This is a good post on the above question; it nails one of the biggest problems that we have internally; building out our internal cloud so that we can provision from it. Not technically an issue, we know how to do this, well more or less!
Our projects tend to be point-funded; charge-back to projects is also hard to implement; this is not a technical problem, this is a budgetary problem. Building cloud infrastructures does not fit with with either current budgtary practises but also it does fit well with the current annualised budgetary cycle.
Hi Martin,
Nice blog. I’m curious your perspective on how a private cloud storage vendor can work to overcome these issues. Is it simply an accounting problem? I have some ideas that we are implementing but real world feedback is always welcome. Feel free to drop me an email if you would like to chat.
MwM
———
Mike Maxey
ParaScale
Is it just an accounting problem? Actually, pretty much at the moment; building the cloud to support the agility which is beginning to demanded by the business is actually technically not especially challenging.
Virtualisation of servers is now well understood and accepted; there are a number of good namespace products to support file-level virtualisation. Put it like this, if you gave me a few million bucks, I could build you an private cloud (assuming I don’t need to build a new data centre).
But it is hard to convince our friends in the finance departments that build it and they will come is a prudent approach.
With a few million dollars you could build the mother of all clouds. Our customers can get started for under $20,000 and scale as needed. Which raises the accounting question of what is more painful. A giant “build it and they will come” capital outlay or incremental purchases throughout the year. My gut tells me incremental is more palatable and enables the company to leverage the latest capacities at lower costs. Do you agree? But doesn’t that create challenges for the charge-back model?
MwM
Internal charge-back models are a nightmare for any large company; well they can be anyway. For example, a lot of companies have a ‘use it or loose it’ approach to budgeting; so at the end of the year, everybody goes insane and tries to burn all the money they’ve got left over. So what happens if you have an internal charge-back model, is that all the projects try to spend their money in the internal market; it’s quick, easy, doesn’t involve procurement etc. Unfortunately, the poor internal supplier ends up with a massive surplus in the last weeks of the financial year and no-way to spend it; so they loose it.
Now, the way to get around this is to allow the cloud/infrastructure to be purchased up front or go down a capacity on demand model. Problem with most of the capacity on demand model s is that they require a leasing framework. I happen to work for a cash rich organisation and they don’t like leasing but they also don’t like spending lots of money up front without knowing precisely what they are going to do but they also want to be agile, react to market demand etc, etc. Doing lots of incremental purchases doesn’t help really; sure, it helps financially but I’d end up implementing/procuring increments almost on a weekly basis. And the overhead that’d put on me with regards to following ITIL processes would be a killer.
Its a hard problem to solve and it is almost entirely a budgetary/financial one. Build-out is the right thing to do but it requires a mindset change from point-based project funding and an understanding what a true shared infrastructure means and brings to the business in the way of agility. Still I remember having these discussions 10 years ago; I think it is going to just be hard.
One of the most outrageous schemes I’ve heard of is a “compute resource marketplace” where potential users “bid” on bundles of cpu, memory and storage for a defined time.
Intellectually fascinating, but entirely unpragmatic.
That being said, I believe there are many classes of IT investment (and all forms of infrastructure) that suffer this classic problem: networks, email, phone systems, etc.
Given my longevity in this business, there were times when we debated “how to pay for the network” and “how to pay for file servers” and “how to pay for email”.
Somehow, strangely, they all got paid for, and we’re all using them today. And I don’t get an itemized bill for my email traffic (thankfully!)
As I talk to many customers, they’re starting to realize that they just need to bit the bullet, build a big (virtualized) playpen, and offer it up to people — and figure out who pays for it later.
I’m sure we’ll get there eventually but at the moment, it does take a lot of my time and energy trying to convince our friends in the finance teams that we need to do it in this way. I’d much rather be trying to build the cloud and solving the technical challenges than solving budgetary problems. If I’d wanted to do that, I would have carried on with an unwise foray into accountancy.
Chuck, I agree on your auction comment. People want predictable pricing with declining costs as the technology components get cheaper. An always auction pricing scheme is impossible to plan for in the future as demand spikes impact COGs.
Although the build it and they will come method seems to be an artifact of a time when the only options for systems were $1 Million plus. Look at the success of SaaS providers like salesforce.com or Google. Solutions that start small but are scalable enough to run the enterprise. New classes of storage architectures are changing the infrastructure game in a similar way as these application providers. Don’t get me wrong, I’m not advocating storage-as-a-service as the security and lack of data control issues are rampant but the architectures that enable these types of services are compelling and will be deployed inside the firewall. The accountants of the world will likely get heartburn but the ROI and cost savings are highly compelling.
MwM