Microservers take the familiar concept of the blade server,
which essentially puts multiple servers into one physical chassis for shared
power and cooling, to another level. A Dell microserver chassis, for example,
houses 12 physical servers in 3U of rack space, theoretically lowering space,
cooling, and hardware costs. However, there are several challenges to
widespread microserver adoption. Here are some of them.

1: Unfamiliarity

Microservers bring a fairly new hardware form factor to the
data center, eschewing the blade concept that’s been in place for a number of
years. While the concepts are fundamentally the same, everyone from data center
designers to front-line operators will have to learn new design and operational
considerations.

2: A gunfight at the processor corral

The mainstream X86 server market has basically standardized
around Intel processors, with AMD grabbing limited market share. Intel’s server
chips are powerful beasts with well-understood capabilities. In the microserver
market, Intel has introduced lower-powered versions of its traditional server
chips, as well as a “server-grade” version of the Atom chip. However,
the ARM-based chips that dominate mobile phones and tablets are also making a
play for the data center, and workloads that are targeted at microservers seem
perfect for ARM designs. Companies may consider waiting out the first few
generations in the microserver market until a processor architecture becomes
standardized.

3: The curse of the Atom

One of Intel’s two classes of microserver processors uses
the company’s Atom branding. Recent Atom processors are excellent in terms of
power consumption and processing capabilities for the right workloads, yet the
brand still engenders skepticism, particularly from people with bad memories of
the early days of the netbook craze. While it’s patently unfair to discount the
current generation of Atom processors, the brand alone will cause some to
dismiss microservers.

4: Sizing

Sizing traditional servers is a known commodity. Most
vendors and implementation providers have proven sizing tools that quickly
determine server and processor configurations. With microservers being somewhat
new to the market, you or your partner may be taking a bit of a guess with
sizing a microserver deployment — a guess that could be fairly expensive if
done poorly.

5: Lack of clarity around workloads

Microservers have a compelling story around power and space
savings, but when it comes to the types of workloads they’re best equipped to
handle, the story becomes a bit less clear. Since microserver processors
generally eliminate some of the technology embedded in higher-end server chips,
it’s fairly obvious they’re not the best choice for complex analytics or
graphical calculations. Outside some well-defined niches, it’s hard to tell if
a microserver is the best choice. Most vendors mention areas like web serving
or serving up virtual desktops, but they lack compelling use cases that show
the savings that can be achieved by a microserver. After all, if a workload
that could be accomplished with two traditional blades takes four or five
microservers, savings may be illusory at best.

6: Open Compute

Technology companies like Facebook have been pioneering nontraditional
servers for a number of years and have launched the Open Compute platform, which specifies designs for servers targeted toward
high-density web service providers. While Open Compute is arguably a distinct
physical platform versus microservers, it’s targeted toward a similar market
and presents early adopters with a major, divergent choice between Open Compute
and microservers.

7: Competing against virtualization

Facebook and others have largely proven that a giant farm of
microservers or a similar technology is the best way to handle millions of
concurrent web requests, but that doesn’t mean a move away from virtualization
is appropriate for every data center. Many companies have invested
significantly in their virtualized infrastructure, and microservers return us
to the days of the physical server. There is certainly some overhead to
virtualization, in both technical and human terms, but a pool of physical
servers also poses its own challenges.

8: Managing workloads

One of the great benefits of technologies like
virtualization is that you can dynamically allocate processor and memory
resources as workloads shift. With microservers, you are physically committed
to a fixed level of resources once you place your order. With effective
planning and workload balancing tools in front of your servers, this may be an
easy task. But for other workloads, you may struggle to break them into chunks
that can be processed in parallel.

9: Moving to the cloud

Companies are increasingly hesitant about owning fixed IT
assets like data centers and their racks of servers. In extreme cases, I’ve
encountered clients who want to own no
IT infrastructure other than networks that connect their leased end-user
devices to cloud services. A rack full of microservers now competes against a
cloud provider, which in some cases allows you to simply write a check and let
someone else worry about the debates concerning which hardware to select and
how to provision and maintain it.

10: A product looking for a market?

I can’t shake the feeling that the microserver might be a
product in search of a market. Sure, there are niche markets where the
microserver might fit nicely. But you’re paying a premium for technology that’s
yet to be fully standardized, has complexities around workload, and is new to
the market. The upshot for vendors is that they can charge a slightly higher price
since microservers have not been subjected to the brutal competition and
commoditization of the standard X86-based server market. Unless you’re a
high-volume web provider or highly sensitive to power requirements, make sure
there’s a compelling case for the move to microservers before jumping onboard.