Cloud Computing – Hype or Here to Stay?
Summary: Cloud computing is over-hyped but it is important and businesses will increasingly be run on cloud resources, both because new businesses will start up on the cloud and because existing businesses will slowly migrate onto the cloud.
The other day I read a review for an inkjet printer – the Lomond Evojet 2. Standard inkjet printers have about 4,600 nozzles on a print-head that moves back and forth across the paper. The Evojet has about 70,000 nozzles on a stationary print-head under which the paper travels, allowing for crisper graphics and faster print speeds, but at a higher cost of construction. It turns out, though, that over time the cost per printed page is quite economical and in many cases actually lower than competitive inkjets that may have lower up-front purchase prices. So what has Lomond done about this, and, in general, what’s a company to do when they have high purchase price but lower total cost of ownership? For a hint, let me digress even further, and then I’ll get to the fluffy white stuff.
A company once made drill-bits that lasted 5 times as long as their competitors’ drill-bits, but cost twice as much to make. Market research and real-life experience taught them that the average consumer, standing in the tools section of the hardware store, 9 times out of 10 would choose the lower cost drill-bit. This company should have stopped selling drill-bits and started selling holes! It’s a hypothetical example to make a point, but consider the very real example of Rolls Royce. You may think they sell jet engines, but they actually sell thrust, or hot air ejected out the back of jet engines, what they call and have trade-marked as ‘power by the hour’. Rolls Royce provides jet propulsion as a service, a remarkable and successful business model. They take on all of the up-front costs and risks associated with development and manufacturing, as well as ongoing costs for maintenance and spare parts. Clients avoid high costs for purchase and maintenance programmes, and can more accurately predict their operating costs. It’s a pay-per-use model where clients pay for the amount of thrust they consume. And if you’re still interested in the printer example, Lomond will provide you with a printer and charge you per printed page.
In business computing, not PCs, but data centre computing where organisations run their core business applications, there have been previous attempts to adopt this model. Some have been successful, like salesforce.com. But most have fallen by the wayside, for many reasons that we’ll explore later in this article. But in the meantime, does anyone remember ASP? And no, I don’t mean Microsoft’s Active Server Pages. I mean “Applications Service Provider” which, for those of us who were in IT back then, was widely touted as the next big thing in computing at the tail end of the 90s. Although the Internet actually dates back to ARPANET in the 70s, it didn’t really pick up and become mainstream until the mid 90s with the development of the World Wide Web and the decommissioning of NSFNET, removing the last restrictions on the use of the net to carry commercial traffic.
I can’t find the original quotes on-line any more, after all it was 14 years ago, but I remember headlines like “ASP will dramatically change the way companies do business” and “I can’t imagine why any company would ever buy their own hardware or software when they can purchase business applications on-line for a fraction of the cost and hassle”. Well, I for one can imagine why they would continue to do so. Security and legacy integration are huge issues that need to be dealt with before a company will feel comfortable running their business on shared infrastructure.
I love Google Trends. Since 2004 they’ve been collecting and indexing data about what people search for on the Internet, and now you can perform trending searches against a whole plethora of topics. This chart tells us what’s happened to ASP. Quite likely more of the hits against ASP are to do with ASP.NET than they are with Application Service Providers, but either way it’s dropped almost completely out of the public consciousness.
Now take a look at ‘cloud computing’:
Basically nothing before 2008, peaking at the beginning of 2011, and now appears to be heading fairly rapidly in the same direction as ASP. So is that it for cloud computing? Probably not. Google Trends is just a tool that tells us what people are searching for and doesn’t necessarily correlate to what businesses are buying.
Amazon Web Services was on track to earn $1.5bn in 2012, now apparently accounts for 1% of global business computing, and has 70% market share in infrastructure-as-a-service. Their closest competitor is Rackspace at 10%, with everyone else making up the difference. Everyone else is Microsoft Azure, HP Cloud Compute, Google, Oracle, IBM, Dell and numerous others. It seems that hardware vendors are recognising that more and more of their customers want to buy holes as well as drill-bits!
