The real-world end of Moore’s

Moore's Law, and Shugarts Law, and a pile of other IT laws which suggest that you get more bang for your buck with each passing year do not account for the reality of large enterprise IT.

If you slavishly follow their dictates, what you'd expect is that the price of IT should tend towards nothing over time. There should be so much processor, network and disk capacity available that the optimal decision is to use the largest amount possible in any situation to minimise the other key input to any particular system: people.

But the economics of this break down in a large enterprise, and here is the reason.

Firstly, IT budgets are expected to contract, year on year. That's because, in theory, the cost of all the hardware and stuff comes down. Its also because we are supposedly getting more efficient. The mantra is "do more with less", and by hook or by crook, most IT organisations manage to do so.

In the meantime, of course, demand for IT resources is increasing. It is increasing at a very, very rapid rate. I know we are in a constant rush to provision more storage, network and processor to keep up with the demand, usually remaining just one step ahead. And a key driver of this demand is end-user computing. Transaction volumes aren't going up so much, but all those end users building little applications that move pictures and videos, OLAP Cubes and large Excel data sets certainly know how to consume resources.

The resource requirements are increasing more quickly than the IT Laws are providing discounts. Couple that with shrinking IT budgets, and what you have is increasing internal scarcity. The demand doesn't just evaporate, however.

The gap between the price to satisfy all this demand, and the discount provided by the IT Laws are a pair of open jaws that are getting wider. And in the past, we've used our IT budgets to subside things to the point where no one noticed. IT budget shrinkage is knocking that on the head, now.

This is going to lead to a pretty big price hike for large corporate IT pretty soon. What else would you expect? Unsatisfied pent up demand for resource coupled with scarcity of supply always leads to price increases.

I know some readers will be thinking that Software as Service and various other new models are the answer to this problem. They're not though. All they do is shift the ever-expanding resource requirement somewhere else. Instead of needing more capacity in disk and processor, now you need it in network.

But there probably is an answer, and its one that's been staring us in the face for decades. Use local resources. Make use of the capabilities that are already in the technology on the desk. Peer to peer at the edge is fantastically more efficient than centralisation of everything.

I think the time is coming when we'll have to accept the end of centralised IT. Bigger is not necessarily better, and this may be one case where scale efficiencies simply don't apply.

Comments

[Magnus Lundgren]

Hi,

Seems you're describing a type of solution along the lines of what OpenSpan has to offer. That's basically all about reusing what you already have and tying it all together.

/Magnus

[John Burgess]

James,

Concerningly confused post from an Innovation Head!

Think you're confusing technological observed trends with IT cut justification - though given the current climate, can understand you looking for ammunition!

Moore's law simply states the rate of processor-advancement, it doesn't state anything on cost. Indeed most costs as you'll know are a two-way curve with cost coming down until hardware starts getting so old, it's increasingly expensive for companies to upgrade/replace.

In addition, the law doesn't account for any advancement in systems/technological demands, which as we know permanently rise to strain the current baseline build...

In all the focus on technology, maybe you've missed the point - it's the people and the technological capabilities they offer that need to be decentralised.

In most people's eyes, where the system itself sits is the the least of the worries...

[Hootnholler]

Another component you should consider is the burden of the legacy data/applications/content that the business requires IT to carry forward through advances of technology. This baggage train was accumulated with technologies that had a specific cost at a point in time in the past. The legacy imaging system scanned documents with a less efficient compression technology that is available today... but we have to keep them around for regulatory reasons and converting them will be expensive... The marketing department says that they "have" to keep the data on that direct mailing campaign they did 15 years ago "in case" they need it (I've never met a piece of data that a marketing department didn't like) and they want IT to shove all of the historical data into their brand new marketing system or they won't get value from the new product.

It is this slow accretion of corporate legacy and the difficulty and reluctance of the businesses to prune and groom it that causes it to become a sea anchor to overall IT efficiency.

[Colin Henderson]

I agree with hootnholler and would add one more component to that of increased end user usage. That is complexity - end users require more volume, and more complexity.

