How Green Is Your Data Center?

Ideas and trends come and go on the IT landscape.  Some stick, some don't.

Usually the trends are measured in quarters and years, and not just short months.

What's really caught me by surprise is just how much visceral momentum has built around the need for green data centers.  And just how quickly it's gone from vision to mandate.

It's a hot topic (pardon the pun) in almost every conversation with large customers.  And -- to be honest -- it's caught me a bit by surprise in terms of intensity and urgency.

I'd like to use this post to look at some of the reasons why, and -- of course -- what people are doing about it.

Way Back When ...

I remember a distant era, long long ago, where we actually marketed computer products based on environmental factors: footprint, power consumption, cooling, and so on. 

It was a hard way to differentiate yourself -- showing a three or five year spreadsheet about how much money you'd save vs. alternatives. 

Hardly emotional, visceral stuff.   But what is old is new these days.

The First Wave

I noticed the first wave when energy costs shot up significantly a few years ago.  When gas prices spiked, so did other related energy costs.  It wasn't gradual and insidious -- people noticed the rapid delta.

A few foresighted customers started asking about power and cooling issues. 

We had a few answers at the time, but -- seriously -- we hadn't taken the whole issue all that seriously.  At the time, we weren't fully aware that a fundamental and precipitous shift had begun in data center concerns. 

But we put a few engineers on the task to begin looking at alternatives.

The Second Wave

There was no defining moment, but I noticed at the end of 2005, the discussion evolved a bit.  Data center planners were looking at 2007 and 2008 and were facing a harsh reality.

They were simply out of power and cooling.

It wasn't about paying higher rates -- they were facing major data center rebuilds simply because there was no feasible way to meet the new energy envelopes they were facing.

On the one graph was the power and cooling growth estimates for the data center. 

On the other graph was the timeline to build a new data center, move everything, and spend tens of millions of dollars in the process.

The two graphs weren't lining up.  There was no way that they could escape the trap.

The implication was no new IT projects or equipment -- period. 

Why?  No place to plug it in.

As a result, things got a bit more urgent in the focus on environmentals back at EMC.  Customers had an entirely new perspective on the problem, so that meant that we had to as well. 

The Third Wave

This last wave has hit me hardest in just the last few months. 

Given increased concerns around global warming and "energy citizenship", many large companies want to be able to show that they're taking active steps as good corporate citizens.

They have a real need to show the world that they're making significant, measurable investments to reduce energy consumption, and -- not surprisingly -- large data centers are attractive targets.

I noticed it first from European-based large global customers, but it's spread rapidly to US-based companies, as well as more than a few from the far east.

Energy consumption is now a corporate responsibility issue.

And now there's a whole new sense of urgency in the mix.  People outside of IT are now involved.

And I don't think it's going to fade anytime soon.

What To Do

I think we all know what the basic recipe looks like here.

First obvious target: use server virtualization and current energy-friendly server technology to attack the biggest offender: poorly utilized, energy inefficient server farms.

A few barriers here that are worth mentioning.

First, the common practice of three-year depreciation schedules (or five, or sometimes seven!) have got to go.  Server technology three years ago and server technology today is night and day. 

My politcally incorrect suggestion on this one is to have your corporate responsibility folks go beat up on the finance guys on this one.  Stay a safe distance away, though. 

There's no way to meet this new mandate unless IT has the ability to put in current technology.  The results will simply not meet the corporate objective unless the rules change.

Second, there's only so far you can get with server virtualization in attacking more demanding applications before the information infrastructure around it has to change: storage, storage networking, management, backup and restore, replication and more. 

I've written a post on this subject, which probably needs to be updated a bit ....

But -- hidden in this new requirement to address infrastructure is the opportunity to go after the next big energy offender, and that's storage ...

Second obvious target: storage. 

Yes, that round, brown spinning stuff consumes its fair share of energy, right after the servers.

But here, I think the approach is different.  Yes, technology can play a role, but I think that process can make bigger impacts.

As an example, for a few years, we've been offering a suite of storage services to help customers rationalize storage expenditures.  The focus had been on capital expenditures and operational efficiency, but the same techniques have shown the opportunity for subtantial energy savings as well.

Let's start with tiering.  A modern 500GB drive is more than twice as energy efficient as the 250GB drive it replaces -- if you can get away with not impacting service levels.   

I don't want to even think about those old 36GB drives you might be running.  You are using the newer drives, aren't you?

