10 techniques from Design for Manufacturing that could apply to Data Center Operations

Design for Manufacturing is an old concept.  The ideas of Design for Manufacturing have been applied to construction in the Lean Construction ideas.

Lean Design and Construction is a production management-based approach to project delivery -- a new way to design and build capital facilities. Lean production management has caused a revolution in manufacturing design, supply and assembly. Applied to project design and delivery, Lean changes the way work is done throughout the delivery process. Lean Construction extends from the objectives of a lean production system - maximize value and minimize waste - to specific techniques, and applies them in a new project delivery process. As a result:

Data Centers are some of the most expensive construction space, consuming a large amount of power and hosting equipment well over the cost of the building.  Over the life of a building many systems may be completely replaced which multiplies the ease or pain of install as you remove the old equipment and install the replacement.

Wouldn't it be great if data centers were designed for repair and replace type of operations.

Here is Audi rebuilding a Le Mans car in 15 minutes.

Here are Ten Guidelines for Design for Manufacturing that will give you an idea of the details that need to be addressed if you want to solve the operations issues.

1. Reduce the total number of parts.  The reduction of the number of parts in a product is probably the

best opportunity for reducing manufacturing costs. Less parts implies less purchases, inventory, handling,

processing time, development time, equipment, engineering time, assembly difficulty, service inspection,

testing, etc. In general, it reduces the level of intensity of all activities related to the product during its

entire life. A part that does not need to have relative motion with respect to other parts, does not have to

be made of a different material, or that would make the assembly or service of other parts extremely

difficult or impossible, is an excellent target for elimination. Some approaches to part-count reduction

are based on the use of one-piece structures and selection of manufacturing processes such as injection

molding, extrusion, precision castings, and powder metallurgy, among others.

2. Develop a modular design.  The use of modules in product design simplifies manufacturing activities

such as inspection, testing, assembly, purchasing, redesign, maintenance, service, and so on. One reason

is that modules add versatility to product update in the redesign process, help run tests before the final

assembly is put together, and allow the use of standard components to minimize product variations.

However, the connection can be a limiting factor when applying this rule.

3. Use of standard components.  Standard components are less expensive than custom-made items. The

high availability of these components reduces product lead times. Also, their reliability factors are well

ascertained. Furthermore, the use of standard components refers to the production pressure to the

supplier, relieving in part the manufacture’s concern of meeting production schedules.

4. Design parts to be multi-functional.  Multi-functional parts reduce the total number of parts in a

design, thus, obtaining the benefits given in rule 1. Some examples are a part to act as both an electric

conductor and as a structural member, or as a heat dissipating element and as a structural member. Also,

there can be elements that besides their principal function have guiding, aligning, or self-fixturing features

to facilitate assembly, and/or reflective surfaces to facilitate inspection, etc.

5. Design parts for multi-use.  In a manufacturing firm, different products can share parts that have been

designed for multi-use. These parts can have the same or different functions when used in different

products. In order to do this, it is necessary to identify the parts that are suitable for multi-use. For

example, all the parts used in the firm (purchased or made) can be sorted into two groups: the first

containing all the parts that are used commonly in all products. Then, part families are created by defining

categories of similar parts in each group. The goal is to minimize the number of categories, the variations

within the categories, and the number of design features within each variation. The result is a set of

standard part families from which multi-use parts are created. After organizing all the parts into part

families, the manufacturing processes are standardized for each part family. The production of a specific

part belonging to a given part family would follow the manufacturing routing that has been setup for its

family, skipping the operations that are not required for it. Furthermore, in design changes to existing

products and especially in new product designs, the standard multi-use components should be used.

6. Design for ease of fabrication.  Select the optimum combination between the material and fabrication

process to minimize the overall manufacturing cost. In general, final operations such as painting,

polishing, finish machining, etc. should be avoided. Excessive tolerance, surface-finish requirement, and

so on are commonly found problems that result in higher than necessary production cost.

7. Avoid separate fasteners.  The use of fasteners increases the cost of manufacturing a part due to the

handling and feeding operations that have to be performed. Besides the high cost of the equipment

required for them, these operations are not 100% successful, so they contribute to reducing the overall

manufacturing efficiency. In general, fasteners should be avoided and replaced, for example, by using tabs

or snap fits. If fasteners have to be used, then some guides should be followed for selecting them.

Minimize the number, size, and variation used; also, utilize standard components whenever possible.

