Lazy Energy Elephants

On the face of it, energy savings is simple. 

Conserve. 

Turn things off. 

Increase utilization.

Plug in less.

Real lasting success comes from your strategy.  You know, the one you have carefully planned.   You've looked at each part of the complete data center eco-system and found ways to reduce. 

Like a diet, right?  Research. Resolve. Reduce. Keep the load off.

Then relax.

Just tie objectives-to-plans-to-actions-to-measurements-to-success. 

On the energy front, the most successful companies create a strategy that includes an energy budget: planned, monitored, measured and revisited frequently. 

But I submit that there's an elephant in the room. 

Maybe invisible, but an elephant nonetheless.

And it is a very lazy elephant.   

In the typical data center, the facility itself consumes the largest amount of energy.  That's the first place we look.   

For IT equipment, the first best energy improvements come from improved utilization.  That really means make some parts work harder and shut the rest down. 

Chuck it.

Virtualization. Consolidation.  Process management.

Then, acquire anything new with a certainty that it is more efficient hardware and choose software like deduplication or automated archiving to improve your energy outlook. 

So what about the elephant?  That lazy elephant?

The elephant:  Application code 
A storage architect for a major bank recently described his experience with lazy application code. 

The bank looked at one particularly important and huge but inefficient application and then rewrote it.

The old version took 17 hours on 400 machines for its job to run. 

The rewritten version reduced the time to 20 minutes on 60 machines.

That means they reduced the number of machines to just 15% of the old number and cut the processing time by to just 2%! 

No, I am not aware of any independent validation of his claims.  But even if he was bragging a bit for dramatic effect, these differences are huge.

Even if you cut all his numbers by  half.  Wouldn't you be very happy with that?  Suppose you could cut your application processing time and machine count in half, wouldn't that be worth the trouble?

What called his attention to the issue was some combination of the energy crunch in his data center and a recognition that "hardware is getting faster more slowly than software is getting slower."

But it isn't just this particular application team, but everyone writing code that needs to understand how the efficiency of code has some fairly dramatic impacts that they probably never think about. 

Now rewriting applications can be a huge undertaking.  Seems it can have huge payback too.  Because the real benefit to his bank wasn't the time and machine reduction.  It was the competitive advantage.gained through dramatic improvements in code efficiency.

That's a three-fer.

Any more elephants?

A colleague at a partner companies recently pointed out another elephant that I am looking at now.   (A Zoo seems an easy fit here, doesn't it?)

This one has to do with how, in the midst of a data center building boom, good design is being ignored.   Seems I was incorrect to assume that, given the times, any new data center would incorporate all we have learned about best practices.

Doing it right.

Building to last.

Guess not.