If we focus just on email, employee collaboration, and documentsfor a moment for individual groups within a large corp, I kind of agree ... if the tech the solution is the precise opposite of what is suggested - not going local. It has to be in the cloud that the answer lies - reasons:

1. outsource the complexity to others. Our small company has TCO of $100 per user per annum (Google Apps). No company support offerred, and believe me, not all are geeks in our ecosystem.
2. establish a way to abstract large corporate data from user functions. Outsource the user function part to "the cloud".
3. Definition of insanity - corporate run email system. The costs and difficulties associated are exorbitant, and .... why? What protection and security is really offerred?

Clearly this would offer new challenges, but the benefits would be interesting. What if a bank dept ran:

- gmail
- google docs
- postini (doc security & recovery)
- google calendar
- diigo for shared links
- friendfeed rooms (private) for discussion

Just thinking ....

[Neil Robinson]

The case for Software as a Service.

First, your way, James, without SaaS.

If you want to watch a football match, you build the stadium, you hire the teams, teach them to play and schedule a game in two years. Trouble is, your people then decide they want to watch tennis.

If you want a rock concert, you find someone to learn to play guitar, build the amplifiers and write some songs. Then you build a concert hall. By the time its ready, everyone's got bored and are into House, Rap and Garage.

If you want learn about cooking, you build a kitchen, get in some food and find a chef. but everyone's starved to death or gone vegie.

Now, my way, with SaaS.

I turn on my TV.

If I want a football match, I change to that channel. If I want to listen to a rock concert, I switch to that one. If I want to cook - you get the picture, literally.

What's more my football's premier league, my rock band are grammy winners, my cooking's got a Michelin star. Your's is a pub kickabout, your band are subway buskers and your food is McDonald's quality.

You see, TV costs a lot to provide, but I don't care, I only pay for what I want... Shall I go through that again for you?

Oh, by the way, I didn't need to buy a TV, we both had those. Only we call them browsers.

[James Gardner]

@John: your points are valid ones, but its hard to argue with the observational data available. I agree with the premise that people need to be decentralised, but I think there is an inflection point beyond which centralisation of the technology becomes uneconomic for most of us

@Neil: Your analogy collapses because you fail to consider that free to air television doesn't have a cost for transmission. Instead, a better analogy would be on-demand cable for 100k users in one place. I suspect the benefits you allude to evaporate at that point.

[Neil Robinson]

James, why worry about the SaaS provider's operational cost if they aren't passing them on to you?

With electricity, you buy per unit. Do you stay up at night worrying what it costs to run a power station?

Look let's look at specifics. I use Zimbra mail, delivered as a service. This has all the functionality of Exchange. It costs me per mailbox, £3.70 per month.

Microsoft provides a tool to ascertain the true cost of email delivery for Exchange 2007. Here's the link to it...

http://msexchangeteam.com/archive/2008/11/04/450039.aspx

To save you the time to calculate this, Novell have summerised it here:

http://www.novell.com/communities/node/2776/consultants-corner-special-edition-exchange-2007-facts-update

From this, you'll see that to deliver Exchange 2007 to 5000 corporate desktop users costs a staggering $1,116,692.70.

So an individual mailbox (that sum divided by 5000 and converted) costs around £170.00 per month.

Note: It's actually much higher, when redundancy and support are factored in. Energy requirements (expected to exceed basic software costs within the next 5 years) are not included here, either.

What would a company CFO say if you came to him with a choice between £170.00 and £3.70 per user or a total budgetory ask for 5000 users of £893,354.70 for Exchange or £18,500 for Zimbra. What do you think their answer would be?

Just to summerize. To take up SaaS you don't need any extra network as long as there is an existing Internet connection, so there are no extra delivery costs to be factored in. What you see is what you get!

James. if you or anyone reading this has the slightest interest in what your enterprise could be doing in the next two years, I beg you to get Nicolas Carr's book "The Big Switch" and read the first chapter. I guarantee, you will understand it all.

Microsoft knows this will happen. I will send you the memo Bill Gates sent in 2005 about this if you like...

[James Gardner]

@neil:

"you don't need any extra network as long as there is an internet connection"...

You have clearly been out of banks too long.

It is almost always about the network when you have to run thousands of branches.

[Neil Robinson]

Not quite as long as you think, James! ;-)

Surely you aren't suggesting that your branches are currently standalone. I would naturally assume they connect in some way to a corporate backend.

If the connection does exist, then you already have all you need at branch level to accommodate a SaaS model.

Centrally, if your Corporate Internet pipe (PoP) isn't up to the job, expanding it becomes a low cost, once-only action bringing many other benefits!