Microservers offer benefits such as low power and space consumption, and have a number of potential roles in the data center — generally ‘scale out’ workloads that can use large numbers of relatively lightweight server nodes. But they have their limitations, including moderate computing capabilities, the potential need for software rewrites to handle clustered configurations, a lack of standards, and ambiguity about workload capacity and management. They also face competition from virtualization and cloud services (although it should be noted that multiple physical microservers can have advantages over a collection of virtual servers running on a single full-size server).
Limitations notwithstanding, microservers have made significant inroads in certain fields as their advantages have solidified. Here are 10 key facts about microserver technology.
For the purposes of this article, a “microserver” isn’t the handsome little box pictured below, even though HP calls its entry-level (Celeron- or Pentium-powered) ProLiant systems “MicroServers.”
What we’re referring to is a considerably larger and more expensive beast — “any server with a large number of nodes, usually with a single socket or multiple low-power processors and shared infrastructure,” as Kevin Huiskes, director of cloud computing at Intel, puts it. So be aware that the definition of a microserver can vary.
The system shown in Figure B may look like a chassis loaded with blade servers, but those are actually microserver “sleds” (in Dell’s parlance). It’s important to note that these two types of servers are different — and that the microserver is trying to eat the blade server’s lunch. A blade server has more horsepower, but it’s larger, consuming more energy and requiring more cooling. A microserver, by contrast, is smaller and less resource-hungry but has less computing firepower. It might seem like the weaker candidate would be overlooked, but as companies become greener, the notion of higher density and reduced cost becomes increasingly attractive.
You might argue that blade (or full-size rack-mount) servers are still a better bet because they can handle more demanding workloads. However, microservers can work together to pool their limited individual resources via clustering or shared processing. This is why microservers are prospering in the data analytics space, where shared resources can play a key role. Microservers can also boost uptime thanks to increased redundancy.
YourPCUniverse says that microservers cost 63% less for the setup and use 85% less energycompared to traditional servers. This sounds appealing, but it’s important to keep in mind that companies have to find the right balance between microservers and traditional servers. It does no good to replace full-size servers with too few — or too many — microservers. Cost savings can be attractive, but performance issues can be much less so, not to mention the danger of finding that your cost savings merely represented cost shifting.
Before replacing a fleet of workhorses with shiny new microservers, get professional guidance from the vendor, IT staff, and/or outside consultants to ensure that you’re not just trading one set of problems for another, unable to leverage the benefits. Resist the urge to go all-in until every avenue is explored, including application, network, storage, database, and processor/memory usage.
Microservers may fit into relatively small spaces in the data center, but they’re not exactly “affordable.” For example, in 2011 the list price for a “base configuration” of SeaMicro’s Atom N570-based SM10000-64 microserver was $148,000.
Most microservers offer internal storage, either on the server sleds themselves or in separate storage slots on the shared-infrastructure chassis. HP’s Atom-based Proliant m300 Server Cartridge, for example, accepts one 2.5-inch hard drive (500GB or 1TB) or SSD (240GB). With the Moonshot 1500 chassis accommodating up to 45 cartridges, that gives a maximum of 45TB of internal storage per chassis. SeaMicro’s latest SM15000 chassis has eight storage slots, each of which can hold eight drives, for a total of 64. If more storage is required, SeaMicro’s separate Freedom Fabric Storage system accommodates up to 1,344 2.5-inch or 3.5-inch drives (HDD or SSD) for up to a massive five petabytes all told.
Global Industry Analysts says that “media storage is currently the leading application area” for microservers, which are providing access to photos, audio, and video files over the network or internet — a relatively simple yet significant process for consumers and businesses alike.
8: The use of microservers for cloud computing is expected grow (a lot)
RnRMarketResearch.com reports that cloud computing application among microservers “will grow at a phenomenal CAGR of 62.3% from 2013 to 2018.” In an article last year, TechRepublic’s Nick Heath said, “As use of public cloud services grow, the demand for microservers suited to handling the lighter cloud service workloads is also likely to grow.” It appears he was right, as cloud companies are seeing the fulfillment of a clear need by microservers.
According to Silicon Semiconductor, IC Insights said that “between 2012 and 2017, microserver sales are projected to rise by a CAGR of nearly 72 percent, totaling $1.2 billion.” Meanwhile,Markets and Markets expects sales to exceed three billion in 2020. In other words, it’s possible this market segment could almost triple between 2017 and 2020 if the predictions are accurate.
Depending on the data center environment (racks in use, power distribution, heating/cooling), a microserver rollout might involve changing the layout and resource usage. For instance, traditional racks might have to be removed in favor of alternatives geared to hold microservers. Cooling may have to be adjusted to ensure that it remains appropriate for the needs of the systems. Power distribution units might have to be revamped to accommodate the changes. This can translate to greater costs to include in any provisioning project.
Microservers are well positioned across an array of business sizes as well as data centers and cloud companies — at least for those with specific application or functional requirements. Most interesting is their potential to rekindle the physical vs. virtual server debate, which many considered closed once the advantages of virtualization became well known. Many prominent server manufacturers, such as Dell, HP, and IBM, are in the game, and their efforts are helping drive advances in microserver technology. At present, North America is the largest market for this industry.
Jim O’Reilly of DatacenterAcceleration recently shared some interesting insights about microservers: “Microservers are in the embryonic stage of growth. The next few years will see ARM 64 multicore processors and much beefier Atoms replace the 32-bit processors. DRAM will grow, and SSD will speed the IO side no end. We can expect 10GbE or better Ethernet as standard in 2014, which will give us enough bandwidth for networked storage connection.” He added, “The current trend towards software-defined datacenters opens an opportunity for low-powered simple servers in large quantities.” Furthermore, a key point in the microserver future may be “the conjoining of GPU [graphics processing unit] and CPU that some microserver players are involved in.”
If your company is considering microservers, it seems a safe bet that the field will continue to bring new developments in the areas of distributed resources, space/cost savings, and customized workloads for certain business needs. It’s an equally safe bet that as with any new technology, a careful assessment of the proper strategies and best practices to meet those business needs will be essential before embarking upon a deploymen