For quite a while, we've offered the classification services that help folks understand what portion of their workload can use these more capacious drives.  And the savings are substantial and quantifiable -- not only in terms of capital, but energy footprint.

And then there's the overprovisioning problem. 

Why?  Basic storage management is still relatively primitive in many environments.

I used to tell the "chain of lies" example to illustrate the problem.

• Business guy says 100GB to application guy who doubles it to 200GB.
• Application guy says 200GB to database guy who doubles it to 400GB.
• Database guy says 400GB to server guy who doubles it to 800GB.
• Server guy says 800GB to storage guy who doubles it to 1.6 TB.

And the business guy ends up only needing 10GB, rather than 100GB.  And terabytes go unused.

Simply put, ITIL-based end-to-end storage management processes can take a mind-boggling bite out of storage requirements if applied properly.  And the result is largely independent of who's storage you're using.

So if you're not feeling that you're making the grade here, time to invest in either a services engagement, or an on-site storage management resident.  The old justifcation was capital savings -- the new justification is energy savings.

Then there's active archiving -- using specific tools to pull inactive stuff out of email environments, files systems and databases, and put it on low-cost -- and energy efficient -- media.  Again, what we once justified in terms of capital expense, we now justify in terms of energy footprint.

If you're more aggressive, there's data deduplication for backup and recovery.  We've been justifying Avamar on network savings and media savings, but -- at the same time -- through this lense there's considerable energy savings as well.

There are those that would tell you that storage virtualization is a means to an end here. 

Use storage virtualization and you'll end up using less storage.  That's the premise, anyway. 

I have to disagree.

If using less storage is your goal, there are far more practical ways of getting after that result than adding another piece of (energy consuming) equipment, or -- worse -- keeping your old energy-inefficient storage on the floor.

So what are the barriers here? 

Well, the same arcane depreciation schedules that are slowing server rationalization are hurting storage rationalization as well.  You're not going to get to big wins anytime if you're facing three year (or five year, or seven year!) depreciation schedules.

Sorry to say, but three years is a long time in disk technology, as well as the controller technology that supports them.  Especially when one considers environmental footprints.

I also think that many organizations find it hard to come by the skills needed for an effective storage rationalization project.  EMC and other consulting organizations can help with this.

The third target: networks.  Yes, networks.

I do a lot of work with Cisco.  They too have a good focus on environmentally friendly computing. 

And although their impact probably comes in behind servers, and perhaps storage, it's a significant contribution in the landscape, and shouldn't be ignored.

They were showing me the before-and-after numbers for the overall energy take-out around a network rationalization project.  Between newer, energy-efficient products, denser pipes, and so on, it was quite an impact.

I think one of the challenges is that a fair amount of network equipment is outside the data center, thus a bit outside the radar screen.  But, taken together through the lens of "corporate energy citizenship", it's all part of the equation.

Network guys face the same problem the server and the storage guys face.  Long depreciation schedules.  Complicated consolidation and migration plans.  It's the same general story with a slightly different set of issues.

A Few Suggestions

First, ask yourself -- do we see this coming at us?   

Are our power bills increasing over time?  Will we run out of data center energy envelope anytime soon?  And -- will our company need to make a statement showing good energy citizenship?

Second, are we taking an integrated approach?

Do we have a combined initiative to look at servers, storage and networks together?  Or are we attacking the problem piecemeal, with more effort and less results?

And -- most importantly -- have we reached outside of IT for support?  Have we engaged the PR guys, the corporate community folks, maybe the board of directors on this issue?

But there's a hidden opportunity here as well that I think we should all appreciate.

Wouldn't we all like to have dense racks of energy efficient servers, fully virtualized?  Attached to cost-effective, well-managed tiered storage?  And all connected throughout the enterprise through consolidated, well-managed networked pipes?

Maybe there's a silver lining in this particular cloud ...

Comments

[Tarry Singh]

Very rightly said, Chuck! And an excellent post. I did (sometime back) muse about CSR myself (http://tarrysingh.blogspot.com/2006/12/virtualization-where-is-your-csr.html).

[Chuck Hollis]

Hi Tarry

Went back and looked at your piece -- very well put, so thanks!

I don't think that either of us are the first (or the last!) to write on this growingly important topic.

Thanks for commenting, and am enjoying reading your blog!

