Is Ubiquitous Internet Access Gating Public Cloud Adoption?

 

Article written by Leo Reiter
 

I recently had a friendly email debate with an experienced executive and investor who I respect tremendously. I argued that I would never pitch a private cloud idea to a venture capitalist knowing what I know now, and that new investment should go into public cloud technologies as a bet on the future. It’s true that private/hybrid cloud is strong today and will continue to be relevant, but the really explosive growth – the transformative growth – is in the public cloud space. As large scale computer systems and networks mature and we spend more time solving problems with them than putting them together, it’s only logical that we simply outsource their operation to others (who specialize in that sort of thing). Obviously private cloud operation is commonly outsourced/hosted as well, but the context of the discussion generally assumed such clouds would be operated on premises.

Inevitably the objection of ubiquitous internet access came up, as did my standard rebuttal of “not long ago we had the same concern about electricity”. I wholeheartedly believe this comparison is valid – Internet connectivity is as key to our progress today as electricity was a century ago – so I decided to do a simple comparison of the evolution of the two megatrends. Specifically, what are the inflection points we can compare? And… if we agree that the (developed) world assumes ubiquitous electricity, how long before the same can be said about reliable Internet connectivity?
 

The futurist in me argues that ubiquitous Internet access is already here. My “day” job as a CTO of a public cloud company demands this. But to be truly objective, I wanted to do a little research. Obviously I’m not going to overthink something that I already accept to be true, but I felt that there must be some straightforward data to draw from. I was both pleased and somewhat surprised by what I found. Note that this is a very simple, limited comparison of two very different things, so think of it more as a thought exercise than any sort of absolute truth.
 

For the purpose of this study, I’m focusing only on the U.S., but most of the developed world followed a similar trajectory. Without getting into alternating versus direct current, etc., we can say with certainty that the first electrical power plant in the U.S. opened in 1882 – the Pearl Street Station in lower Manhattan.

 

For the first time, consumers could actually purchase electric power from what would eventually become “the grid”, albeit for a price. Adjusted for inflation, the $4-$5 per kilowatt hour (in 2013 dollars) of the day was quite cost prohibitive, and limited access to the wealthy. No doubt these were the same consumers who could afford the latest electrical devices (e.g. lamps) in the first place. We’d have to fast forward to about 1918, when the average cost of a kilowatt hour dropped to the $1 range (in 2013 dollars), and remained there until 1930, to reach a level of affordability that would support wider use. In the 1920s independent utilities started to consolidate, delivering economies of scale to the industry, and eventually bringing us the power grid we know of today. For the sake of argument, let’s take 1930 as the inflection point where electricity started to become ubiquitous in the U.S. (at least available to those who wanted it), and manufacturers could start assuming it was there when creating their latest products. This means that in less than 50 years, the U.S. witnessed the rise of the power grid as we know it today, from the very first power plant. If I was an aspiring entrepreneur in 1920, I would likely have pitched an electrically-powered idea versus a steam powered one, for example. When placing bets on markets, you want to bet into the (tangible) future, not the present (or the past) – at least if you want to make money in the long term!

 

Today, we can agree that virtually every home and business in the U.S. that wants electric power has access to it. True, there are outliers, like the very poor who simply cannot afford electricity, to those who prefer to live off the grid. In both cases these are minorities – electricity is considered just as critical as running water, both per conventional wisdom and laws. For those living off the grid, this means precisely that in almost all cases – they are not using oil lamps, but rather generating their own electricity rather than buying it commercially. But they are still using electricity – whether it comes from solar panels on the roof or the substation down the street, it’s still ubiquitous.
 

While the Internet itself was in place decades earlier, we can trace the first commercial Internet Service Provider in the U.S. to 1989. In Brookline, Massachusetts, The World served its first customer in November of that year. Just like commercial electricity, the rest of the developed world followed a similar path – e.g. Australia established its first ISPs in 1989 as well.

 

 

2 original Brookline headquarters of The World
 

By the the mid 1990’s, Internet access in the U.S. was spotty at best. Most applications were limited to dial-up access, on low speed copper circuits, connecting to a very limited set of services. In 1995 regulatory restrictions were lifted and the world-wide web was free to become fully commercialized. We all know what happened next. An important inflection point was in 1997 when DOCSIS 1.0 (for cable TV lines) came about, paving the way for cheap and reliable broadband to start displacing dial up service over the next few years.
 

