“Peak Oil” and the death of the Data Warehouse

Multi-tasking. Some of us do it better than others. And unfortunately, I fall into the "others" category – so much so that I now regularly close my e-mail client whilst I'm working on a document, as I have come to understand that it takes me three times longer to complete a task when a "foreground" activity has to compete for my attention with a "background" workload.

Those of you who are struggling to empathise with me right now are highly-likely to have one thing in common: your gender. Women, you see, have been proven to have superior multi-tasking abilities to men. It turns out that there is a scientific basis for my wife's ability to win the domestic disputes that we carry-on whilst I am driving and she is navigating. Either that, or she is just much smarter than me. But to protect my fragile male ego, let's stay with the former explanation!

Multi-tasking is, of course, very important in a multi-user computing system. If we can't share computing resources equitably, then users are highly likely to demand their own, dedicated computing systems. And Enterprise Data Warehousing – bringing multiple applications and users to one copy of the data, rather than the other way around – is almost the prototypical multi-user computing application. A "Data Warehouse platform" that doesn't have very robust mixed-workload management capabilities isn't a "Data Warehouse platform" at all – it's a "scalable Data Mart platform". Now there is a place in very many organizations for one or two Data Marts - which is why we introduced a scalable Data Mart platform of our own several years ago and, more recently, a middleware framework ("Unity") that increasingly aims to tie together multiple analytic computing platforms (more on that another time). But let's not be confused about the market segment that most of our "competitors" are really targeting with their products, optimized as they are for high-performance, low-concurrency table scanning.

Anyway, I digress. Some of those self same competitors have managed to persuade some vocal analysts who-really-should-know-better that the lack of this robust mixed-workload management capability in their products doesn't matter, "because, the Enterprise Data Warehouse model is dead" – and the future is multiple, overlapping redundant Data Marts.

This is a completely wrong-headed analysis for all sorts of reasons – not the least of which is the intentionally lunatic definition of "Enterprise Data Warehouse" that it relies upon - but so that I can get back to concentrating on the e-mail argument that I am currently conducting with my wife about my latest expensive luggage purchase, let's consider just one.

Most experts now think that we are at or past "peak oil", meaning that we will be able to extract less oil and gas from under the ground and under the sea year-on-year in future than we have been able to in the recent past. The mathematics of all of this is complicated by the fact that the current very high price of hydrocarbons means that the extraction of some formerly economically unviable reserves is now feasible; but equally, the rise of the BRICs means that supply and demand are now very tightly balanced. We are decades away from serious supply shortages, but equally there is very little prospect of any sustained relief from high oil prices. Sell your gas-guzzler!

Perhaps that doesn't matter, I hear you say. As George W Bush so memorably put it in his 2006 State of the Union Address, we are all of us "addicted to oil" – and for the sake both of energy security policy and the environment, oil is a habit that we need to kick. Renewable energy to the rescue!

Well, maybe.

Last time I checked, the European Carbon Trading Scheme price was $8 per metric tonne of CO2. That means that a European polluter has to buy $8 worth of carbon credits to release a metric tonne of CO2 into the atmosphere. The estimated carbon price required to stabilize atmospheric CO2 levels at 450 parts-per-million – and so ensure an average global temperature rise this century of "only" 2 degrees Celsius – is $40 / tonne today, rising to $80 / tonne by 2050. And the carbon price required to make offshore wind generation – the UK government's preferred method of renewable energy generation – profitable without subsidy? That would be $136 / tonne.

So between expanding demand for dwindling supplies of fossel fuels and the high cost of renewable energy, the only thing more certain than the fact that I am completely wasting my time trying to persuade my wife that my new bag really does represent excellent value for money is that we will all be paying (much) more for our energy in the future.

What on Earth (pun intended) has all of that got to do with Enterprise computing?

Well, the 3-year cost of powering and cooling an Intel standard high volume (SHV) server in many geographies already exceeds the acquisition cost of that server. And with the upwards trajectory in the price of energy, "Green IT" is no longer merely a question of Corporate responsibility – it's an increasingly urgent issue for the CFO.

Now consider that the US Department of Energy estimates that average server CPU utilization is at or below 20%. That's right, 20%. Those multiple, overlapping, redundant Data Marts running on multiple, redundant servers - each merrily drawing full power for 24 hours a day, 365 days a year - are each only 20% utilized at best.

If the environmental argument against this lunacy doesn't move you, then the economic one should. One – or close to one – copy of the data, shared by multiple applications and users and maintained on a server running at 95% plus utilization is very clearly a much, much better deal than the alternative.

Providing, of course, that you have the robust mixed-workload management sub-system required to make sharing possible, so that those multiple applications - each with a different Service Level Goal (SLG) - can be supported from one analytical database platform, without relatively long-running analytics trashing the production service supporting frontline knowledge workers. Teradata, of course, has just such a sophisticated mixed-workload management sub-system – and if you want to know how it has just got even better with the release of Teradata 14.0, then you should come and see Teradata CTO, Stephen Brobst, present on our European tour this year. Visit www.teradataemea.com for details - we look forward to seeing you soon!

Martin Willcox
Director of Platform & Solutions Marketing, Teradata (EMEA)