There is an amazing amount of interest in the microserver market of late but very little actual substance to back up the bottomless well of hype. Lets take a look at the pieces that make up the market and hopefully separate the hype from the substance.
The first question you should ask whenever someone mentions microservers is exactly what they mean by the term. Microservers can be anything from a NAS box with a downclocked cell phone chip to a multi-rack behemoth meant to run the largest data centers in the world. There is no single definition of microserver, it is more nebulous than cloud or Web 2.0.
Similarly the solutions that people are touting for those interested in microservers run the gamut from 1S boxes that barely warrant the term “widget” to something with a software stack meant to run a social network. On top of these solutions there are yet more layers of stuff being offered, and yes that term is vague on purpose but it is mercifully less nebulous than the products that it encompasses.
If anyone uses the term microservers in a conversation, requirements document, or worse yet a sales pitch, stop them dead in their tracks right there. Before they dare mention what they are doing with the category, make them define it, hard. SemiAccurate provided a hard definition, or at least listed the most basic questions you should ask before looking at Microservers and it took 17 pages.
Once you have come to a basic understanding of what a microserver is, be it a pocketable personal assistant or a multi-row rack installation at your NOC, then you need to figure out if the paradigm will work for you. Microservers are not guaranteed to work on any given workload much less your problem no matter what promises a vendor may make. Likewise they may not necessarily save you money or energy when real solutions are tested regardless of claims made in sales presentations.
Things that seem obvious when evaluating a single big core vs a single small core often don’t scale to larger numbers of devices, a key prerequisite for microserver deployments. Likewise there are enough hidden roadblocks in even the simplest and cleanest architectures to surprise the most seasoned data center veteran. Microservers are a whole new paradigm that follows some of the old rules but departs from the expected path more often than not.
Anyone evaluating microservers has four basic tests that need to be applied before any solution can be evaluated. Some of these are based on the architectures of the CPUs being offered, others are more complex and nuanced. All four tests are listed in the PDF linked above, and any solution that passes all four is not necessarily right for anything, it is simply not ruled out for server tasks.
At this point if you have defined a microserver as it relates to your needs, you can go on with the discussion. With the hard part of the discussion done the hype starts up. You may be able to define what microserver are, how they will work for you, but who is delivering a solution? Does that solution actually do what the company offering it claims? Even more importantly is it actually available to test or purchase? How about volume?
There is no shortage of companies making promises about microservers, CPUs, SoCs, servers, software, tools, and everything else. Each day brings new entrants, new configurations, and new markets promising to revolutionize everything you do. This ever-growing flood of microserver solutions is years old, but once you stop and look at what is really available now, reality throws cold water on everything. Worse yet if you compare the claims from a year or two ago to the market today, it is enough to make you cringe. Through all this the roadmap claims are getting more incredible every day. Lets take a look at what is real in the microserver market and who some of the key players are.
Note: The following is analysis for professional level subscribers only.
Disclosures: Charlie Demerjian and Stone Arch Networking Services, Inc. have no consulting relationships, investment relationships, or hold any investment positions with any of the companies mentioned in this report.
Latest posts by Charlie Demerjian (see all)
- HyperX ships it’s 60 millionth enthusiast memory module - Oct 15, 2018
- Bittware/Nallatech water cools 300W of Xilinx FPGA - Oct 12, 2018
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018