Why Amazon Hired a Car Mechanic to Run Its Cloud Empire | Hacker News Reader

Why Amazon Hired a Car Mechanic to Run Its Cloud Empire

wired.com

145 pointssk2code45 comments

It's kind of frustrating that otherwise interesting articles like this get buried under misleading titles like this, with a guy who some 30 years once worked on cars (high-end, cutting edge cars - not changing oil and brake pads on VW Jetta's) now working for a company that didn't even exist when he was a mechanic being described as 'Amazon hired car mechanic'. Michael Dell used to stock shelves in his youth, should newspapers run headlines like 'shelf stocker taking computer manufacturer private in 25B leveraged buyout'? Please.

</grump_mode>

Agreed. I was looking for something particular about why they hired a mechanic, since that's what it suggests in the title.

Probably because designing and running a big DC crosses discipline boundrys that a pure CS guy would not have the skills.

As Seymore Cray said he was more a plumber that a computer guy - well strictly speaking cooling HPC's is more fluid dynamics rather than a domestic plumber.

I used to be a mechanic, so maybe I can shed some insight. My work experience was very varied, but I mostly focused on luxury(Porsche, BMW) cars towards the last years or so.

Being a mechanic gives you a different perspective on software. You don't see code ad code, but as an engine. If the engine is not tuned properly, the engine will not be as efficient as possible. If there is too much cruft in the code, the engine will perform poorly. Troubleshooting is also easier, I think. Due to the fact that all that mechanics due is troubleshoot. I once spent a week chasing an electrical bug on a Porsche. For those curious, it was the car's computer going bad (a circuit connection somehow broke under stress and it shorted). It also makes design a bit easier, because I have seen many different design patterns IRL. When you can touch and feel a design, working with them is much easier. For example, the camshaft is a function that makes valves open and close. The camshaft lobes are the parameters being passed as data to the camshaft function. The parameters tell the camshaft function how far should valves open, and for how much time.

It also removes any fear from working with any system. I've seen and fixed it all. There is no way any codebase will turn out to be harder to fix than changing the spark plugs on a 1996 Porsche 911 Turbo. No way.

> "There is no way any codebase will turn out to be harder to fix than changing the spark plugs on a 1996 Porsche 911 Turbo. No way."

Famous last words

I have to agree, the spark plug task is pretty deterministic in that a shop manual will specify the number of hours to charge, which a well-practiced mechanic can easily beat and often pocket his share of the difference.

Question, and I'm not being pedantic: Have you ever changed the spark plugs on a 1996 Porsche 911 Turbo?

I have not done so myself, but I always imagined that it would require dropping out the engine. As such I'd call it a tedious but well-defined task for someone with a lift and plenty of time. If the job can actually be done without removing the engine, yes that sounds incredibly frustrating.

The 993 series in general is still a fantastic car today but the turbo model in particular looked absolutely stunning when it was introduced. They gave up a lot of character when they re-tooled in 1999.

I agree. The 996 was just too bland, but still sold much more units than the 993. Go figure.

OT: I once tried to talk a client into dropping a 3.6 993 engine into a 996 Carrera. Everything lines up, but he was afraid of the resale value of the car. Would have been a fun experiment. :)

Does it involve dropping the engine?

I can barely see the plugs on my non-turbo 911SC.

I imagine that the plugs for a turbo are buried under some gnarly manifolds.

There are two ways to do it:

You can remove the manifolds (and turbochargers and all related shit), or you can lower the engine a bit and get your hand in there with a collection of flexible 3/8 ratchet joints, a shaved spark plug socket (if its too long it wont work), and lots of patience. The issue is that the upper head faces the frame rails.

By the way, I love the SC. Very nice car. I dropped a 930 engine into one, but it keeps breaking transmissions (600+ HP). :)

http://p-car.com/diy/sparkplug/

I've done the change =)

Reminds me of a friend who took the hole saw to the inner guards on his STi to make getting to the plugs easier.

Oh yeah, I own a Subaru. I know the pain of changing those spark plugs, too.

Unless you've got a lot of experience changing spark plugs on a 1996 Porsche 911 Turbo, I think that statement applies to both you and the GP. It's reassuring to think our problems are harder than those of other professions, but it's not always true.

Oh, I get his point, and somewhat agree. But I think fixing cars is harder than software. Personal opinion of course.

That may be because software is your profession, and you spend more time doing it.

If it were between fixing an unknown issue with said car and fixing some code, then I would say it might be a wash. Fixing that particular issue with the car though? Regardless of the difficulty in fixing it, at least you've got a book that tells you what to do (I assume. I would be very surprised if that wasn't the case. From the sound of it, it looks like he knows what needs to be done). Diagnosis and design of a fix will almost always be more challenging them implementing a fix, be it with cars or code.

In auto repair, what the books tell you != apply to real life. Never does. I think its a joke played on by the auto manufacturers.

I'm still going with merely knowing that your job is to replace the sparkplugs gives you a leg up. You seem to have a pretty firm grasp on how complicated the task is and what is involved, which is already a head start.

In most cases no. You do have an idea of what you have to do (as in software), but nothing too exact. Worse when its an European brand that pleases itself in over-engineering everything.

Well as I said, "If it were between fixing an unknown issue with said car and fixing some code, then I would say it might be a wash."

If we're comparing an unknown with an unknown, that seems reasonable, but comparing a known (replacing spark-plugs on that particular car) with an unknown seems foolish, no matter how physically difficult that known may be.

If it takes longer than a week to change a spark plug on a 1996 Porsche 911 Turbo, I'd be shocked. If it takes longer than a month, I doubt anyone would ever bother. A year? Unimaginable.

For software, on the other hand, I have seen all three.

