A possible green data center metric, carbon footprint of a Cloud VM

I was at a student panel discussion with Seattle Pacific University's M.A. for Social and Sustainable Management program. Rob Greenwood and I were discussing Environmental topics. Rob would discuss things at a broad industry, and I would discuss specifics within the data center industry and technology companies.

Part of what Rob expressed which aligns with things I observe as well is that many companies have difficulty showing the business value of sustainable efforts. In data centers with cost of power and cooling systems, improving PUE is a standard. Carbon footprint reporting is also gaining momentum. So the data center industry is different than the overall industry in that it is clearer when the efficiency pays off.

Examples are:

Google has been carbon neutral for 6 years.

Google has been carbon neutral for six years. This means that all of the work we do has a total carbon footprint of zero. This includes serving 100 billion search queries and 6 billion hours of YouTube videos a month, supporting 750 million Chrome users and delivering Google Maps in 194 countries. It also includes efforts such as developing self-driving cars and launching Wi-Fi-enabled balloons into the atmosphere.

Apple has one of the largest solar installations at its data center. And, Greenpeace has friended Facebook for its environmental efforts.

So how about this for a game changer. If users wanted more transparency from Cloud providers and required to know the power and carbon impact of a VM? It is not the standard now. This is easy for Google to say the carbon impact. It is zero. Currently no cloud provider makes this disclosure as they don’t want users to know how much power a VM uses. Why? Let’s say you are in AWS with a small VM. What would you think if your VM power use was 20 watts? 1/12 of a 240 watt server environment. Dual processor 6 core AMD with 1 VM per core. But, are some actually putting 2 VMs per core, and the power may be 10 watts per VM. You can see why no cloud provider wants to tell you how much power the VM uses.

Warning playing yourself in the DC is not something you usually promote

The following is good for a laugh, and a different way to see a press announcement.

One of my data center friends sent this link to a DatacenterDynamics post.

IO'S MODULES DEEMED MORE EFFICIENT THAN IO'S RAISED FLOOR

Year-long data collection effort finds 19% efficiency gain

20 August 2013 by Yevgeniy Sverdlik - DatacenterDynamics

 
IO's modules deemed more efficient than IO's raised floor
An IO.Anywhere module

IO, data center provider known primarily for its modular offering IO.Anywhere, has conducted a comparison study to see which design approach – modular or traditional raised floor – is more energy efficient.

 

The provider has both types of data center space at its Phoenix facility, which it used to collect power usage data over a period of one year. IO handed this data to Arizona Public Service (APS), the electrical utility serving the area, for third-party evaluation.

 

The results, confirmed by APS, showed a 19% reduction in overall energy use by the modular solution over raised floor. The average annual Power Usage Effectiveness (PUE) of the raised-floor environment was 1.73, while the modular environment's PUE was 1.41.

So, if play by yourself you win.  Or do you lose as you accuse yourself of cheating? You argue you are playing fair.  Watch this video when a person plays by yourself.

I wonder how many of you out there will follow the lead of IO to play with yourself to outperform your past.

NewImage

NewImage

NewImage

The press announcement says the new PUE saves 19% energy.

IO, a global leader in software-defined data centers, today announced that a third-party evaluation by Arizona Public Service (APS) showed that a modular data center technology installation in IO’s Phoenix facility has achieved 19% energy cost savings quantified by its improved Power Usage Effectiveness (PUE) ratings.

1.73 - 1.41 = .32 / 1.73 = 19%

Another way to calculate the numbers is .73 for overhead - .41 = .32 divided by .73 = 44% reduction in power used by electrical and mechanical overhead.

If you are going to play by yourself to show you can beat yourself, it sounds better that there is 44% reduction in electricity in the power and mechanical systems.

Who is crawling GreenM3 and some other metrics

I don't spend as much time as I used to on site analytics.  This morning though I was looking at what crawls www.greenm3.com and decided to post.

Over the last month there were 37,945 hits from robots

NewImage

Of the robots here is the breakdown of who and how many.

NewImage

Browsers

NewImage

OSs

NewImage

If I was making a significant amount of money on ad revenue related to impressions I would probably worry about traffic, but I don't.  I took a bit of a break from blogging so much, and have some new ideas to focus on.

NewImage

Driving Data Center Performance, focus on your team, not on others

At 7x24 Exchange a panel discussion with Google, AOL, and IBM discussed the challenges for Driving Data Center Performance.

NewImage

NewImage

Panel Discussion

Moderator:
David Schirmacher
Senior Vice President of Operations
Digital Realty
and
President
7x24 Exchange International

Panelists:
Joe Kava
Vice President
Google Data Center Operations

Todd Traver, CDCDP, PMP
Data Center Strategy and Energy Efficiency Executive 
IBM Global Technology Services 

Michael J. Manos
Chief Technology Officer
AOL

You those of you who haven't seen Mike and Joe on stage, they both have lots of energy.  Todd was also chiming in.

NewImage

NewImage

NewImage

It is difficult to write about all the ideas discussed.  Here are nuggets made during the session.

Joe - Dashboards allows visibility within Google to see operations including executives.

Mike - PUE as a metric has hit the diminishing return point.  The next issue is work per watt.

Todd - integration across the organizations is the next issue to address.

Manos -  It costs 30% overhead to develop the software infrastructure for redundancies and operations. but it pays long term. 

Overall message on measurement - wakeup create your own metrics for your team.

Kava - greatest cause of unavailability are the choices made to make changes, not failure of the electrical or mechanical systems.

There were many more things discussed and this blog post would get way too long if I tried to cover all the ideas.

FYI, I am biased on this blog post, because Mike Manos and Joe Kava are friends that I moderated 1 1/2 yrs ago at 7x24 Exchange, have known many years, and we had plenty of time in the bar and chatting in side conversations during the conference.  Oh and the moderator David Schirmacher is also a good friend.  So I am able to interpret and think more about what is presented, yet I need to go back to watch the video at some point to see more of what I missed. For those of you not at 7x24 Exchange Boca, you'll be able to watch the video to see what you missed too.

Below is a picture with Jack Glass, Mike Manos, and Joe Kava from the conference 1 1/2 years ago. 

NewImage

Convention Photography are courtesy of Professional Images Photography Joe Rodriguez 2013. 

 

Surprise Surprise, a networked based energy solution gets acquired by Cisco

Two years ago I first had a chat with folks at Joulex on their energy efficiency solution.  Their solution is an agentless approach that inventories the IP addresses of all devices on the network to identify the power uses.  There are very few devices running in a data center that don't have an IP address.

And, Cisco has announced intent to acquire Joulex.

JouleX’s energy management solution, together with Cisco EnergyWise™, will provide customers with a simple way to measure, monitor and manage energy usage for network and IT systems across the enterprise, without the use of device-side agents, hardware meters or network configurations. JouleX’s software helps to reduce energy costs by monitoring, analyzing and managing energy usage of all network-connected devices and systems through a set of policies derived through analytics tailored for an enterprise’s needs.

I haven't run into many who have used Joulex in a data center, but with Cisco's acquisition there could be a new push into the data center market.