william gibson

Cloud Computing, Data Centers, and the Meat Locker Matrix

Today is Tuesday, May 18th, 2010. During the last two days (Sunday and Monday), two of our employees spent 40+ man hours in our data center. We use Paetec's data center (also known as a Co-Lo site ) in Fort Myers, Florida to house our internal business applications, our R&D infrastructure, our phone system and our Hosted ASP software application. Paetec provides the bunker, the security, the bandwidth, the power,  and the cooling. DataWorks provides the servers and software.

If you are looking for cloud computing  this is where it exists.

Cloud computing is not some vaporous thing that floats magically on the internet. It is racks and racks of noisy, hot, glowing  servers,  stacked one on top of the other in an enclosed silo of steel and cables. Our Co-Lo site is cold -- meat locker cold. You wear a warm coat when you are in the cloud.

99% of the time, the cloud  is an uninhabited space where the steel doors are locked, access is electronically keyed, servers stay hot, hard drives spin,  air conditioners whip away the heat, and the lights are off.  The cloud  is not a place for humans.

William Gibson, the novelist, (who has penned much of our computer slang) calls the walking-talking-awake-world the "meat" world. The meat world is the world we inhabit when we are not jacked into the "matrix" (Gibson penned that term too).

Our typical business activities are 98% matrix based  - very little muscle ("meat") is used in the pursuit of profits. The last two days these two Data-Workers efforts equated to an intense work out.  They are now physically exhausted - having stood on hardened concrete floors, their feet (especially their heels) feel like raw hamburger. Their hands ache. They have cuts and bruises  on their hands and arms from dead lifting 150 pound battery back up units out of service and replacing them with newer (and lighter) units.

By the way, they are in good physical shape, they hit the gym at least twice a week with personal trainers, they hike, they bike, they camp. In the course of two days, they used a host of low tech tools like wrenches, screw drives, pliers, and  pocket knifes  to lift racks, adjust rails, twist electrical leads, and cut cables.

After about an hour of this work their fingers go numb; after about 4 hours,  their minds start to go numb.  That's when the mental mistakes start to kick in.

In a moment of mental meat enlightenment they decided to move, re-rack and re-cable a group of servers - they moved them from one rack into another so they would be "logically" together. Makes sense. Makes sense now. Made sense then.  But in moving the servers, some mistakes were made. 1) Untested patch cables were used to re-cable the servers 2) cables were plugged into unmarked dead switch ports and 3) the primary firewall was re-connected with one cable in the wrong port.

It took  an additional 14 man hours and an additional 23 elapsed hours pass our scheduled down time to totally diagnosis and solve the cabling issues.

The obvious mistakes that were made: they changed more than one thing at a time (no way to trace the diagnosis  by working backwards -  that was working , but now its not working...) and not enough labeling and documentation of the servers, cables and ports. If they had taken an hour or two to label all the cables before moving the servers,  they would not have had the problems that they encountered after the boxes were spooled  back up.

In that perfect vision that is known as hind-sight, I thought about how military officers do not typically carry rifles or heavy packs. In the field, they are lightly equipped with just a side-arm. Why? So they are not physically burden, they do not become physically fatigued,  and their minds can remain fresh and agile. When your "meat" is on the line in combat, you want the pointy end of the kabob to be deadly sharp and accurate, and you most definitely want the thinking end of kabob  to be cool to the touch and ready to think through a crisis.

The cloud is were the matrix world meets the meat world - Bring a coat.

And don't forget the label gun.

