High-density data centers are all about efficiency. IT managers look for technology solutions that deliver high levels of performance with minimal energy consumption and operational overhead. Disaggregated servers deliver on all three counts while dramatically reducing server refresh costs.
First, a little history. Servers used to be large, monolithic machines until the introduction of rackmount servers in the late 1990s. The “pizza box” design maximized the use of space within the data center compared to traditional “tower” services. Next came blade servers, which collapsed the entire server architecture onto a single motherboard. Multiple blades can be mounted within a single chassis that shares common resources, including power and network connectivity. The design further increased space and energy efficiency.
But all of these server designs have their own CPU, memory, I/O and other components, which are tightly interconnected. When it comes time to refresh a blade or rackmount server, you have to replace the whole machine, even if some of the parts still have many years of useful life remaining.
With disaggregated servers, CPU, memory, I/O, storage, power and cooling are broken up into modular subsystems that can be sized, scaled and refreshed independently. This makes it possible to design a system that precisely meets workload demands and upgrade individual components to take advantage of advances in technology.
Disaggregated servers enable significant capex savings by eliminating waste, and reduce the amount of staff time required per sever refresh. In addition, disaggregated servers improve power and cooling efficiency through their common, shared infrastructure, and the reduce environmental footprint associated with manufacturing and shipping servers and their containers.
Supermicro is an industry leader in disaggregated server solutions. In a large server deployment, Supermicro’s disaggregated MicroBlade systems have delivered 86 percent improvement in power and cooling efficiency and 45 percent to 65 percent capex savings per server refresh cycle.
One Fortune 50 company has deployed more than 50,000 MicroBlade servers with a power usage effectiveness (PUE) of 1.06. Compared to a traditional data center running at 1.49 PUE or more, this new data center achieves an 88 percent improvement in overall energy efficiency. When the build-out is complete at 35 megawatts of IT load, the company expects to see more $13 million per year in energy savings alone.
Supermicro recently extended its SuperBlade lineup with a new 6U system that supports 10 or 14 blades per chassis. With up to 98 dual- or single-socket blade servers per 42U rack, the 6U SuperBlade systems are perfect for a wide range of enterprise and cloud applications where density and compute performance are prime considerations.
The two-socket blades have 24 DIMM slots, and the one-socket blades have 12 DIMM slots. Each enclosure supports up to 28 U.2 NVMe or 42 SAS solid-state drives and up to four 25G or 10G Ethernet switches. The enclosures use the same Ethernet switches, chassis management modules, and software as other SuperBlade and MicroBlade systems for improved reliability, serviceability and affordability. Additional resource savings are achieved with up to 90 percent reduction in cabling.
Disaggregated servers are the darlings of hyperscale environments but can benefit any organization interested in data center efficiency. Contact Rahi Systems to learn about Supermicro’s MicroBlade and SuperBlade systems.