Server form factors: A guide to rackmount, blade servers and more
A comprehensive collection of articles, videos and more, hand-picked by our editors
A new generation of microservers challenges the x86 norms, and not every application will be happy about it.
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
System on a chip (SoC) configurations are changing the compute capabilities of data center systems and end-user products. An SoC is an integrated device combining all or most of the components required to build an electronic system: processors, memory, peripheral controllers and onboard timing devices. SoCs also can include analog, digital signal processing and radio-frequency capabilities, such as 802.11n Wi-Fi for wireless networking.
SoCs in enterprise computing
Modern servers deliver an enormous amount of computing power to data center workloads, and virtualization optimizes resource usage. Renewed interest in reduced instruction set computer architectures such as Advanced RISC Machine (ARM) processors comes from enterprises rethinking the way they handle, scale and power workloads; this has paved the way for server-type SoC devices designed for relatively simple microserver tasks, such as Web serving.
Instead of provisioning dozens or even hundreds of Web servers from an expensive x86 virtualized server, enterprises can use a box full of inexpensive and efficient SoC microservers right from the start. Intel's 64-bit Atom Avoton chip supports virtualization, includes up to eight processor cores, and integrates Serial Advanced Technology Attachment 3 and Gigabit Ethernet, for example. These 1.6-GHz to 2-GHz SoC devices use 8.5 W of power, compared to an Intel Xeon processor that demands 85 W or more. Chip vendors such as Calexda and AMD license ARM's A57 or ARMv8 64-bit cores for microserver-targeted SoCs.
With such dense server boxes, compute power is more scalable than in traditional x86 server architectures. Rather than provision more CPU cycles from a pool of several x86 cores, an IT administrator provisions more Atom or ARM cores to tackle a particular workload or computing problem. However, this approach is new, and most operating systems and applications don't readily support this kind of scalability yet.
SoCs offer many advantages, but won't suit every workload. Web hosting, Hadoop and other Java-based applications have low processor utilization and small memory footprints; these are generally good choices for SoC deployment on microservers. But workloads with heavy processor utilization demands, large memory requirements and poor scalability across multiple processors are typically deployed on traditional x86 platforms.
An SoC can include one or more digital signaling processors for analog signal processing and a graphics processor. Memory blocks on the SoC include some combination of volatile and non-volatile memory, such as dynamic RAM and flash. PC-class SoCs also access external memory.
A full-featured I/O controller supports a suite of standard interfaces including Ethernet, USB, FireWire and so on. All these components are operated with a set of timing sources (clocks), counters, regulators, power management circuits and other ancillary devices, each interconnected with a proprietary bus or a recognized standard such as the Advanced Microcontroller Bus Architecture.
Sophisticated SoC devices designed for tablet and microserver systems will support desktop-class operating systems, such as Windows or Linux, and rely on external memory and peripheral devices.