The data center on the cusp of change

Micron Technology | January 2022

In recent years, the exponential rise in the volume and types of data and new services has driven a significant investment in data centers and related cloud infrastructure. Additionally, an evolution in Artificial Intelligence (AI) workloads is now central to unlocking value and insights from this valuable data. Consequently, organizations are increasingly focused on building the infrastructure they need to meet these demands—whether on-prem, at the intelligent edge or in the cloud—to gain greater efficiency and scale. These conditions create a unique opportunity for cloud service providers.

A transformational challenge

For many providers, scaling resources to adapt to the evolving nature of workloads has created a tricky challenge: how do they support the known demands of today while planning for the increases in scale, efficiency and performance requirements of tomorrow?

Addressing this challenge has been made more difficult by the nature of “typical” enterprise-class data centers. Many of these have been built over a long period of time and represent a spectrum of technical architectures, from monolithic legacy systems to more open and modern virtualized environments and increasingly cloud-native containerized infrastructure.

The need for flexible infrastructure

Efficient scaling poses one solution to these increasing demands. As workloads evolve to include more complex tasks that leverage AI capabilities such as image recognition at the intelligent edge, machine learning in high performance clusters, natural language processing in the cloud, or video encoding/decoding on content delivery networks, the traditional balance of compute to memory or storage access and connectivity (I/O) must shift. This means that some portions of existing data center infrastructure may unfortunately not be well suited for these new workloads.

The imperative for new infrastructure investment today is flexibility. Architecting a data center for efficiency is also key. The ability to effectively serve applications and leverage data from widely disparate sources with different performance, scale and latency needs is more important than ever before.

Architecting transformations without performance bottlenecks

Historic infrastructure innovation has relied on a maniacal focus on more compute, more cores and, consequently, more cost. Today’s opportunity feels much different as constraints in system architecture have limited the effectiveness of simply throwing more compute resources at the challenge. This time, the industry response will need to focus on a balanced approach where data and services are provided quickly and efficiently, with minimal performance bottlenecks between compute, memory/storage and I/O resources, maximizing valuable resources and pooling these resources where possible.

Tackling evolving workloads

The “compute first” hierarchy of workload characterization in traditional data centers is evolving. A growing volume and variety of these resources (such as CPU, GPU, IPU, DPU and accelerators) are emerging in the market. Clearly, the heterogeneity of compute and evolving workloads means that innovative memory, storage, and interconnect technologies will be as important, if not more so, than traditional compute when it comes to rolling out an efficient data center infrastructure in the future.

A new memory and storage hierarchy in the data center is key to future innovations, such as Compute Express Link (CXL), and are central to maximizing the value and resource utilization in future data centers. CXL is a high performance, low latency, memory-centric protocol that can be used to communicate between devices in a system. It’s an industry-wide effort to unlock the opportunity of composability.

Take pooling of resources as an example. Pooling enables memory resources to different hosts depending on the workload. Memory can be allocated and deallocated to different hosts. This minimizes overprovisioning in the data centers while improving memory performance.

Composable data center architecture is the key

This is where Micron steps in. We are creating innovative memory and storage technologies that address these challenges and are delivering them through broad ecosystem partnerships and standards leadership, bringing over 40 years of innovation in memory and storage to support the industry in a new era in the data center.

In fact, Micron’s technology innovations of today are unlocking the opportunity for tomorrow.

For the first time in our history, Micron has delivered industry-leading process node leadership across memory and storage. The introduction of 176L NAND and 1 alpha DRAM represent major technology breakthroughs for our company and reflect our focused R&D investment.

Earlier this year, Micron introduced its industry leading 1 alpha memory node, the world’s most advanced memory node. This advancement has been realized across our standard compute DRAM and low power DRAM product lines, delivering significantly higher memory density versus our previous generation.

We have also developed the world’s first 176-layer NAND which leads to better power efficiency and write times, and higher data transfer rates.

Micron is also leading the industry in the CXL-attached memory era by pioneering innovations to enable low-power DRAM in the data center, but success requires strong industry collaboration. CXL-attached memory needs to be co-optimized with leading compute platforms, so we are working with our ecosystem partners and through industry consortiums to do this and more.

The industry is collaborating to address challenges and deliver the data center of the future: a composable data center architecture that will give organizations the ability and flexibility to dynamically allocate resources to workloads as needed. This composable future will provide greater efficiency of infrastructure, allow workloads to scale up and out as needed, and serve up and store data faster and more efficiently than is possible today.

Future data centers will unlock new possibilities

Though these challenges seem daunting, we are well on our way to a composable future. This will enable us to not only be faster and more efficient with today’s applications and processes, it will create compute environments where resources are maximized in an efficient manner and companies can gain every benefit from the infrastructure investments that they make.

Most importantly, the future data center will help data scientists, innovators and researchers solve some of our biggest global challenges—whether it’s cutting-edge work being done in science around climate change, genome research to cure cancers, or saving coral reefs. The future data center will also be key to unlocking smarter cities, factories, and efficient living for all. Delivery of more services, more insights and more innovation will wholly rely on the future data center; It will be profound and life-altering.

Learn more from industry insiders

To learn more, watch the fireside chat: “The Future of the Data Center.” In this fireside, host Patrick Moorhead of Moor Insights and Strategy talks with Micron’s Raj Hazra and Jeremy Werner about the challenges and opportunities that the future data center will bring and how Micron is stepping up to lead.