Avoid screws that are too long, or too short, separate washers, tapped holes, and round heads and flatheads

(not good for vacuum pickup). Self-tapping and chamfered screws are preferred because they improve

placement success. Screws with vertical side heads should be selected vacuum pickup.

8. Minimize assembly directions.  All parts should be assembled from one direction. If possible, the best

way to add parts is from above, in a vertical direction, parallel to the gravitational direction (downward). In

this way, the effects of gravity help the assembly process, contrary to having to compensate for its effect

when other directions are chosen.

9. Maximize compliance.  Errors can occur during insertion operations due to variations in part dimensions

or on the accuracy of the positioning device used. This faulty behavior can cause damage to the part and/or

to the equipment. For this reason, it is necessary to include compliance in the part design and in the

assembly process. Examples of part built-in compliance features include tapers or chamfers and moderate

radius sizes to facilitate insertion, and nonfunctional external elements to help detect hidden features. For

the assembly process, selection of a rigid-base part, tactile sensing capabilities, and vision systems are

example of compliance. A simple solution is to use high-quality parts with designed-in-compliance, a

rigid-base part, and selective compliance in the assembly tool.

10. Minimize handling . Handling consists of positioning, orienting, and fixing a part or component. To

facilitate orientation, symmetrical parts should be used when ever possible. If it is not possible, then the

asymmetry must be exaggerated to avoid failures. Use external guiding features to help the orientation of

a part. The subsequent operations should be designed so that the orientation of the part is maintained.

Also, magazines, tube feeders, part strips, and so on, should be used to keep this orientation between

operations. Avoid using flexible parts - use slave circuit boards instead. If cables have to be used, then

include a dummy connector to plug the cable (robotic assembly) so that it can be located easily. When

designing the product, try to minimize the flow of material waste, parts, and so on, in the manufacturing

operation; also, take packaging into account, select appropriate and safe packaging for the product

The next big thing in data centers is low maintenance

Saving energy in the data center is old news.  The smart companies can easily achieve 1.2 PUE or lower.  They are being careful with their capital costs as well to build data centers.  More companies are looking at the data center holistically.

Google built its first data center in 2006.

Google is proud to call Oregon home to Google's first owned and operated data center.

We opened our data center at The Dalles in 2006—investing $600 million in the facility and establishing a long-term commitment to the region and state. Now a fully operational site, we've created over 80 full-time jobs on site, and we work hard to support the communities in which our employees live and work.

After 7 years of operation and opening another thirteen data centers, Google has plenty of data to make the next move to design data centers for low maintenance.  Google has made no announcement of this idea focusing on maintenance.  

I am making this statement because it is the logical progression in the life cycle of running efficient operations.

First phase is you want the facility to work.

Second phase you want to make sure it is available and performs.

Third phase is you drive efficiency and cost reduction while accounting for functionality, availability or performance.

The fourth phase is maintenance affects all these factors, but many times low maintenance was not designed into the data center.  You can only reduce maintenance costs so far.  Think of the per mile maintenance costs of your Ferrari vs. your Toyota.  Audi wins Le Mans because the cars are design for maintenance events.

With years of operation data, you can now design for operations and maintenance.  Low maintenance should use less resources which is greener and ideally lower cost.

Aiming at the wrong results, increase rack density vs. NOT stranding power

I was listening to a data center analyst and they made the point increase rack density as one of the top things to do to increase efficiency in the data center.

This is a clear target to hit, but it is the wrong one.  This situation reminds me of the 2004 Olympic target shooter who had one bullet to shoot to win the gold.  He hit the bulls eye in lane three.  He was in lane 2.

Emmons fired at the target in lane three while he was shooting in lane two. When no score appeared on the electronic scoring device for his lane, he turned to officials and gestured there was some sort of error.

"I shot," he appeared to say with a quizzical look as three officials in red blazers approached.

The officials went back and huddled briefly before announcing that Emmons had cross-fired — an extremely rare mistake in elite competition — and awarded him a score of zero.

It is easy to claim you increased the rack density, hitting the bulls eye.  But in the same way that Emmons lost because he shot the wrong target, there is a different target to aim at that would be more important if you have had judges.

If you deployed 2 kW in a rack, but also stranded power in the process why should you claim the increased rack density as a win?  The bigger picture to look at is did you use power without stranding power.  There are many other factors that influence where a piece of equipment should be placed.

Increase rack density is a great way for the Blade vendors to sell more blades.  But, the smartest data center operators don't make rack density a target to hit.

Do you?

