Scaling: Why Giants Don't Exist | Hacker News Reader

Scaling: Why Giants Don't Exist

galileoandeinstein.physics.virginia.edu

35 pointssoundsop16 comments

On Being the Right Size is in the same vein:

"You can drop a mouse down a thousand-yard mine shaft; and, on arriving at the bottom, it gets a slight shock and walks away, provided that the ground is fairly soft. A rat is killed, a man is broken, a horse splashes."

http://irl.cs.ucla.edu/papers/right-size.html

This topic was investigated many years ago and was known as "The Physics of Lilliput."

For those who are only familiar with Gulliver's Travels as a movie, the 18th century book also includes a visit to a land of giants (Brobdingnag) and other places (http://en.wikipedia.org/wiki/Gulliver%27s_Travels).

Thus the physics studied applied to both giants and mini-people (re: @AngryParsley).

(only reference I could find: http://www.brookscole.com/physics_d/templates/student_resour...)

The overall field is called allometry. Kleiber's Law is another reason behind our size and shapes: http://en.wikipedia.org/wiki/Kleiber%27s_law

"for the vast majority of animals, an animal's metabolic rate scales to the ¾ power of the animal's mass."

On a planet with lower gravity, the same rules would apply, but the ratio would differ. A much larger human could exist there with the same proportions as an Earth human.

Perhaps this is how Galactus exists. He lives in space, where there is effectively zero gravity. Thus, he can be gigantic without any problems.

On a related note, there are some geologists who believe the Earth has increased in volume/mass over millions of years:

http://en.wikipedia.org/wiki/Expanding_Earth

For instance, dinosaurs grew very large because the Earth was much smaller with less mass and gravity.

Dinosaurs and Google. 'nuff said.

I thought of Google when I read the article, when I got to the line about large animals being made possible by the development of the circulatory system.

It seems like there has to be a qualitative change to deal with scale. Moore's Law has multiplied computing power by a factor of roughly a million over 30 years. In most of the industry, that capacity has been eaten up by productivity enhancements. We have CPU cycles to burn, so we burn them in interpreted languages like Python and Ruby and IDEs like Eclipse, for some marginal increase in productivity.

Google chose to burn them by processing larger and larger data sets, instead. And there're certain emergent phenomena that occur as the data set gets bigger. Almost nothing in search or ads would work with the data available at desktop-scale.

The interesting thing is that by pushing on data size instead of productivity, Google's had to find other ways to scale productivity and the number of engineers in the organization. I've mentioned some of the public ones in a previous comment (http://news.ycombinator.com/item?id=1101443); there're some others that I don't think I can talk about. But I find this whole notion of where you allocate resources fascinating. There's this conventional wisdom of always putting them into developer productivity, but Google's succeeded because they sometimes (judiciously) choose to move them into processing more data and find other ways to work around the productivity hit that entails.

I've continued to be perplexed by the lack of other companies copying Google's revolutionary style of data center operations. It makes such a huge difference in Google's capabilities (and their bottom line), yet data center operations outside of Google haven't changed much in a decade.

it's because Google's data center are geared for one single purpose: crunching massive amounts of data. They can get away with just one hardware platform, fine-tweak it, then design the racks.

Most other companies (99% of the web companies) other there have different needs. For instance, Facebook needs one hardware platform fine-tweaked for the cache, another one for pics and video storage, another one for chat, plus mysql etc.

It's the same for your regular web2.0 company, they want different kinds of servers, so they will ask data centers specific hardware and different server formats. It's therefore very hard, if not impossible, for a data center to copy Google's design, since they have to host so many kind of hardware.

Even so there are many companies that own their own data centers outright. It's intriguing to me that in an industry where a successful company can't exist for longer than a week without being cloned 8-ways from Sunday Google has managed to maintain its uniqueness over the course of a decade. In some ways that's a testament to Google's excellence, but in other ways it's a testament to the very dysfunctional nature of the state of the industry.

It's interesting how potent, and how rare, cross-domain expertise is in this industry. Google deftly coordinates data center operations (including server hardware) and software mastery, giving it a unique and enormous edge over its competitors. Apple weaves together expertise in device hardware, software development, usability, and aesthetics beyond anything its competitors can achieve, giving them significant market dominance and enormous brand prestige. Amazon combines competent large scale web-application hosting and development with highly efficient warehouse and fulfillment operations. It seems as though in this industry melding together expertise in related but seemingly far separated areas is a good recipe for attaining a near unassailable advantage over your competition.

good point.

I think it has more to do with the company's core values than the company's expertise.

Most data centers have expertise in data centers operations, but they don't know where to focus. Google is a master in optimization, so they took it to the next level, even in data centers. Apple values design and user experience that it sets the focus on future product development etc.

Google stores video and pics as well. They also use caching. I think your analogy is incorrect in this case.

Google does massively distributed computing, whereas Facebook doesn't have the same priorities.

But I agree, the analogy is not the best

Not everything Google does is about massively distributed computing. But the work on massively distributed computing payed off in serving images, videos, and doing caching content.

I see the point you are trying to make and even agree with. But it is still a little odd that no one saw the success that is possible with what google did and worked to replicate it.

While that bit may seem flippant, I also wonder about dinosaurs and scaling.

They don't seem to have half the supposed skeletal size that they 'should' according to simple scaling.

My understanding is that the earth's atmosphere used to be far denser and the oxygen concentration far higher. This is what allowed giant roaches and the like to exist at one time. Dinosaurs evolved in a different world than we live in.

I'm not real up on this information -- it's my sons who are really interested in it, not me -- but this detail sticks with me because I used to really be bothered by giant spiders in movies. One day, while watching something on TV that talked about dinosaurs and other historical critters, they showed a very large spider and said "If this spider were alive today...(it would the size of a house cat or hunting house cats or something)" and I said "If it were alive today, it would be on life support." My sons cracked up. And I no longer get freaked by giant movie spiders. They can't exist in earth's current atmosphere, which is why they don't exist on earth today.

Reply on news.ycombinator.com