Samsung is making server memory expansion simpler, more flexible, and more efficient with its CXL Memory Module-Box (CMM-B) solution. Designed to occupy dedicated slots in a storage rack, CMM-B integrates seamlessly into Supermicro plug-and-play rack-scale platforms, enabling faster productivity gains while reducing total cost of ownership (TCO).
As the industry’s first rack-scale composable memory pooling solution, Samsung CMM-B addresses memory underutilization by allowing capacity and bandwidth to be dynamically shared across multiple servers. Developed through close collaboration between Samsung and Supermicro, this architecture represents a major step forward in disaggregated and composable infrastructure.
What Is CMM-B? #
CMM-B is a rack-level memory pooling device built on Compute Express Link (CXL) technology. It enables flexible memory resource allocation by supporting up to 24 E3 connections, with Samsung CXL Memory Modules – DRAM (CMM-D) installed inside the rack device.
This architecture delivers several key capabilities:
-
Disaggregated Memory Allocation
Memory capacity is decoupled from compute and exposed as a shared pool, allowing multiple servers to access memory over CXL. This eliminates memory silos and improves overall utilization across the rack. -
Composable Memory Orchestration
Memory resources can be dynamically combined and assigned to different hosts based on workload demand. -
Scalable Memory Expansion
Multiple memory modules can be aggregated into larger pools, delivering on-demand capacity and bandwidth where it is needed most.
CMM-B is managed through Samsung Cognos Management Console (SCMC), which provides an intuitive interface for rapid rack-scale deployment and streamlined operations.
Rack-Scale Memory Architecture #
A complete CMM-B rack-scale solution includes:
- Top-of-Rack (ToR) switch
- One or more CMM-B devices
- Application servers
- Samsung Cognos Management Console (SCMC)
By consolidating hardware and management software within a single rack, this architecture enables fast, intuitive memory provisioning while laying the foundation for fully composable infrastructure.
Rack-Scale Management and Orchestration #
CMM-B becomes truly composable through integration with rack-scale management platforms such as Supermicro SuperCloud Composer. SCMC coordinates memory allocation across hosts, enabling centralized control and automation.
This integration allows operators to treat memory as a shared, software-defined resource, rather than a fixed component tied to individual servers.
Advanced Memory Pooling with CXL #
CMM-B is specifically designed for CXL-based compute environments and supports both CXL 1.1 and CXL 2.0 specifications. These protocols allow memory devices to attach coherently to hosts and be centrally managed via REST APIs when connected to the SCMC platform.
This standards-based approach ensures compatibility with emerging CXL-enabled CPUs and platforms.
Scalable Device Configuration #
CMM-B supports highly scalable configurations:
- Up to 24 TB of memory capacity using CMM-D modules
- Connectivity for up to three host systems
- Flexible scaling of both capacity and bandwidth
This makes CMM-B suitable for memory-intensive workloads such as AI training, in-memory databases, and large-scale analytics.
Improved Power Efficiency #
An integrated SoC switching chip manages CXL traffic efficiently under both CXL 1.1 and CXL 2.0. By minimizing power consumption at the interconnect level, CMM-B helps reduce overall system power draw and lowers TCO—an increasingly critical factor in modern data centers.
Dynamic Memory Allocation with SCMC #
One of CMM-B’s most powerful features is dynamic memory allocation via the Samsung Cognos Management Console:
- Memory can be assigned independently of the physical server
- SCMC agents integrate seamlessly with application modules
- Memory is allocated precisely where system or application demand requires it
With a consistent GUI and REST API support, SCMC simplifies memory configuration while enabling orchestration across heterogeneous memory devices.
Memory Allocation Analytics #
SCMC agents run on application modules within the rack-scale environment and provide a visual dashboard that displays:
- CMM devices
- Switches
- Hosts and their memory usage
These analytics make it easy to identify imbalances, optimize memory distribution, and determine when reconfiguration is required.
Software Designed for the AI Era #
CMM-B’s software stack enables dynamic discovery and utilization of CXL memory across rack-scale deployments. As AI workloads increasingly push data processing closer to memory, Samsung’s CXL-based solutions offer a practical path to scaling, managing, and fully exploiting memory resources in modern data centers.
By combining CXL hardware innovation with intelligent management software, Samsung’s CMM-B solution sets a new benchmark for rack-scale, composable memory infrastructure.