It is a sad fact of life that all of us, at some point in our lives, will experience the dissonance between what our mind envisions and what our bodies allow. If you have not yet experienced this “rite of passage,” good for you. But don’t revel too much because your time is coming. For the rest of us, examples abound in our own lives and those around us. I attended high school with a girl who was an Olympic-caliber gymnast; but I’m sure she would pop a “hammy” now if she tried to touch her toes. Another old friend of mine was a notorious late night party animal when we were in college. I asked him if he wanted to go to a concert with me a couple of weeks ago, and he declined stating the he is “worthless the next day if he stays up past 10 p.m. on a school night.” As I contemplated the dichotomy between the capabilities of my youthful past and my middle age now — climbing a flight of stairs has become akin to summiting Everest — I couldn’t help but think how analogous this is to a number of data centers that I’ve seen lately.
A lot of companies that are seeking new facilities tell providers how they’re going to “put this in the cloud,” and how they want to support new latency-sensitive applications, and how they want to get closer to their customers. Believe me, we’re as excited to hear about it as they are to tell us, but every data center provider can’t help but notice a recurring theme in what we’re hearing: While customers have given a great deal of thought to everything their new data center will require, their vision becomes a little less rosy and clear when the topic turns to the network. In other words, while their data center may be willing, their network is telling them no.
In many instances they are unaware that their existing network structure won’t support what they want to accomplish. For example, we met with one firm where the CIO was so enthusiastic about what the new facility they were planning would enable them to do that I thought he was about to start speaking in tongues. But when he described his network, it was immediately obvious that not only wouldn’t it support their more bandwidth intensive applications but they were paying way more than they needed to for what they had already. Since no one wants to hear that their baby is ugly, we politely explained that he may want to explore some alternatives.
This doesn’t mean that the CIO was incompetent. In fact, based on what we’ve heard from a number of his counterparts, he’s pretty normal. Due to the increasing volumes of data that pass between your servers and to the cloud, having “big pipes” just isn’t enough anymore.
IDC estimates that 50% of existing networks will be bandwidth-constrained as a result of the rapid increase in data volume for IoT. When you consider Gartner’s estimate that there will be over 25 billion devices connected to the IoT by 2020, you have to wonder if maybe they are being a little too conservative.
Not only will this thirst for bandwidth come from the information generated by IoT devices themselves, but from its capacity-consuming underlying components as well. The information flowing back and forth from the many private clouds that will be implemented will require large amounts of bandwidth. Further, enhanced technologies such as Docker that can support up to four to six times more server applications than a hypervisor running on an equivalent system will correspondingly require more capacity than most operators currently have available to support them.
The fundamental problem with the historical telco approach is that even the biggest pipe must be shared. Even a 1G or 10G line circuit ultimately hits a switch or router with an “oversubscribed” backplane. When you combine cloud architecture with the changing nature of data and enduser expectations and requirements, even the biggest pipe typically cannot support the sheer volume of organizations that are all using public cloud SaaS at the same time. For practical evidence of this phenomenon, try using an Xbox service at Christmas or watching Netflix on New Year’s Day. As a result, the associated delays in processing information lead to customer dissatisfaction in the case of video for example, or the value of the data being severely degraded — even to the point of worthlessness — in the case of IoT. Thus, organizations relying on this traditional data center network model are incurring costs that cannot be justified in terms of the ROI derived from their performance.
To alleviate these problems, companies will need to view their data center and network plans holistically. Fortunately, services such as direct cloud connection via “cloud exchanges” and technologies like 10, 25, 40, 50, and 10GB Ethernet switches either or will be available to enable you to construct a network capable of supporting high bandwidth requirements to ensure that you are able to address both your current and future needs. In short, unlike in our own personal experiences, your data center and your network will both be willing and (more importantly) able.