It is normal as we change decades to reflect on many things. Firstly, I’ve just realised that this is my fourth decade in the computer industry, which is alarming to say the least! But more relevantly, I feel that now is the time to reflect on the significant changes in our industry in the last decade.
I do apologise to Joni Mitchell for my title; however, my view is that the “Cloud” has probably been the most significant. They came in “like thunder”, were “surrounded by fog”, and eventually they cleared! I’ll stop now…
Key Characteristics
So, what do I mean? The early days of hyper-scale (Amazon Web Services, Azure etc) cloud adoption were quite remarkable, characterised by immediate, “at scale” adoption by many. There’s two key characteristics that were, and still are, attractive. The first was agility- the simple ability to “stand up” a resource immediately and cope with uncertain scale. Second comes “Pay as you use” economics, Opex vs Capex.
I was told just a few years ago by many senior people that they, “have a cloud first policy” or “are migrating everything to the cloud”. Well anecdotally some tried but learnt fairly quickly that the “promise” or “illusion” of the cloud was not necessarily as it initially appeared.
I want to repeat that the cloud still is unrivalled by any other approach in terms of agility and cloud economics. However, over the first few years many companies learnt that the cloud, once scale developed, was not a “low cost option”. The reason is simple, the costs persist incrementally, the benefit of scale is never achieved. If this were not the case, who would ever operate a cloud?
Subsidiary Benefits
There are of course many subsidiary benefits that may still be critical. These include availability, security, compliance etc, but these we will consider separately at another time. There are of course some other “quirks” of cloud operations that most learn through bitter experience. “Shadow” IT is the phenomenon where employees simply “do their own thing” and set up accounts to populate with company and often client data. This is often without the knowledge of the operations community responsible for compliance.
The second common outcome is that instances are “spun up” and left unmanaged. They are not spun down when finished with, or scaled down when appropriate. In all cases these behaviours have the potential to increase cost and risk. Neither is a good outcome.
The clarity that came from the “fog” was that for most companies a Hybrid was the answer.
So, What Have We Learnt?
Whilst I absolutely accept that the scale and curve is meaningless, the idea is to convey the concept. My point is to illustrate that whilst the initial “entry” cost of the cloud is relatively low, the cost is incremental. Conversely, the cost of building a platform, assuming the capital cost and growing, at some scale the trend reverses. If this where not true, no one would build and operate a cloud.
The intersection of the two curves does however illustrate why the “Hybrid” may make great sense. Take advantage of the cloud characteristics of agility and low cost in the early phases of projects. Also, we must not underestimate the ability to “kill early”. Over the years I have witnessed projects that seemed like a great idea, but the platforms built were never used. They continued to “spin and blink” in racks for extended periods as the cost was sunk in the infrastructure. In “Cloud Economics” simply turn it off if it doesn’t work.
The Intersection
The key learning point from this discussion is that there is an optimum point that we should strive for. We need to understand when the initial benefits of agility and start-up cost are outweighed by a growing operational cost. Mitigation is brought in two ways- firstly by the use of Hybrid. Secondly, by using tools and techniques that once achieved, this state is maintained over time.
It is also interesting that in the Hybrid world, manufacturers have recognised this effect and are providing platforms with cloud economics. Lenovo seem to be leading the field currently, in my view with their excellent TruScale offering which does scale up and down. Many others have similar offerings which simply assume “Scale Up” not really in the spirit of “Pay as you go”.
Here at DONMAC Data we focus on helping our clients optimise operations at scale. Our experience and partnerships allow this exquisite balance to be achieved and maintained. If you wish to discuss this further please get in touch.
Opinion: David Leyland, Business Development Director, DONMAC Data
