Rational Survivability

Hoff,

I'm pretty much in agreement with you on the fact that there is a place for both public and private clouds and that cloud is merely a waystation on the "Trail Of Technology" but I think the battery analogy isn't the right one. Batteries are temporary, disposable, and really aren't designed for intermediate to long term solutions.

Perhaps I'd better liken private clouds to home solar installation (or, even better, the solar power systems you find installed on the roofs of some large office buildings)… You can generate power enough for yourself, it's technically difficult (but not impossible) to do, but isn't a panacea for all your power needs…

Still thinking…

/Armorguy

There is no doubt it was a useful thought in 2002 when Nick blazed the trail, none of my snark on this topic is borne out of disrespect for that watershed moment.

While not a battery in a small scale sense, I find it hugely ironic that the one last bastion of on premise, albeit emergency, power generation and large scale battery usage are data-centers themselves.

Data-centers and hospitals are too critical to trust purely to public utilities even today a century into the technologies evolution. I suspect private clouds will have an even longer and more useful lifespan, for similar reasons.

Hoff, I understand where you are going with this. Armorguy, the battery analogy can be right for exactly the reasons you stated it was not. temporary, disposable are exactly what private data centers are. It is the perception of what we have to deal with to maintain a constant stream of information (analogous to power) In private clouds, SAN hard drives fail at 30% in 3-5years, CPU/RAM longevity is <3 years and cloud software (ESX/Xen) changes 1-2 times per year. In a public cloud if you made an EC2 instance, they (Amazon) probably experienced the same failure rates, but you as the consumer did not notice.

I would imagine a gas or diesel generator could be considered a hybrid cloud with a focused dependency on data residing in the public cloud. The solar installation would be a hybrid cloud with a stronger dependency on the data residing in the private cloud.

Great topic for discussion!

@randyjcress

Hoff, you are falling into the trap of those people who cling on to the electricity analogy in spite of its limited value. You don't need a battery analogy. Those who are hell bent on using the electricity analogy is similar to people who are hell bent on clinging to the idea of free mind in spite of evidence to the contrary. They cannot (or wont) go beyond the simplistic models. If we take an example from Physics, it is equivalent to clinging on to free electron theory to explain superconductivity, in spite of the fact the the electrons inside the superconductors are strongly correlated. Many people cling on to such simplistic theories because it uses much lower processing power of their brain and they can get away without upgrading their CPU. My suggestion will be to not find an equivalent example to electricity model. By using the battery comparison, you are giving them a chance to say "battery can only take you so much and you need public utility to have a more efficient workflow, blah, blah, blah.

@randyjcress,

I'm sorry but I don't see private data centers as 'temporary' – at least for the government or Fortune 500 (and probably even farther down the food chain). But that veers towards the religious side of the whole "public cloud vs. private cloud" and I want to avoid that…for now. 🙂

I'm also thinking that any analogy breaks down when analyzed too deeply – which is where I think your deep view of the generator and solar installation goes. That being said I, too, am diving down the Rabbit Hole…

I'm looking at this, admittedly, kind of simply. It seems to me that the "Go Public Cloud!" group uses the electric utility analogy to emphasize ease of use, availability, and standardization. But I think that if we really look at that analogy we may want to back away from it. Electricity didn't always get distributed the way it does. The early years were difficult, painful, and required massive government subsidies and allocation of monopolies to get deployed ubiquitously. Is that the road that public cloud needs to go down? (I really hope not, to be honest…)

But what does that say for private cloud? I think it says that just like the electrical power grid many people will subscribe to it and suck up the occasional loss of power there will be users that decide (for any of many valid reasons) that the risks of depending on the grid are too great and will expend the resources to build their own infrastructure. Some of that infrastructure will be supplementary (like UPS/generator) or in some cases complete replacement (like some solar).

In my mind neither is inherently "right" or "wrong". They're just two different ways to get the customer what they want at a cost and risk they are willing to accept.

@Krish Krish–yeah super agree on the urge for simplistic models overwhelming peoples genuine curiosity about the actual nature of a phenomena. Could have given a better example than your physics ones.

I think /hoff is just curating the discussion with this post more than helping the electricity agenda though.

Sh*t sorry, should say "couldn't have given a better example"

Batteries are simple (far simpler than even connecting to the electricity grid) and they store a limited amount of energy rather than generating it. That's why I use the example building/running/fuelling/maintaining a generator, which I think is a much more appropriate analogy.

