As companies pursue greater efficiency in their data centers, particularly with regard to lighter workloads such as those resulting from the burgeoning mobile market, the microserver is stealing the spotlight from the traditional server form factor. Microservers are set for explosive growth in coming years, so here’s a brief look at what you can expect.
Microserver: Sorta Like a Blade Server
Microservers are a natural outworking of blade-server technology. Blade servers are server chassis containing multiple electronic boards (“blades”), each of which incorporates the functions of a typical server (excepting things like mass storage). The chassis may contain local storage and basic infrastructure like cooling. Blade servers enable much higher densities than traditional servers, and they can cut cabling requirements in data centers.
Given the tremendous growth in mobile computing, which produces numerous light workloads (as opposed to the heavier workloads of content development, research and high-performance computing), data centers (particularly cloud facilities) are seeking efficient means of handling the demand. Because mobile workloads do not require the heavy lifting that is the province of powerful servers (often running Intel Xeon processors), the benefits of blade servers have manifested in a related form factor: the microserver. A microserver emphasizes power efficiency for workloads with many smaller jobs that can be processed in parallel. Because the horsepower of, say, a Xeon chip isn’t required, these devices can employ more-efficient chips, such as Intel’s Atom or variations of ARM’s proprietary architecture.
By one definition, a microserver is “any server with a large number of nodes, usually with a single socket or multiple low-power processors and shared infrastructure.” According to ServerWatch, microservers generally “have two or four slots for main memory, two high-speed Ethernet ports, and Sata ports to connect up to four disk drives. Unlike a blade server, it usually does not have a management processor.” Like a blade server, microservers aim for higher densities, but they are balanced by lower-power processors.
IHS iSuppli is forecasting global microserver shipments to triple in 2013 year over year, reaching almost 300,000 units. Furthermore, the research firm expects tremendous growth to continue, eventually exceeding 1.2 million units in 2016, for a compound annual growth rate (CAGR) of roughly 230% between 2011 and 2016. GigaOm notes that in 2016, microservers will constitute about 10% of all server shipments.
Microservers are certainly not suited for every application. In some sense, microservers are to mobile devices what traditional servers are to PCs. As much as pundits might declare the PC to be dead, mobile devices still lack the content-creation capabilities and other beneficial features of PCs. Thus, PCs will remain an important market, even if that market does shrink (since many users get everything they need from a mobile device). In the same vein, microservers offer many benefits, but other server form factors will still carry a major share of the data center. In the case of heavy workloads, a single powerful server can be more efficient (measured in performance per watt) than an array of lower-power servers.
Therefore, the microserver, like the mobile device, may simply be capturing the portion of the server market that is best suited to its capabilities. The data center of the future probably won’t be just an array of microservers, though—just like the end user of the future won’t be limited to just a tablet and a smartphone.
Competitors for Microservers
The x86 architecture is dominant in data centers, so Intel and AMD are obvious participants in the microserver market. AMD will, as in other markets, trail Intel, but it is branching out from x86 by adopting other architectures, particularly for the data center market. Although Intel certainly has the resources to follow a similar approach, it stands to lose much more face by doing so. The company is the owner of the x86 intellectual property and has only reluctantly granted other companies permission to use that architecture (AMD being a case in point, although Via Technologies is another even smaller example). Adopting an alternative architecture would be tantamount to admitting that x86 doesn’t cut it for all applications. Such an admission would be far from unexpected, since no technology can serve every market well, but it would put Intel even more on defense—in marketing, if nothing else.
And anyone who follows technology news probably knows the most touted competing architecture for servers in the data center: ARM. The king of mobile, the ARM architecture is seeking to make headway in servers by bringing low-power operation and improved efficiency to companies desperate for ways to cut their energy costs. The microserver is one category in which ARM will compete.
Even though ARM has gained extensive marketing buzz, however, moving into the server market is not simply a matter of rebranding processors from mobile phones, sticking them in a server box and selling them to data center operators. ARM-based server chips must increase performance, expand memory capacity and offer additional features compared with mobile processors—and that entails greater power consumption. ARM and Intel are therefore both moving toward the middle in pursuit of microservers. ARM, the king of small and light, must offer beefier performance, whereas Intel, the king of heavy lifting, must offer better efficiency for lighter workloads.
The ARM architecture is being peddled by a variety of vendors, including AppliedMicro, Calxeda (formerly Smooth-Stone) and Marvell, for example. Whether ARM-based processors can deliver on their promise remains to be seen, but Intel at least is maneuvering to counter the threat to its dominance in the data center. Eventually, the buzz around ARM taking on server giant Intel will die down, leaving just the matter of which of the two “extremists” (light and mobile versus heavy and powerful) will best serve the market in the middle.
The microserver market is on a path for tremendous growth over the next few years, taking a significant (but not overwhelming) bite out of the share of other form factors. One of the critical battles will be between the light, fit contender, ARM, and the heavyweight incumbent, x86. By adopting ARM, AMD may have a unique opportunity in this market to find a foothold against long-time (dominant) competitor Intel while holding off other vendors. ARM has yet to prove itself in the data center, however; despite press buzz and marketing hype that paints the architecture as the savior of the data center, it has a long road to follow before achieving a solid and lasting position. ARM may end up in a technological draw with Intel, in which case the incumbent’s greater and more-established ecosystem for the data center may carry the day, simply by virtue of momentum.
Photo courtesy of aleutia