud containers on general-purpose networks (Akash/Vast) and for most standard tasks.
Option 2: Top-Rated (Maximum Capacity for Long-Term Use)
If you have the financial flexibility to expand your budget at the start, we recommend installing a 3.84 TB drive right away (listed as 4 TB in the specifications).
Estimated cost: approximately $600–$750.
What this means in practice:
Higher Network Rating (Multiplier): When distributing client orders, decentralized networks evaluate servers comprehensively. The more free enterprise-grade disk space you can make available to the network “with a single click,” the higher your server’s internal technical score will be. Consequently, the network will prioritize offering you more lucrative and high-value rental contracts first.
Perfect for “Quick Start”: When running without graphics cards (using only CPUs and memory), a server with a 4 TB drive automatically ranks at the top of search results for customers who need to deploy resource-intensive databases. Your revenue will approach the upper limit ($600/month).
A Head Start on AI and Rendering: Heavy neural network models can be hundreds of gigabytes in size. When you upgrade your graphics cards and move up to the big leagues, a 3.84 TB drive will save you from having to rebuild your system. You’ll be able to run several AI projects simultaneously on the server.
⚠️ Technical Note: There are 7.68 TB drives available, but their current price is exorbitant (starting at $1,200 and up), which undermines the entire business model for return on investment. We also do not recommend using two separate 2 TB drives—it’s best to use only one U.2 slot on the motherboard so as not to take away PCIe lanes from future graphics cards.
Summary: If you’re on a tight budget, go ahead and choose the 1.92 TB base drive—the business model works great on it. If you want to immediately top the network rankings and have the highest priority when tasks are assigned, go with the 3.84 TB model. The choice is always yours.

*Note: All software-related tasks—from preparing the bootable USB drive, installing Ubuntu Server, correctly partitioning the disks manually, and mounting the /data directory, to registering unique UUIDs in the system’s fstab file and configuring Docker—has been fully moved to a dedicated software module. At this stage, our hardware is fully assembled, securely mounted, connected, and ready for its first power-on.