Retail 101. Fewer Choices equal More Sales

Our local mega-movie-complex figured out many years ago that if they offered too many candy choices, they actually lowered their candy revenue.  What they probably learned in a Retail 101 class  (or a corporate manual) was that if you have too many choices, the customer takes longer to make a selection, the line moves slower, and because the movie start time is fixed, folks bounce out of line and head for their seats without making a purchase. I have noticed a similar problem at our local Subway franchise.  Folks  line up for their 6-inch meals during the lunch rush.  Subway newbies struggle with the menu matrix variables.  A programing language is spoken under the "Order Here" sign:  syntax needs to be in the proper order to get the sub built quickly.  Start with size.  Follow with sandwich type.  Delineate the bread selection.  Keep it moving,  one side step after another until you belly up to the cash register.  Get any of the code out of sequence and you will get an  onion operator mismatch or a division by pickle error.  If you get too many noobs  queued up, forget about the quick turn and burn, you are stuck in the thick of the sub-plot.   After a couple of long sessions of staring at the  potato chip rack,  I now come prepared with a trade magazine (Hospitality Upgrade and Wired are my popular periodicals)  or my current novel (large helpings of William Gibson have been consumed in the midst of the Subway sub-culture).

If you are a retail manager,  give this some thought:  at the cash wrap you can display a lot of snap item choices - but at what point are there too many choices? When are you creating counter clutter and slowing down the point of sale?  My aesthetics tell me that your counter should have a maximum of  five SKUs  to pick from.  Odd number of choices have more visual power  then even numbers - three is better than two, and  five is better than four. But more than five is just noise.

The same odd-versus-even thing works for product facings too.  Better to have a facing of three rather than two.  I would line up a facing of one rather than a facing of just two.  Call me crazy, but that is what going to art school does to your sense of  "what-looks-right".

I was trained as a fine artist during the 1970's - minimalism and conceptual art were among the vanguard.  I admit that I am jaded - I was trained to see, think and believe that "Less is More".  (For those who want to know more about those art movements, here is a sampling of artists who shaped the visual vocabulary of the 1970s: Sol LeWittChristo, John Baldessari, Donald Judd, Carl Andre and Dennis Oppenheim.)

Boiling your presentation down to the iPod-ian essence can result is an elegant, simple, and beautiful end product. Apple certainly has nailed down the clean minimalist style with their recent products.

The challenge we all have is this: in a world of many choices, how do we focus on what is important?   As a kid, the top of  my dresser was always crammed with every toy or object that was important to me:  plastic models,  coin bank, world globe,  jar of BB's,  comics, army men,  photo of Alan Shepard, spent Estes rocket engines, and Lego masterpiece - not a square inch of empty space.  At a certain density the pedestal-ed objects cease to function as a display. Objects break or topple when you go to pick them up, mom complains about the dust,  and the whole presentation looses its luster and appeal.

It is the goal of good design to filter out all the extra objects, features, requests, buttons, knobs, dials and other do-hickeys until you have a  product distilled  down to the fewest - and most important - choices.

What has never been published until now is this - NeXT® back in 1999 was actually much larger in scope. It had  additional sub-systems that dealt with watch-dog alerts, work-flow integration, n-levels of hierarchy and a separate financial ledger that were removed from the design scope because they put too many additional tasks into getting version 1 out the door.  Since I was the designer I can admit that I suffered from the second system syndrome . I knew about the tendency since I had read Fred Brook's masterpiece The  Mythical Man-Month, but figured since I knew about the follies of those who had proceeded me, I would be immune from the disease.

I was still putting too many things on top of the dresser.

About 20 months into the data design we realized that at the rate we were modeling, it might be another 24 months before we finished with the project scope. I think I may have even re-read a chapter or two of Brook's book at the time.  We evaluated the project and started hacking out entire modules and trimmed down the effort. 10 years later those features are now being introduced back into NeXT. All of those removed features will eventually show up in NeXT they just needed to be introduced in a  more managed context.

That decision to cut back was important because we got NeXT installed in 2002 versus 2005.  We are glad we got it out the door.  What is interesting is that many of the forms have now undergone a cleansing and scrubbing process where the number of tabs and controls have been reduced and cleaner, simpler hyperlinks have replaced them. We now make sure there is plenty of free space (negative space) that gives the controls some breathing room and more visual importance.

That's how fewer choices helped in our software design - I hope  you can use it with your own merchandise presentations.