[Josh Maher]

Are there any plans for EMC to create perscriptive guidance around how to implement storage or information management in an ITIL (CobiT, MOF, ISO) fashion? I have seen a you talk about EMC's services and other internal folks talk about EMC's services, but no guidance for customers to do this on their own. I understand services make EMC money, but sometimes it makes more sense to use internal resources.

[Jeff Oliver]

Chuck,

Great article. This encapsulates many issues we currently face.

I manage the corporate data center at a large retail company. I'm curious if and when you see these issues impacting non-technology companies? Perhaps a better question is when do you think non-technologies companies will start noticing these issues? There is no doubt in my mind that virtualization, server efficiency, depreciation schedules, and storage management practices affect us. But when I raise any of these issues, I tend to get blank stares from my colleagues.

Thanks again,

[Chuck Hollis]

Hi Josh -- good question.

The answer is "somewhat". EMC offers pretty extensive education on how to design, implement, etc. storage and information projects. Yes, you have to pay for the classes, but it's an alternative to the services.

The framework discussion presents a challenge. There are many to choose from, and most seem to focus on operations and change management, rather than design and implementation. I am not an expert in this regard.

That being said, I think EMC could offer more "self help" guides that allow a proficient practitioner to do it themselves.

I was heartened to see a string of solution guides (blueprints) for things like Oracle on IP storage, storage considerations for Exchange, and the like.

My understanding is that more are under development for VMware, SAP and additional environments.

These don't directly address your suggestion, but it's getting closer.

Thanks for the comment, Josh

[Chuck Hollis]

Hi Jeff

Where I'm seeing the most dramatic take-up for "green data centers" is in larger companies that take their social agenda (and its appearances) very seriously.

The usual arguments were always there about power savings, etc. but never really got acted on until the issue landed on the corporate agenda.

Now, I don't know how large your company is, or how important social optics might be, but my guess is that within a few years, you'll see it becoming all the rage. Especially when your competitors start doing it.

Until then, you're probably better off advancing the idea, but not trying to break too much glass. All good ideas need a place and time to be realized, no?

Thanks for reading, and thanks for the comment!

[Bill Bonin]

Chuck

Great blog today on energy savings. I think your best one. Energy is a case of high principals meets economic pragmatism with a big serving of PR benefit. This usually drives behavior.

I have one little quibble. Upgrades and lease rollovers are two of the biggest uses for virtualization. If IT shops go to an 18 month upgrade cycle without virtualization, they will be working a whole lot of nights and weekends.

I'll be interested to see the feedback you get. I'll bet it's a record.

[Chris M Evans]

Chuck, I talked about this subject last May (http://storagearchitect.blogspot.com/2006/05/green-datacentre.html) specifically at the time around SAN switches. I wanted to understand the efficiency based on bandwidth as a per port measured solution seemed unreasonable. Based on my numbers, why would anyone install anything other than Brocade as the switch product? They are hugely more efficient than Cisco for instance. Well, I think it is because environmental cost is just another factor in the mix of making a purchasing decision. I don't think there will be a wholesale rip-out of technology to be more green, however I do think there will be pressure to show you have chosen the most green product as long as it offers the feature set, reliability and value for money you are looking for.

[Chuck Hollis]

Hi Chris

My suspicion -- based on the numbers I've seen -- is that the SAN devices are really just a small portion of a much larger landscape of servers and storage.

Perhaps your argument makes sense in a standalone fashion, but -- consider a situation where someone doing aggressive server virtualization for a part of their environments wants to use a mix of iSCSI and FC.

The potential energy savings from the use of iSCSI in a virtualization context could easily outweigh the energy differential from choosing Brocade over Cisco.

Maybe that's not the greatest example, but I think the recipe here is to look at all the components together, as the interactions may outweigh the individual elements.

As a preliminary step, EMC has updated all their power calculators to include servers and networking gear in addition to storage.

I think we need to take a further step and show the idealized power-efficient architectures end-to-end: servers, networks and storage. We're not entirely there yet.

Thanks for the comment!

[Jen]

Hi Chuck,
I am just learning about this issue and have a question: Could you please clarify the term, "energy envelope." Thank you!

[Chuck Hollis]

Hi Jen

I hear the term used when customers are evaluating the entire energy footprint of a technology component, e.g. not just the power consumption but cooling requirements as well.

Going a bit further, over time there seem to be wider differences between normal power usage, and peak power usage for more and more technology components. More sophisticated models take both into account as well.

Thanks for your question ...