I've done both and while it's a bit of an exaggeration, changing those spark plugs were a giant pain in the ass. I still have the scars.

My new benchmark for self-inflicted pain is trying to get IPSec up and running from an iPad to the house going through dd-wrt terminating in a Linux VM. Still pounding my head on that one.

I'll see your luxury carpart and raise you one AbstractAbstractWidgetFactoryBuilderFactory.

Good point. Though I then raise you one 2002 Nissan Pathfinder V6 with a broken thermostat housing. You basically disassemble the uppder part of the engine (minus the heads) to change a stupid little part.

So you're saying that as long as I'm taking the heads off my 2001 Xterra, I may as well replace the thermostat while I'm at it ;-) ISTR it's the same engine.

Definitely do so. Along with all the cooling system hoses. Problem is those hoses are expensive, and will add about $200 to your parts tab. Worth it, though. Wait, I would actually replace the actual thermostat housing due to corrosion. Old coolant is a bad mother on your cooling system. Make sure to also flush out the radiator with running water.

Oh, and since these are aluminum parts, go to a Honda auto parts dealer and ask for a tube of Hondabond. Its a silicone dressing that is 100 times better than anything available from Pep Boys. You want to make sure your parts seal properly.

I think you are slightly confused as to what the other commenters are saying. You are assuming that they are saying that working with cars is irrelevant and that's not the case. I think we can all understand that outside experiences can aid in building software. I personally feel that all high schoolers should take industrial arts classes before graduating, whether it's auto shop, metal shop, or wood shop.

That said, the original posters were complaining that the headline is misleading. Jeff Bezos wasn't sitting at the BMW dealership waiting to pick up his car and found some diamond in the rough that could handle his AWS business. He found a candidate who was highly experienced with the right knowledge base to do the job and that candidate just happened to be a mechanic decades ago.

Orwell's famous quote needs an update:

In a time of universal link-bait and ironic hipsters, being genuine is a revolutionary act.

Every cruiser between Ketchikan and Bellingham has a copy of the Hamiltons' book on board: http://www.amazon.com/Waggoner-Cruising-Guides-Secret-Coast/...

I didn't know he had a day job. A true renaissance man.

Edit: I don't want to geek out too much over his boat, but Nordhavns are awesome. Comfortable, tremendous amounts of internal space, long range, and designed for real cruising. Here's a magazine article on the Hamilton boat [pdf]. The photos of the engineering spaces will make you cry. http://www.mvdirona.com/TechnicalArticles/TheHamiltonNordhav...

Yea there's only one problem with them. They're not sailboats :)

Car Mechanic? He was a Lead Architect for DB2, worked in various Lead/Architect roles on Microsoft SQL Server since SQL server 7.0, etc http://www.mvdirona.com/jrh/Work/default.htm

From 1980-1982 he worked as a mechanic. Its in his resume.

Actually, the title:

"Why Amazon Hired the Lead Architect for DB2 and SQL Server to Run Its Cloud Empire"

would instill much less confidence than the current one, consider how bad those products are compared to one that has proper architecture (i.e. PostgreSQL).

Care to elaborate ?

"When Amazon’s EC2 and S3 web services arrived in 2006, they changed the computing business. Big server makers — most noticeably Sun Microsystems — had been tinkering with the idea of selling computer power in much the same way that utility companies sell power or water, but much to everyone’s surprise, Amazon — a seller of books and CDs — managed to produce a set of services that the market was willing to pay for."

This reminded me of the Jonathan Schwartz (Sun's CEO) blog mentioning "Chief Electricity Officers" - I tracked down this archived copy of the March 2006 post: http://web.archive.org/web/20060720095546/http://blogs.sun.c...

Tracking those words a little further back, I found this 2004 snippet: "Peter Gassner, Salesforce.com. Build or buy? No, subscribe. On-demand utility. Chief electricity officer was once a real position." ( http://www.jaycross.com/wp/category/just-jay/page/130/ ), besides Nicholas Carr and others elaborations of similar vintage.

The ideas were definitely in the air, the problem was execution and delivery.

I wonder, why doesn't Amazon venture into actual utilities business? As in making their own power plants(nuke/solar/whatever), and renting power out?

I did remember reading about Bezos investing in a few energy companies, but nothing about Amazon.

First off, in some areas it is a regulated market which would prevent them from being able to do so (electricity at least). Second, it would be very expensive for very little gain. It wouldn't scale well and they would have a ton of new infrastructure to support.

On top of that, they would have to compete with companies who already have customers. The electricity market has recently become unregulated in Illinois, and despite being able to save money using a different provider, I haven't switched from ComEd. The reason is that the savings isn't enough for me to justify the time spent actually comparing and seeing if I save and then switching providers. Amazon would be up against this same kind of apathy, which would only occur after they get some sort of infrastructure up and going. Meaning they would have sunk capital into a business for nothing.

Funny to see how frequently the word "grid" is sprinkled into that blog post. Now we use "cloud" the same way. It illustrates how silly the industry gets about using/abusing whatever the latest buzzword is.

Larry Ellison, of all people, nails it:

http://www.youtube.com/watch?v=0FacYAI6DY0

James is amazing at getting together people inside these big companies. His talk series at MSFT was legendary. His asking me to present at his talks when I was just a couple of years out of school opened the door for many good things for me at MSFT.

This link from the article on his analysis of preventing the Super Bowl power failure is pure gold:

http://perspectives.mvdirona.com/2013/02/04/ThePowerFailureS...

nice read. It's also worth mentioning that James Hamilton's personal website [1] has been hosted by Godaddy ever since, including the domain's name servers.

[1] http://mvdirona.com

Best hair ever

http://mvdirona.com/jrh/work/JamesHamilton.jpg

Because software people get's no respect

Reply on news.ycombinator.com