Just as $1/kilowatt hour electricity in 1918 didn’t suddenly mean it was widely deployed, the invention of DOCSIS didn’t immediately lead to widespread broadband adoption either. In 2000, only 3% of the U.S. adult population used broadband Internet access. But in 2013 that number became 70% – virtually anyone who wanted and could afford the relatively low cost options had access to commercial broadband just over 15 years after it was standardized. Comparing to the rise of the electric grid, this means that in just under 25 years, Internet access not only became fast, but also fairly ubiquitous. This is also true in the rest of the developed world, which often leads the U.S. in adoption. It’s not coming, it’s here. Let’s also not forget that this only tells half the story, at best. The explosion of mobile device use in the last decade (such as smart phones and tablets) means that the market has fully embraced and in some cases takes for granted this ubiquitous availability. If access is seen as a gate to public cloud adoption, I would argue that consumers (businesses or individuals) are not paying enough attention. Today we have a myriad of access options, including satellite, mobile broadband, wired, and even peer to peer networks. If you can’t get broadband from your cable or telephone provider, you can likely get it from your mobile carrier instead, regardless of where you live or where your business is based. But is it reliable?
 

No complex system can offer 100% reliability in all conditions. This is as true of the power grid as it is about “the Internet”, whether it’s the backbone itself, the local access point, or something in between. But this doesn’t mean we must assume systems are unreliable – it’s important to design services for the rule, not the exception. For example, an electric kitchen appliance doesn’t come with a generator for backup because of occasional interruptions of commercial power. We treat that as an outage – even if it means we can’t do something critical at that very moment (such as prepare food). The same is true for the Internet and by extension the public cloud. We don’t stop using email, posting social media updates, downloading content, backing up files, etc. – just because there are times when networks are unavailable. In most cases service providers offer very strong SLAs, promising (and delivering) very minimal downtime throughout a calendar year. Save for a natural disaster, where everything is interrupted, not just Internet, it’s rare to lose access. Obviously there are times when we are underground or in the air and momentarily lose connectivity, but it either comes back quickly or we have access to alternative networks (such as in-flight WiFi, etc.)
 

Nevertheless, there are some activities that are just too critical to interrupt, and we need seamless backup systems to mitigate outages. This is true with hospitals, who have reliable backup power so that operating rooms don’t go dark in the middle of surgery for example. For power grid outages, the common solution is backup power from batteries and/or generators. Usually there is limited runtime from these systems, because again, outages are the exception, not the rule. Save for a cataclysmic event that takes out the power grid entirely, the “lights” will come back on, generally fairly quickly.
 

The generally accepted mechanism for mitigating Internet access outages is caching. That is, devices store some relevant information to both speed up operations as well as continue operating while momentarily offline. An obvious example of this is songs downloaded from a streaming music service. On another note, this is where the public cloud adds even more value, with its “designed to fail” philosophy leading to stateless services. Even if a client disconnects, services continue to run. For example, if doing a simulation in the cloud, just because the client disconnects doesn’t mean the simulation stops.  When the client reconnects, it has access to the latest simulation results. This is strong evolution from the client/server days, where everything was mostly session based and state depended on connectivity.
 

More advanced users opt for redundant paths to the Internet, subscribing to multiple providers so one can be used in place of the other should there be extended outages. Yes there is cost involved with this, but if systems are critical enough, the cost of being offline is much higher (and sometimes incalculable).
 

What happens if all networks are down for a prolonged period of time? Chances are a natural disaster took place, and you are home with your family rather than trying to figure out how to keep using cloud services. Again, this is the (rare) exception, not the rule. Even in this case, once access finally returns, the public cloud will still be there where you left it. Unfortunately the same cannot always be said for physical assets (be they computers or personal items) affected by fire, flooding, winds, etc. In fact, the public cloud keeps your data and processes safer. Sure, you can build a private cloud that is just as resilient, but why? Unless you have a regulatory reason to do so, you are not gaining any advantage from duplicating what public cloud providers already offer at much lower up front cost and much broader scale.
 

Private and hybrid clouds will continue to exist for some time – the world does not change overnight. We still depend on mainframes for many of our daily services, even if they’ve been generally displaced by more modern technology, for example. But the argument that Internet access is not reliable nor ubiquitous enough for public cloud is dated, and I would suggest no longer valid in almost all cases.