Infranomics: Non-technical Considerations for the Cloud

For those of you confused by the title of this post, don’t be… one of the things that keeps me interested in my job is the diversity I am allowed. My senior management likes to use an analogy of a capital letter “T” – that is, you should have a deep level of expertise in one technology, but just as important is to have breadth across many. It doesn’t just apply to technology though – there are many other facets that we need to consider, and economics is one of them. Me being an infrastructure guy at heart, I’m going to coin the term “Infranomics”.

So in a similar vein to this previous post of mine, this is another aspect of the cloud that you can and should start thinking about now. And it doesn’t require any expenditure up front other than a few Joules of brain power. And, like that other post, I’m not claiming to be the first person to think or talk about this but I will put my usual Enterprise slant on it, including the usual disclaimer that this probably is completely and utterly irrelevant for smaller shops.

There are 2 aspects I want to cover. The first relates to the cost of infrastructure in the Enterprise compared to some of the Cloud based services you can get externally. The second pertains to how the cost models potentially differ within an Enterprise, compared those outside.

Infrastructure Cost
Even within the cost of infrastructure, there are 2 considerations. This recent post by Stevie Chambers, Esq. is a great example of one – the total cost of infrastructure. As Steve points out, the guy who makes the compute purchasing decisions quite often doesn’t give any consideration to the downstream infrastructure costs. This is better than what it used to be in the pre-blade days, I’ll bet my bottom dollar that nobody who recommended the 4U behemoth with 10 NICs and 4 HBAs as the ESX hardware standard 3 years ago gave a second thought to the port and cabling costs for all that connectivity (that got choked once it passed the access layer anyway), let alone the power costs. But it’s still far from good. The Enterprise has a bad habit of over-provisioning, it’s the reason virtualisation became such a sucessful technology in the first place (of course there are many other reasons now, but the core value proposition of VMware has historically been about less physical kit and less datacenter space). This is exacerbated by the technology silos that are evident in most large organisations. The server team wants to fattest pipes from the network team and the fastest most redundant disk from the storage team, often without a clue as to what the actual requirement is or the cost implications outside of the server hardware. I propose that this is not so prevalent in external hosting companies – their business model is built on the notion of sharing and performance SLA’s. They’ve been doing this way longer than the Enterprise, allowing them to better size their infrastructure and thus get price advantage #1.

The second price advantage that hosting companies may achieve on the hardware front comes from their focus on using software to overcome problems that an Enterprise typically uses hardware for. Lets take storage for example. Online backup provider Backblaze recently posted up their storage hardware design – a custom built box from commodity hardware that allows them to get 67TB of raw (note that, RAW) disk for under $8K USD. Now, the pricing comparisons in that post are a little disingenuous in my opinion, because they must have built a lot of intelligence into their software layer to overcome the obvious hardware layer deficiencies in their design (funnily enough, the first thing I asked them was why they didn’t use OpenSolaris / ZFS too ;)). Which is absolutely fine, but it’s not free. It’s not like an Enterprise could implement the Backblaze hardware design without the software layer to overcome the corners cut in the hardware layer, and so any comparison with products from EMC / NetApp / Sun etc is not really valid. Enterprises just don’t work that way – the argument is always that it’s cheaper to shell out more cash for “Enterprise class” hardware and software from 3rd party vendors, because of staff turnover in the Enterprise. For the record, I’m planting both my feet firmly in the sceptics camp with regards to that argument, but that is nonetheless how it is today. Same concept applies for the likes of Google and Amazon – we have the same “design for failure” philosophy in the Enterprise, but we implement it moreso in hardware where they prefer to do it in software. So there you have price advantage #2. Yes, it’s a little flaky due to what I mentioned before about the higher software costs needed to overcome limitations of lower hardware costs, but Enterprise hardware is not without software associated costs anyway. I heard a rumour that the core of Amazon’s EC2 team was around 30 developers, who handle everything from the custom Linux / Xen platform that runs on the hardware, to the custom software that handles everything from infrastructure level monitoring to VM lifecycle and integration with the other Amazon Web Services. If that team had a human cost of around $3 million a year, I bet it would still be an order of magnitude cheaper than the software costs that an Enterprise would incur to get the same functionality from 3rd party vendors. And that’s on top of the more expensive Enterprise hardware. Yes this is pure speculation, I have absolutely nothing to back either side of that argument up other than a hunch. But it’s a helluva hunch ;).