Sam

@Armorguy Agreed, best to stay out of the rabbit hole with simple analogies to explain a very complex topic. But if you already have the concept and are not new to the ideas, it is good to reflect on different aspects and think about it from a different viewpoint.

Unfortunately sales and marketing will classify the cloud to their needs based on what the consumer will accept and purchase. Think back to the physical appliance model vs. the virtual appliance. Cisco still sells a majority of items as physical appliances because they can be procured by the networking teams that can't play in the VM arena.

My concern is that non-IT departments will become the majority consumer of public clouds leaving IT with the task of defending the slower-moving private cloud. Let's hope that "governance" doesn't fade as a buzzword while "cloud" continues it's momentum.

so you guys know why people use analogies? To try to explain a complex concept to people in a context that they can consume. In this case, the electric grid analogy is quite simply describing using computing as we use electricity. You use what you consume. That is the beginning and the end of the usefulness of the "cloud as electric grid" analogy. Discussing this topic ad nauseum is a phenomenal waste of time. Unless, of course, your goal is to write a book/paper/definition of cloud computing and are trying to select a good analogy.

Just remember, analogies convey big picture concepts, not details and minutia.

I think it's awfully arrogant and condescending for some folks here to insult the intelligence of anyone trying to explain a complex topic with a simple analogy. An attitude that like comes off as: "Your too stupid to understand". Much like asserting Private Clouds being DOA is to say to the consumer: "Your too stupid to build/buy your own cloud." Attitudes like that have been historically rejected by the free market.

Cheers,

Brad

@randyjcress,

I don't think letting "marketing define cloud" is, in and of itself, a bad thing. Marketing does what marketing does (hopefully sell more widgets) and IT does what IT does (hopefully get the right widgets to make the company work better). The consequences of the marketing/sales/support cycle, intended and not, are dealt with just like they have been for years and years.

I think that the key is exactly what we're doing here – have conversations about concepts, ideas, and particulars so that as we design and implement these clouds we're getting what we wanted. I think the most valuable conversations avoid the extremes and focus on developing systems that business wants at costs they are willing to spend and at risks they are willing to take. Prejudging any of these 3 things in an attempt to stereotype *all* business as doing *something* is not productive, IMHO.

@Armorguy

I agree with your comments around public vs private cloud. If I HAD to make this analogy (which BTW has been abused) then I would say public cloud is the utility (electric) company, which everyone subscribes to, pays for, and deals with the outages etc. I would say wind, solar energy is more like private cloud, in that because public utility (electricity) is not sufficient enough, I would want to have an option that provides more(?) reliability, QoS (CS) and I can control (more so than in public).

The bottom line is that it's not one or the other that enterprises will go with. I do hope that they will learn to take advantage of both, because I've said many times before, they both have benefits and advantages, depending on the functions/tasks/workloads at hand. The bigger challenge will be how these 2 will interoperate/integrate with each other, in order for enterprises to be able to extract value from both. Virtual private cloud is the closest thing to achieving this now, however, it remains to be seen just how seamless this will be.

The analogy of self-generation of power is the better analogy for private cloud than acquiring stored energy generated from someone else, IMHO. Solar panels, wind power, even gasoline-powered generators are fine examples of why private clouds are viable long term.

What happens when there is a major disaster in any developed part of the world (especially the United States), and public power distribution is disrupted? There is a run on generators at the local hardware store (or equivalent), that's what. Those that already owned generation (either in the form of solar or an existing gas generator) have a distinct advantage, as they are ready to deal with the situation well before the competition for portable generation begins.

This is a strategy for gaining an advantage in a heavily electrified world. Similarly, owning compute capacity compatible with your public-hosted compute payloads is a distinct advantage in a world where public networks are disrupted. Sure, you can distribute your payloads to another cloud provider, but if the *network* is disrupted, that won't do you much good, will it.

For necessary continuous "local" access to computing power (e.g. in hospitals, payroll processors, banks, research institutions, etc, etc, etc), private clouds are a no brainer. However, even in other situations, precisely because data != electricity, private clouds may still make sense. "Governance" is a key word here (though I hope that's a transitional situation), as is "existing investment".