That Olympian had instant feedback he missed the target.  Here is the story after and how he met his wife after the mistake.  

At Athens 2004, Matt Emmons missed his target but found love. He recalls:

"That was the last shooting event of the Games so a bunch of athletes and coaches went up to this beer garden between the ranges.

"We were taking it easy and relaxing, I'm there with some friends, and Czech shooter Katerina Kurkova [who had been commentating on the final as Emmons missed the target] came up to say how sorry she was about what happened, and how she admired how I'd handled the situation.

"At that time I just knew who she was, we'd never really spoken. But we hit it off really well, we started dating a year later and we were married in 2007. She's now Katerina Emmons."

Jeff Bezos invests in reinventing the Physics of the newspaper business buying Washington Post

Jeff Bezos has spent $250 mil of his own money to buy the Washington Post.

The LA Times discusses the possibility of Jeff Bezos creating a new business model for newspapers.

Washington Post buy: Can Jeff Bezos fix newspapers' business model?

Amazon's Jeff Bezos, buyer of the Washington Post, has shown success at experimentation and great patience about turning a profit.

I've worked on Publishing technologies way too long, starting in 1987 on displays, printers, fonts at Apple, then continued working on the publishing technologies at Microsoft.  Back in 1997 Bill Gates was focused on winning the battle for publishing vs. Apple.  Microsoft didn't win that battle.  Google changed the game sucking the air out of print advertising model.

Jeff Bezos in 1994 started Amazon.com and has seen the transition of Books and DVDs to digital. While this all going on Jeff has got an insider's look at the business models of media companies.  Jeff Bezos started college as a physicists and switched to computer science.

Bezos often showed intense scientific interests. He rigged an electric alarm to keep his younger siblings out of his room.^{[citation needed]} The family moved toMiami, Florida, where Bezos attended Miami Palmetto Senior High School. While in high school, he attended the Student Science Training Program at the University of Florida, receiving a Silver Knight Award in 1982.^[9] He was high school valedictorian.^[10] He attended Princeton University, with an intention to study physics, but soon returned to his love of computers and graduated summa cum laude, with a Bachelor of Science in Engineering in electrical engineering and computer science. While at Princeton, he was elected to the honor societies Phi Beta Kappa and Tau Beta Pi. He also served as the President of the Princeton chapter of the Students for the Exploration and Development of Space.^[11]

Jeff's love to understand how things work and make them better is what he did to retail.  Why can't he do the same with newspapers.  Once he figures out newspapers he can move on to other things that shape human perception.

Here is a progressive view of journalism.

So then what the hell is journalism?

It is a service. It is a service whose end, again, is an informed public. For my entrepreneurial journalism students, I give them a broad umbrella of a definition: Journalism helps communities organize their knowledge so they can better organize themselves.

Thus anything that reliably serves the end of an informed community is journalism. Anyone can help do that. The true journalist should want anyone to join the task. That, in the end, is why I wrote Public Parts: because I celebrate the value that rises from publicness, from the ability of anyone to share what he or she knows with everyone and the ethic that says sharing is a generous and social act and transparency should be the default for our institutions.

Is there a role for people to help in that process? Absolutely. I say that organizations can first help enable the flow and collection of information, which can now occur without them, by offering platforms for communities to share what they know. Next, I say that someone is often needed to add value to that process by:
* asking the questions that are not answered in the flow,
* verifying facts,
* debunking rumors,
* adding context, explanation, and background,
* providing functionality that enables sharing,
* organizing efforts to collaborate by communities, witnesses, experts.

40 Days of Dating goes Viral, Could 40 Weeks of a Data Center work too?

I was reading Om Malik's post on 40 days of dating.

40daysofdating: An awesome new kind of long-form story telling

by Om Malik

AUG. 4, 2013 - 12:16 PM PDT

3 Comments

• A
• A

SUMMARY:

40daysofdating is a website that combines text, photos and video to tell the story of two friends Jessica Walsh and Timothy Goodman who after failing at finding love, are dating each other and sharing the experience. It is like reality television, except for the web.

I've been waiting for when someone would have an interesting web site that tells a story.

Needless to say, we have barely scratched the surface. At some level, 40daysofdating is like reality television — reality web if you will — and a good signpost for what could be a more episodic approach to content and story telling. In the past, it was comic books and stories in noir magazines that kept you hooked. So why not web-based episodic story telling?

In the data center world, there could be 40 weeks of a data center.  But, it could be 40 years before the data center world would adopt story telling as a way to share its ideas. :-)