For the technology industry interesting times can bring great opportunity. We have recently seen a pretty major, if evolutionary, shift in architectures over the past decade and particularly in the past year, with the release by both Intel® and AMD® of their latest generation of CPU offerings. It is obvious to me that these changes are an indication by both CPU vendors that while CPU power is important – each has made improvements in compute power – it is becoming more important than ever to ensure that these higher-power CPUs are better able to interact with data and the outside world. Each vendor has made significant improvements in their PCIe and memory capacity as illustrated in the two charts below.
Looking at this jump in capability (especially from AMD), I begin to think about what this means for how we design and deploy solutions that could be dramatically improved by these new capabilities. Here are four examples:
- Software Defined Storage solutions that can take advantage of more memory for buffering, storage class memory (SCM) for caching, and PCIe-native NVMe SSDs resulting in extremely high performance data access.
In fact, we examined the new AMD EPYC® archictecture and looked at how the dramatic increase in the number of PCIe lanes and memory capacity will affect how you design and deploy storage-centric solutions based on maximizing one of several primary performance metrics: Capacity, Throughput and IOPS. Each of these criteria dictate different design strategies and put different demands on the design of a server platform that is suitable for each. You can read the details of our analysis in a new Micron white paper.
- Artificial Intelligence and Machine Learning solutions that depend on high-performance graphics processing units (GPUs) to perform massive calculations on larger node-local data sets enabled by more and larger NVMe SSDs. Many of these solutions are dependent on advanced edge processing capabilities that are data intensive.
- Cloud Data Center solutions that need to host data-centric applications such as SQL Server with the flexibility of hybrid hyper-converged infrastructures that balance compute, high-performance storage and high-bandwidth networking to potentially reduce the number of nodes to do the same amount of work while consuming less power, rack space and fewer software licenses.
- Container infrastructures that are dependent on rapid deployment and scale to meet the growing “on-demand services” requirement from customers. SSDs can allow you to do more with less – devices, power & cooling, and latency.
There are many more solutions in which these new architectures will affect our design strategies. The flexibility to deploy more storage, more networking bandwidth, and more processing power in scale than we have never had access to before will allow our customers to create some really capable solutions that break the traditional cost curve for IT.
We truly do “live in interesting times” and I personally am excited about what these new architectures can enable. Micron’s goal is to provide you with the best possible memory and storage technology for your solutions by analyzing and understanding how these new architectures – and those still to come – will impact your solutions. As we continue to use and understand these new server architectures, we will be releasing more insights into the benefits that each may provide for your solutions and how best to design these solutions to meet your business requirements.
In the coming weeks we will dive a bit deeper into the topics mentioned above. In the meantime, may you continue to “live in interesting times” (for you history buffs, the prior link provides the interesting background on this old quote).