GigaOM just published an article titled Why 2013 is the year of ‘NoOps’ for programmers [Infographic] and since my comment doesn’t seem to show up (pending moderation apparently) I decided I might as well just reply on my blog. If you’re in a hurry here’s why that’s just not gonna happen: a PaaS is a wonderful thing that is giving an opportunity to ideas to see the light of the internet when otherwise they might have never made it. And while it drastically reduces the need for IT staff to almost zero, those (primarily startups) who can benefit from it still represent a small share of the IT market as a whole and that percentage won’t change dramatically any time soon. Now that those in a hurry have left let’s explore this a bit more in depth.
If you go and read the infographic, which is a good read by the way, you’ll realise a few things. First of all there are two points that are quite misleading: 1) the cloud/PaaS has cut costs 2) the cloud/PaaS has made companies more productive. The corollary to that is that nowadays an entrepreneur can launch a startup in a month or even a week by leveraging a PaaS, which is great, but that does not represent the entirety of the IT market.
The cost of the cloud
There are a lot of benefits associated to Cloud Computing and I certainly welcome this new opportunity, but if you look closely have the costs really been cut? What kind of costs? If I had bought a server on day one and installed rails on it and mysql or redis, at the end of the year, including bandwidth, would have I spent more or less? Most likely less and for obvious reasons. Behind the scene the good folks at your PaaS provider are doing all of that for you and they deserve to be paid! What is true however is that you don’t have that big upfront cost and if things don’t work out for your startup you can just terminate your contract and move on. So what’s true is not that the cloud or PaaS have cut costs, quite the contrary, what they have done is to cut down the investment required to get started. And don’t get me wrong, that’s a big deal, I’m leveraging that myself as I try to build my own PaaS based thing, but it’s important to not lose focus of what’s happening behind the scene. If you have managed hardware infrastructures in a datacentre and are now using cloud/PaaS I’d love to hear in the comments what your experience of the costs has been.
Productivity is not a function of what you produce
Likewise the productivity argument is not very solid. If you have 10 people, each one of them producing 1, 7 are devs and 3 are sysops, your productivity is 10. If now you have another 10 people still producing 1 unit each, but all of them are devs, is your group more productive? Certainly not. However, and this is a big advantage, having 3 more people working on the product has the potential of shortening your time to market, improve your product and a lot of other good stuff. But it has nothing to do with making you more productive, which is far too often one reason I hear people using to convince others to jump on the cloud bandwagon.
2013 won’t be the year of NoOps
Back to the corollary, it’s amazing that you can now launch a new product in weeks rather than months or years, but in fact this sort of thing is not new at all, people that made a living off of websites have already seen that kind of thing with one-click installations of wordpress, drupal and so on. The thing tho is that this is just a slice of the market. Yes PaaS are great, yes they almost reduce to zero the need for ops, yes in the years to come we’ll just see more of them, but will 2013 be the year of NoOps? Hell no. Not 2013, 2014 or any other year unless there’s some major breakthrough. There is a huge chunk of the market that has an actual increasing demand for complex environments and performances that are hard, if not impossible at this time, to achieve on any cloud or PaaS. And I’m not talking about enterprises, even the startup that has now grown into a 100 people company with millions of users is likely to have some particular needs, especially if they deal with big data. A PaaS, like all frameworks and abstractions, is great when you play by the book and don’t have special needs, which is almost always true if you’re small and just starting out, but as what you’re building gets more complex and you can’t play by the book anymore it’s game over. However, if you have built yourself something really complex in a PaaS, or know someone that has, I’d love to hear about it in the comments, I am truly excited about this whole development.