I've been railing on Twitter against relying too much on the electric utility analogy for cloud computing specifically when it is used as a justification for ignoring any form of private cloud. Such justification is in fact unfortunate, as it is simply an analogy, and does not reflect on the true complexities that computing has above and beyond electricity generation and distribution.

@beaker

@Brad Hedlund

I think @beaker nails it. It was not an arrogant response to insult people building the model. It was just an early morning response to beating a dead horse. You may not have got my thought process as we have never interacted on this topic. My argument is that as the technology improves further, economics will drive the adoption to public clouds in most of the usecases. I have always argued that one should let the markets decide on public-private debate than one side dismissing the other.

Having said that I also want to argue that the electricity model is too simplistic and there is no analogy to data in the utility model (for a better description of my thoughts on this, please refer to http://www.cloudave.com/link/cloud-computings-ele…. You cannot just jump into public clouds just like how we get our utility turned on (at least at this point of time) and due diligence is important. I suppose I have made my thinking on this clear to you now. My comment is more related to the "dead horse beating" than anything to do with the models. Sorry, if I had come across as someone who is insulting the "defenders of public clouds".

@James Urquhart, I agree.

In fact, I quoted solar power analogy to @samj during one of the twitbates. In spite of the availability of public utilities, some people use solar power or wind power just because

1) they can (think of some big enterprises willing to spend or being able to spend)

2) they want (to help the environment or they are in a geographical location where it might turn out to be comparable to public utilities in the long run)

Yes, public clouds have a great future but it doesn't mean that private clouds have no role in IT. Let the customer decide. Pre-guessing what customer wants by dismissing private clouds is not a market based approach. Thatz the crux of my argument.

Hi Hoff,

Some general responses to the comments etc.

The analogy between computing resource provision and electric and other utilities (gas, water etc) was made by Douglas Parkhill in "The Challenge of Computer Utilities", 1966. Douglas Parkhill also went on to discuss the issue of private / public / government and hybrid forms of utilities (as occurs in other "utility" industries) along with a host of security, legal, data privacy issues and exchange issues.

The analogy pre-dates Carr by some time. However, the analogy has always been accepted to have limitations mainly due to the relationship between the consumer of services and provider. In the case of electricity, the relationship to the provider is minimal (particularly since it has standardised). In the case of computer resources, the relationship varies (according to what level of the stack you're using) and involves some element of data & code (i.e. you're code and data exists on the providers systems).

Naturally, it's slightly more complex than this as in the case of SaaS, you're data exists within the context of the service providers application, whereas in the the case of PaaS you're data exists within the context of your application but the code exists within the context of the service providers platform. The difference is subtle but very important for the issue of portability.

When it comes to the role of public / private, the shift of ubiquitous and well defined IT activities from a product (i.e. "infrastructure as a product" you buy / own) to a service ("infrastructure as a service" you rent / consume) economy is no different from the many other examples of industries which have gone through the same change. As with other industries, the new relationships are consumer to service provider and there exist multiple strategies for balancing benefits vs risks (including transitional risks such as confusion, security of supply, trust to outsourcing risks such as competitive pricing, lock-in, second sourcing etc). This is all standard supply chain management stuff and it has happened many times before.

Companies often combine both private generation with public consumption of utilities (not just limited to electricity) to mitigate some of the risks and balance the overall equation of benefits vs risks in their favour. This was highlighted by Douglas in 1966 and it is still the case to date (some 40+ years later).

It is fairly fanciful to suggest that "cloud" will not follow the same path of other industries and yet provide no basis or reasoning for this.

As for Nick blazing the trail in 2002, whilst it is true he became the poster boy for this change – it is not reasonable to assign either the concept of utility computing provision (Parkhill / MCcarthy 1966-68) or commoditisation, variability of what IT is and declining strategic value of IT (Strassmann 1994-98) to Nick. His original paper was an excellent piece and his subsequent books were great summaries for the field, but that's all.

I was going to post when I first saw this but held off. I am glad I did not now because Vanessa and James were much more eloquent and concise than I probably would have been on the topic.

The private generation of power makes a lot of sense understanding that the analogy has limits presently as it relates specifically to what is defined as "power".

Side note:

I have never agreed with Carr's approach because it seems to drastically oversimplify some differences between the two paradigms in order to make his point. To me that approach is a bit reckless in that one of our greatest challenges is keeping communication open and as clear as possible.

–David