Four Myths of the Container Data Center Market

Mark Thiele of Switch writes that pre-fabricated containers many not be the answer to every data center need. With aggressive marketing for what many vendors are calling "modular data centers" that he would term PODs or containers, vendors are asking, “Why build a data center when you can buy a POD?” This column covers some of the drawbacks of utilizing PODs.

Mark Thiele is executive VP of Data Center Tech at Switch, the operator of the SuperNAP data center in Las Vegas. Thiele blogs at SwitchScribe and at Data Center Pulse, where he is also president and founder. He can be found on Twitter at @mthiele10.

Mark Thiele, SwitchMARK THIELE
Switch

As our industry evolves, most enterprises will probably never need their own data center. Increasingly, it makes less sense to build one, especially considering the wide variety of colocation options, along with cloud providers building giant facilities that can be readily deployed without having to worry about massive investments in hardware.

Still, the marketing is aggressive for what many vendors are calling "modular data centers" that I would term PODs and containers. These vendors will ask you: “Why build a data center when you can buy a POD?”

PODS are different. PODS are often roughly the size of your typical semi-trailer. In my opinion, their best uses are in disaster areas, war zones and temporary situations. They are similar to how you employ a portable trailer for housing or on a construction site. A POD is better than a tent, but not suitable for long term mission critical work.
A POD may seem like a small mansion, but in reality it is a self-contained trailer. With that in mind, here are four myths about PODS that you need to be aware of when the sales people come calling.

Myth 1: Businesses don’t plan well and as a result should sacrifice the correct solution to buy PODs.

Reality: There are tradeoffs to be made for expediency and the consequences can be negative. If you aren’t planning well for your next data center, you will be prone to mis-steps no matter what options you select. This means you need to take a look at your planning process. There are experienced providers who would be willing to help you leverage their scale and work with you to create a road map for your future needs.

Myth 2: PODs are just like a small data center.

Reality: Modern data centers are modular  - that is to say, the IT equipment and supporting power and cooling can be deployed in "chunks." PODS and Containers are not modular in the same way. You can modify a traditional modular data center by increasing the floor space and changing layout configurations, just like you can with a home. Pretty much all you can do with a POD is add another unit to make it a double-wide. You can’t share plumbing or air conditioning capacity. It’s not like a house that has a shared infrastructure. With a house, you can add a bedroom or a fireplace. You don’t have that capability with a POD. A POD or Container data center is not a modular data center and there are constraints in how you will be able to expand as needs grow.

Myth 3: Every company has the need and capability in their IT infrastructure to move applications from one part of the world to another or from one POD to another.

Reality: While some percentage of companies would like to be able to shift workloads from one area of the globe to another, the capability to do so is still extremely limited. Most companies just plain don’t have a serious need for moving active workloads. If they do move a workload it’s one or two applications, not an entire POD or data center. Also, as big data becomes an ever larger part of your IT load and further bottlenecks your network, high density environments in well protected locations will become even more essential.  PODS do not offer that because they are usually less protected nor can you achieve the density that traditional settings can provide.

Myth 4: IT infrastructure is homogeneous and can be scaled in large volume increments.

Reality: Most organizations -large and small - don’t operate on a single hardware architecture strategy. The few that do are companies like Google, Facebook, Yahoo, Microsoft and eBay. Even very large enterprises rarely have a single hardware platform that supports a big enough workload to justify buying an entire POD of servers every time you need more capacity. I estimate that over 95% of enterprise applications still fit into the legacy design category. These legacy apps are generally not easily shuttled around the world from POD to POD. Even if you could move apps, you’d need redundant and duplicate infrastructure in multiple locations because it is usually cost prohibitive.

Let's Continue To Innovate

I’m all for pushing the envelope on data center design and efficiency, but not to the point where it affects a company’s ability to operate. There are a few people in the industry who are pushing PODs. Not surprisingly, most of them are the POD manufacturers, but there are also a few well known industry names who are also advocating for these pre-manufactured units. While I know and respect many of these folks, the majority of them aren’t running an organization that looks anything like 99% of the rest of the market. What works for them, in my view, won’t work for you and I.

Real Risks Associated with Thinking a POD is a Modular Data Center

A POD is a self-contained unit of capacity. In many configurations, each POD provides for its own UPS and cooling. In a large building with fifty (50) PODs, the entire load might be 15MVA of power and cooling, yet if the cooling or batteries in any one POD fail, that POD and its servers fail. There is no way to share the remaining 14.7 MVA of cooling to the failed POD.

There is virtually no stored CFM (Conditioned Air for Cooling) in a POD. In other words, when the HVAC dies, your servers cook. Even in a moderate density environment (300 Watts SF) the temperature can rise over 30 degrees in under a minute. In a high density environment with very limited stored CFM like a POD, the temperature rise could be more than double in that same minute. This high temp will likely negate any warranty on the gear and virtually ensure hardware and data integrity issues going forward. A POD is not a mission critical data center, it’s just that simple.

Benefits of a Shared Infrastructure in a Data Center

Because of the shared infrastructure in a large modular data center hall (which may have hundreds or thousands of cabinets in a shared space), the failure of any one device (UPS or HVAC) has little to no effect on the facility as a whole.  The enormous volume of stored CFM in a large hall can protect against even a complete failure of HVAC for several minutes with little effect on operations. Whether you’re in a colo or your own facility, having a larger space to manipulate and share greatly increases your ability to deal with change and emergencies.

In this modular data center environment, you are free to change HVAC units or replace batteries and increase or decrease availability Tiers as needed all in an active customer environment. In a typical POD, you’re unlikely to be able to change much of anything other than the racks and servers and you certainly won’t be able to make changes while running.

Look Before You Leap

So don’t let a fancy show and tell at a conference fool you into believing that somehow a small little space with some gear in it is a data center. Avoid the POD "trailer park" if you can. Instead, look to a combination of internal and partner resources to help you treat your data center capacity like the critical manufacturing resource that it is.

Industry Perspectives is a content channel at Data Center Knowledge highlighting thought leadership in the data center arena. See our guidelines and submission process for information on participating. View previously published Industry Perspectives in our Knowledge Library.

Hide comments

Comments

  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.
Publish