Amazon pretty much invented the market back in 2004, and their cloud services now run hundreds of thousands of companies, including Netflix, Pinterest and Dropbox. Even organisations like NASA, Samsung and Unilever run some of their business on AWS. Thirteen employees built photo-sharing and social networking service Instagram on AWS and sold it to Facebook two years later for $1 billion, not bad for a company who didn’t have to buy any hardware or software.
But in terms of the global spend on data centre infrastructure to run their businesses, $1.5 billion is a drop in the ocean. So are we any closer to answering the question – is cloud computing hype or is it the next big thing?
Let’s consider for a moment the nature of business over time, and who buys compute infrastructure. New businesses are starting up all the time; certainly more and more new businesses are starting up directly onto cloud computing with no consideration for owning hardware or operating system software and all of the associated management bits and pieces around it. Existing businesses are doing what they do: changing existing systems, building new ones and rolling new ones in the door. More and more, however, the enterprise architects and solution designers in existing IT departments are considering their options for deploying at least some of the new stuff onto cloud resources. So I believe that over time we’ll see an increase in cloud adoption by existing organisations, but I don’t think many companies will actually sponsor the migration of existing applications. This costs money and the business case to change things that already work is often hard to prove. But new stuff, whether it’s new build or new COTS integration, is definitely a candidate for cloud deployment.
Then there’s the consideration of so-called private cloud. To me this is a sort of cloud lite – a different commercial model for hardware manufacturers to sell you their kit. But if it comes bundled with the same tools that public cloud has – allowing users to commission and decommission compute, storage and network services and only pay for what they use – and if it manages to increase utilisation because multiple departments and business units are sharing infrastructure, and if it actually costs less than buying the hardware outright, then okay, it’s a good thing. And I recognise that many organisations will struggle to trust the security of publicly shared infrastructure, and many will also struggle with the ability to connect cloud applications to legacy applications and data, because their legacy infrastructure is in different data centres and possibly different countries to the cloud resources. But I’d be wary of things like lock-in, minimum commitments, scalability, time to commission and de-commission, and interoperability with other cloud services, and whether this was going to give me a real cloud or just a different way to pay. But to give credit where it’s due, if I built servers or storage, I’d be building and selling public and private cloud offerings too. Or I’d be going out of business.
Companies will not only look to deploy their business applications on cloud resources. They will increasingly look to run their businesses on-line using commercial applications-as-a-service. I already mentioned salesforce.com, one of the original ASPs and still the most successful on-line provider of sales and customer management solutions, and who will turn over $3bn in revenue this year. Did you know that in the airline industry, most of the Star Alliance airlines run key operating functions like sales, reservations, inventory management and departure control on a shared applications-as-a-service model. Or that many Swiss banks run their entire businesses on a shared platform where they do not own any of the hardware or software? It’s taking a lot longer than anyone in the late 90’s anticipated, but applications-as-a-service is here to stay and will continue to grow. Already we see more and more migration onto standard business tools like email as a service or collaboration as a service (e.g. SharePoint). This trend will only add to the erosion of traditional hardware and software markets.
If you asked me to look into my crystal ball, then I’d say I have no doubt that in ten years time a majority of organisations will be sourcing at least half of their business applications as a service where they don’t own any hardware or software; and for the applications they want to own, they’ll be running at least half of these on hardware they don’t own. So if I’m right, this means that in ten years time only a fraction of the server and storage infrastructure out there will be owned by companies for their own use. This will have major implications for hardware and software companies, and if they cannot re-invent themselves, they’ll go the way of DEC, Sun Microsystems and many others.
Fluffy white computing is a big topic with many interpretations and players. I hope the above discussion has given you some insights and perhaps sparked some questions. If you would like help with cloud strategy or IT strategy in general, please get in touch.