#memoryexpansion

Presenting the Micron CZ122 and the memory expansion portfolio certified by Red Hat

Micron CZ122

Micron is announcing the release of certification samples for the Micron CZ122, its newest memory expansion module based on Compute Express Link (CXL) 2.0. Along with this noteworthy achievement, Micron is also announcing that the Red Hat Enterprise Linux version 9.3 operating system has certified the CZ122 and its predecessor, the CZ120. These achievements highlight Micron’s dominance in providing enterprise data centers and hyperscalers with high-quality, high-performance solutions.

With a maximum memory capacity of 256GB and a memory bandwidth of 37GB/s, the Micron CZ122 module delivers notable improvements in hardware-based heterogeneous interleaving, security features, and reliability, availability, and serviceability (RAS). The CZ122 and CZ120 are certified with Red Hat Enterprise Linux and have undergone extensive testing with AMD and Intel CPU platforms as well as with server OEMs. A significant milestone in the CXL ecosystem’s evolution, this certification opens the door for production-ready solutions that support new composable data center architectures and speed up workloads of the future.

Red Hat Device Certification’s main advantages for Micron CZ120 and Micron CZ122

Smooth integration: RHEL 9.3 OS compatibility has been certified.

Performance optimization: The most recent RHEL 9.3 OS Linux kernel has improved memory-tiering functionality.

Multi-tenancy: Support for memory allocation and management across virtual machines for KVM and QEMU applications.

Announcing the Micron CZ122 memory expansion module

Micron is sampling its second memory expansion module, the Micron CZ122, to set the stage for manufacturing. The CZ122 is a development of the CZ120 that adds more RAS, stronger security features, increased manageability, and hardware-based heterogeneous interleaving for improved system-level performance.

Key characteristics of the memory expansion module Micron CZ122

With CXL, hardware-based heterogeneous interleaving offers users a more innovative way to attain the best system-level performance.

Data interleaving between native memory and CXL memory is made possible via BIOS configurability.

Applications don’t need to be modified for deployment.

Improved security, RAS, and management features.

Its Technology Enablement Program (TEP) offers white glove assistance for integrating CXL-based technologies into your data center infrastructure.

Interesting applications for CXL

Micron is committed to creating compelling and widespread use cases that result in significant total cost of ownership (TCO) benefits, going beyond ecosystem collaboration for product adoption. Three compelling use cases for CXL have been presented by Micron and partners at the Future of Memory and Storage (FMS) conference:

LlamaIndex framework: A retrieval augmented generation (RAG) workload can be sped up by using CXL memory.

Neural Speed framework: An Intel CPU-based inferencing job can be accelerated by using CXL memory.

RocksDB application: Memory pooling and sharing behind the CXL 2.0 switch can be achieved with the H3 Platform memory chassis and Micron’s suggested Famfs contribution to the Linux open source kernel. Multiple servers or virtual machines running the RocksDB program can access the CXL memory as a shared file system thanks to Micron’s Famfs innovations.

We have learned a lot about the advantages of CXL in certain workloads from its collaborations. Along with offering a route to production deployments, it has concentrated on supplying these use cases as references with enterprise-ready hardware and/or software solutions.

Some of Micron’s primary areas of concentration moving forward are as follows:

Modular RAG: Keep improving the RAG data pipeline in order to create a modular RAG system that needs additional storage and memory.

CXL memory in storage networks: An adaptable and scalable way to increase the average latency and throughput of SAN systems by adding sizable memory pools that serve as read and write cache buffers.

CXL memory pooling and sharing: Evaluation of the performance, TCO, and implementation advantages of pooling and sharing large memory capacity, either directly connected to or behind a CXL switch.

A significant turning point in the development of CXL-based memory extension technology has been reached with the release of the Micron CZ122 module. Micron continues to lead the memory expansion market with its cutting-edge features, smooth integration, and reliable performance, as well as the CZ120 in mass production and customer delivery.