Business / Cost Model
As this recent post from Simon Wardley alludes, the biggest challenge Enterprise IT is likely to face when competing on a cost basis with external Cloud providers is grounded in the way most internal IT organisations are financially run today. This is really the crux of this post too – you need to start thinking about this and challenging the way things are done now, or you will struggle to move to the cloud.

In most Enterprises (and I have worked at more than a few), IT is treated as a cost center. And because of this, IT is not actually allowed to “make money” – the books basically need to be zero’ed out each month / quarter / financial year (the time period may vary, but the end result is the same). Now the only way to actually make this work is to basically wash the cost of IT across the entire organisation. And so when you come along with your internal Cloud proposals and try to show a cost benefit that will be tangible to the consumers (ie “the business”), people on the finance side of IT are likely to just see another bucket in which costs can be washed or even worse hidden, thus eliminating any tangible benefit to the business entirely. I’ve seen this manifest is so many ways over the years with so many things it’s just not funny. The cost of lower tier storage artificially inflated so the higher tier storage can be artifically deflated. Running costs such as existing software license renewals put into totally unrelated project budgets. And on. And on. I bet 90% of you reading this are smiling to yourselves right now because you’ve seen the very same thing.

The flipside of this is that external Cloud providers can play the exact same game of artificially lowering the cost of something in order to attract business, the practice of “loss leading”. Such Infranomic creativity is simply not an option for Enterprise IT. I’m not even talking about the practices of Microsoft using revenue from Operating Systems and Office to fund development of Internet Explorer or Media Center and the like, or the way Sony makes a loss on a new games console for a few years until the component costs come down enough for it to start making a profit at the same consumer price point. I’m talking about really creative stuff.

For example, the way Amazon can offer such low prices on books. This is not merely achieved through volume purchasing as you may think – part of it comes from being creative with payment terms and a mastery of the art of inventory management. Say your average payment terms to a publisher are 60 days. When Amazon places an order, they will likely do so with a very accurate estimate of how many copies they will sell in the first 30 days. And the next 30 days. And the 30 after that. And so on. This gives them a greater ability to recoup the entire cost of their orders well in advance of the actual payment date. Hence, they have the opporunity to make money by investing that cash until the payment is due. Just how much they can make by doing this anyone’s guess, but you get the idea.

Without a fundamental change to the way the internal Enteprise IT cost model is treated as a whole, there is no way it can compete with the outside world on the basis of cost alone. Should it be treated as an external business unit, and be allowed to post profits and losses? Should it have the opportunity to actually offer it’s services to other organisations? Perhaps if that was an option, internal software and hardware development might blossom the way it does in the web startup industry.

And with that, I’ll wrap this post up. Hopefully you don’t feel like I’ve wasted your time, and have a better understanding of some of the non-technical challenges that the Cloud will bring. In my experience, such challenges are far more difficult to overcome than technical ones, and are in many cases more important.


2 Responses to “Infranomics: Non-technical Considerations for the Cloud”

  1. Infranomics: Non-technical Considerations for the Cloud | Software Management Says:

    […] Read more: Infranomics: Non-technical Considerations for the Cloud […]

  2. Derik Pereira Says:

    Obviously, if Enterprise IT adopts the ITIL process framework using an appropriate subset of process for the enterprise then all will be well. In practice, it may be less troublesome for them to adopt cloud and continue the ITIL lip service. The ITIL books themselves map all this out quite well as a published standard. Obviously, such standardising on ITIL should rightly be a CIO driven initiative. Then such initiatives as cloud from a CTO would indeed be driven with solid financial economic decisions. Now, if everything worked so well there may not be chaos and we need chaos to thrive and innovate against all the forces. So, maybe we should just chug along.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: