We first previewed the news of the Dell EMC PowerEdge MX launch at Dell EMC World 2018. Today we have the first concrete details on the PowerEdge MX as part of the formal product announcement. The PowerEdge MX is the company’s next-generation shared chassis (blade) server that is built for future technologies.
Dell EMC PowerEdge MX Addressing Trends
For decades, one of the biggest challenges in the industry has been server utilization. Prior to virtualization, server utilization was extremely low. After virtualization, utilization has gone up, but the industry realizes that there is more to go. Companies like Facebook, Google, Alibaba and others who are operating vast infrastructure know that stranded capacity is an enormous area for cost savings and performance gains. This is the trend that Dell EMC PowerEdge MX aims to address for the enterprise.
The Dell EMC PowerEdge MX is designed to be software defined, use today’s latest technologies, and be flexible enough to utilize future technologies as they become available.
Dell EMC is calling this its “kinetic infrastructure.” The company’s vision is that the PowerEdge MX building block becomes the easiest and most flexible piece of hardware in your data center. Using new technologies, it will aim to minimize the stranded capacities that we see today.
Most large enterprise vendors have some form of this messaging. Dell EMC has one of, if not the most complete articulated vision.
Into the Dell EMC PowerEdge MX
Digging deeper into the PowerEdge MX means that there are a number of technologies used to make a system like this work. Compared to traditional rack servers, this type of infrastructure offers a more tightly integrated design. It also is more modular making upgrades and servicing faster in the data center. As part of Dell EMC’s forward-looking design, the PowerEdge MX is designed for the power and cooling required for next-generation, higher-power accelerators, CPUs, interconnects, and memories.
The compute nodes are both two and four socket nodes. We are told that there are options for not just CPU compute, but also FPGA, GPU compute, and future accelerators for applications like deep learning and AI inferencing. In terms of memory, Dell EMC is planning to support more than just traditional DDR4. That includes storage class memory that is on the near-term horizon. One can install SAS storage sleds in the 7U chassis for direct attach storage.
The interconnect is fascinating. The PowerEdge MX starts assuming 25GbE as base. In the computer industry as a whole, 25Gbps channels are popping up everywhere from networking to internal paths between components. The Dell EMC PowerEdge MX does not have a standard midplane so it is designed for 32Gbps Fibre Channel and 25GbE today and moving up to 100GbE in the future. A key industry challenge is that as signaling moves from low-speed 10Gbps to 100/400Gbps networking and PCIe Gen3 to PCIe Gen4 next year and Gen5 not far thereafter, PCB will become obsolete as it is unable to handle high-speed signaling. Ths is the future Dell EMC is designing for.
Although Dell EMC is focused on today’s fabrics, we have a feeling that the lack of a midplane is a far more strategic choice. The industry knows that there are limits to current DDR4 and future DDR5 signaling. There are limits to PCIe and traditional socket-to-socket interconnects. If you want a chassis to have a long service life, it means being able to accommodate future technologies like Gen-Z. Dell EMC did not state that the PowerEdge MX will support Gen-Z. On the other hand, that is a potential next-generation fabric and not having a midplane would make implementing a future fabric like Gen-Z possible.
Underpinning the system is management. The Dell EMC Open Manage Enterprise Modular Edition is the management solution that makes the Dell EMC PowerEdge MX work. Each system can be managed using a familiar management solution that is customized for the modular architecture that the Dell EMC PowerEdge MX offers. Dell EMC has functionality white box offerings do not have such as a cleaner interface and the ability to manage networking seamlessly with the nodes.