ARM Holdings has been angling for the data center for a number of years but has yet to make significant inroads against Intel’s dominance. But with enterprises of all stripes looking to foster greater diversity across their distributed data footprints, time is growing short for both ARM and its legion of chip manufacturers to show it can deliver what the market demands.
One of the key drivers for ARM’s particular style of processing is the Internet of Things (IoT), which will drive the need for distributed processing architectures to new levels. But as the Next Platform’s Jeffrey Burt points out, it takes more than user demand to pull ARMs into the data center; it requires broad software support as well. So far, however, developers like Red Hat, which leads the commercial Linux market, have yet to extend the level of support to ARM that it gives to other platforms like Xeon, Power and System z. Red Hat has been vocal in its support of ARM, going so far as to back the ARM Server Base System Architecture (SBSA) and to incorporate the architecture in projects like the Ceph object storage platform. But without a full RHEL stack supported on ARM, enterprises will continue to face more hurdles running Linux workloads on ARM than on Intel, AMD or IBM solutions.https://o1.qnsr.com/log/p.gif?;n=203;c=204663295;s=11915;x=7936;f=201904081034270;u=j;z=TIMESTAMP;a=20410779;e=i
Interestingly, it seems that Microsoft is becoming more serious about shifting its cloud infrastructure toward ARM. At the Open Compute Summit last March, the company announced it was working with Cavium and Qualcomm on ARM server designs with an eye toward converting more than half of its cloud center infrastructure before long. As Datacenter Knowledge’s Yevgeniy Sverdlik noted, this is more significant than the typical murmurings from hyperscale providers that they are looking at ARM along with a number of other solutions in the quest for low-cost, high-scale infrastructure. With Microsoft pushing a hybrid cloud vision to the enterprise, its support for ARMs in the cloud may translate to greater deployment in the data center as well.
The biggest knock against ARM is that it is designed to provide low-power support for high-speed, low-packet workloads – the kind that populate mobile environments. In the data center, however, applications often require massive, coordinated processing for large loads, which the ARM can provide but only through additional software overhead to implement large-scale parallel processing. Recently, however, a company called Cadence Design Systems released its Cache Coherent Interconnect for Accelerators (CCIX) that provides an open chip-to-chip standard for data-intensive processing. Working with ARM’s CoreLink mesh solution, the company is aiming to boost multichip connectivity to support advanced functions like machine learning and data analytics.
Probably the best thing ARM has going for it is its low-power design, which the company expects to see in greater demand as the enterprise transitions to software-defined infrastructure. As Director of Market Development Bill Neifert told Semi Engineering recently, developers are increasingly writing code to take advantage of application-specific processors that provide just enough power to get the job done. This will naturally lead to higher demand for smaller processors to take on specified tasks, leading to greater diversity in silicon-level solutions within the enterprise data center and on the cloud. So in the end, ARM might gain more of the enterprise workloads almost by default.
But entrenched infrastructure is difficult to replace, and after nearly three decades of primarily Intel-best data infrastructure, it will require a convincing argument to get organizations to give up what they know for an entirely new chip architecture.
Fortunately for ARM, this is a time of momentous change for the enterprise, and the old ways of doing things are no longer delivering optimal results.
Arthur Cole writes about infrastructure for IT Business Edge. Cole has been covering the high-tech media and computing industries for more than 20 years, having served as editor of TV Technology, Video Technology News, Internet News and Multimedia Weekly. His contributions have appeared in Communications Today and Enterprise Networking Planet and as web content for numerous high-tech clients like TwinStrata and Carpathia. Follow Art on Twitter @acole602.