DF-S04 · SolutionIn development
HYBRID MEMORY · GEN4

Hybrid Memory Expansion.

Extend a compute node's effective memory + storage capacity using PCIe-attached memory devices and NVMe capacity behind one fabric.

Request Rack QuoteRequest Datasheet All Solutions →
Ships inside DF-R1
OVERVIEW

Extend memory & storage without touching the CPU socket.

PCIe-attached memory devices combined with NVMe capacity, behind a single Gen4 fabric — a tiered memory-storage layer that scales per-node capacity beyond DRAM limits.

For workloads that need more than the DIMM slots on the motherboard can offer, a PCIe-attached memory tier plus NVMe capacity acts as an addressable, high-bandwidth secondary memory layer.

Built on PCIe Gen4 Fabric Backplane Shipping today
Hybrid Memory Expansion.
ARCHITECTURE

Solution block diagram.

The block diagram illustrates how Hybrid Memory Expansion extends host memory beyond socket-DIMM limits. The host CPU accesses a first tier of DRAM directly. A second tier of PCIe-attached memory (CXL-ready) is exposed across the Gen4 fabric and appears to the host as an additional NUMA node. A tiering engine promotes/demotes pages based on access temperature. Workloads such as in-memory analytics, large-model inference, and graph databases keep their hot working set in fast memory without buying larger, disproportionately expensive DIMMs.

CPU + DRAMTier 0 Gen4 FabricBackplane PCIe MemTier 1 NVMeTier 2
DEPLOYMENT TOPOLOGY

Connection diagram.

The connection diagram shows the DF-S04 wired into a memory-hungry server. A PCIe host adapter in the server cable-attaches to the DF-S04 chassis. The chassis hosts memory modules across the Gen4 fabric backplane. A dedicated 10 GbE + IPMI/BMC management port carries telemetry to the Datafabrix DCIM. Standard C13/C14 PDU inputs with 1+1 redundant PSUs. Cooling stays inside a standard rack thermal envelope — no liquid cooling required. Use this diagram to plan HBA installation, cable length, and NUMA-topology visibility for the OS.

Compute Node Compute Node PCIe Gen4 Fabric Backplane PCIe Gen4 · Tiered memory/storage PCIe Mem Card NVMe Bay 1 NVMe Bay 2 NVMe Bay 3
INTERFACE DETAILS

Interfaces & connectivity.

InterfaceDetails
HostPCIe Gen4 x16
Memory tierPCIe-attached memory expansion cards
Storage tierNVMe U.2/U.3 Gen4
Access modelSoftware-mapped tiered memory
ManagementRedfish · DCIM-native
HOW IT WORKS

Detailed description.

Modern AI and analytics workloads push against DIMM-slot limits. Adding more DRAM per node is either impossible (socket limit) or economically painful (DIMM prices scale non-linearly with density).

A hybrid memory expansion platform sits on the compute node's PCIe fabric and offers two capacity tiers: PCIe-attached memory (fast, byte-addressable, DRAM-class) and NVMe (slower, larger, block-addressable). Software runtimes tier data between them.

The Datafabrix Gen4 Fabric Backplane provides the fabric connectivity between the compute node and both tiers, with per-slot hot-plug and full Redfish visibility for the DCIM platform.

Applications

  • In-memory analytics — Extend RAM-hungry analytics beyond socket-DIMM limits.
  • Large-model inference — Page LLM weights between DRAM and PCIe-attached memory.
  • Graph databases — Keep hot graph fragments in fast memory, cold in NVMe.
  • Vector databases — Tier billion-vector indexes between memory tiers.
USE CASES

Where this solution shines.

LLM serving

Serve larger models per node by paging weights across tiers.

Real-time analytics

Keep working sets larger than DRAM at near-memory latency.

AI feature caches

Massive feature stores accessible at PCIe-adjacent speeds.

SPEC HIGHLIGHTS

Technical specifications.

CategorySpecification
FabricPCIe Gen4 non-blocking
Memory tierPCIe-attached memory cards
Storage tierNVMe Gen4 U.2/U.3
Access modelSoftware-tiered
ManagementDCIM-native
TARGET INDUSTRIES

Who deploys this solution.

AI/MLAnalyticsDatabasesCloud
HOW TO DEPLOY · DF-S04

Deploying Hybrid Memory Expansion in your infrastructure.

Where it fits in the rack, how it connects, and what you get after installation.

Hybrid Memory Expansion — memory-tier prototype.
Deployment illustration · Hybrid Memory · DF-S04
42U Rack — customer datacenter Existing Host Servers / Compute DF-S04 · Hybrid Memory Expansion PCIe Gen4 Fabric · Hot-plug ◀── This solution ──▶ Existing Storage Existing Network Ethernet Mgmt · Datafabrix DCIM · Fleet Control
RACK-LEVEL ARCHITECTURE

Where DF-S04 sits in your rack.

PCIe / CXL-based memory tiering platform that pools DRAM/PMEM/NVMe behind the Gen4 fabric. The solution slots into a standard 19-inch rack alongside your existing servers, storage, and network fabric — no re-architecture required.

Target deployments: In-memory workloads whose data set exceeds a single host's DRAM capacity.

  • Cabling: standard PCIe host adapter → PCIe cable → DF-S04 chassis
  • Power: standard C13/C14 rack PDU, 1+1 redundant PSU option
  • Management: 10 GbE + IPMI/BMC out-of-band + Datafabrix DCIM integration
  • Cooling: fits standard rack thermal envelope, no liquid cooling required
01

Install in rack

Slide the DF-S04 chassis into a standard 19-inch rack slot. Rail kit included. Cable to your PDU and management network.

02

Connect PCIe host

Install the Datafabrix PCIe host adapter in your existing server. Connect via PCIe cable to the DF-S04 backplane uplink.

03

Populate the fabric

Insert NVMe drives, GPUs, or accelerators into hot-plug slots. Each device trains and appears as a native PCIe endpoint on the host.

04

Manage & scale

Optional: connect to Datafabrix DCIM for fleet-wide visibility. Add more chassis for horizontal scale — no host reconfiguration.

🎯

In-memory databases

Direct rack integration alongside your existing production servers. Turnkey in a single deployment window.

📈

AI model residency

Scale from a single chassis pilot to full-rack production with zero architectural change.

🔧

Analytics warehouses

OEM-ready configuration for organizations building their own branded infrastructure products.

Ready to design a rack around it?

Engineering pilots, reference designs, and OEM co-design programs — we work with your team from concept to production.

Request Rack Quote All Solutions →