Hacker News Reader

Comment by elymar | Hacker News Reader

"I think there is a world market for maybe five computers."

Thomas Watson, president of IBM, 1943

Not Thomas Watson, but A. S. Householder[1] in Nicholas Metropolis et al "A History of Computing in the Twentieth Century"[2]:

https://mobile.twitter.com/benklaasen/status/874543989004881...

[1] https://en.m.wikipedia.org/wiki/Alston_Scott_Householder [2] https://dl.acm.org/citation.cfm?id=578384

There’s no evidence he ever said that though.

Think of data centers as huge mainframes and we will soon come close.

> Think of data centers as huge mainframes and we will soon come close.

I... kinda doubt it. All of the really big players have access to markets that are willing to pay huge margins.

there are a lot of smaller players that would be happy to sell you compute at 75% the price of amazon, and bandwidth at 1/10th the price of amazon. Yeah, the big customers are happy paying the premium, but there are plenty of small customers that will buy service from a smaller player if it means a big discount.

It's going to be increasingly difficult to compete with the big players' economies of scale though. There's a reason we don't buy cars from the neighbor's factory or generate our own electricity these days.

like... half the people I know who own a single family home generate a good portion of their own electricity.

Sometimes centralization helps efficiency, it's true, but sometimes the opposite, decentralization helps, too... it sure looks like for peak-load in sunny areas, decentralized power generation is cheaper.

I think that for commodity compute, the big players are not a lot more efficient than the big hardware manufacturers. Go to a supermicro, a foxcon, or other ODM server manufacturer, and you'd be surprised how cheap it is to buy actual hardware.

Small companies will scrap for margins that the big players, so far, at least, have laughed at, because so far, most of the market seems to be okay paying these large margins. (I used to be one of those small companies scrapping for those low margins; I am not happily and much more renumeratively employed by a big player, outside of their cloud department)

I personally think that long term, smaller players or in-house hosting will be the way to go for your 'base load' compute resource. I think that paying the margins of the big players makes a lot more sense for your 'peak load'

But, of course, we don't know yet; none of this will shake out until the next serious economic downturn. There isn't a lot of pressure to save money when the economy looks like this.

That might have been right at the time.

1889: “Fooling around with alternating current (AC) is just a waste of time. Nobody will use it, ever.” — Thomas Edison.

2019: I have a 100% DC powered datacenter, but the power certainly gets to its location via AC.

Transmission could be done via DC too, and there are quite a few benefits to it. The only field left to conquer is generation of electricity, looks like that's going to be in AC.

It is, but rather difficult and expensive for short to medium distances. Submarine power cables are often HVDC.

https://en.wikipedia.org/wiki/Pacific_DC_Intertie

Great article. The thing's grounding system is lustworthy:

The Sylmar grounding system is a line of 24 silicon-iron alloy electrodes submerged in the Pacific Ocean at Will Rogers State Beach[4] suspended in concrete enclosures about one meter above the ocean floor.

The grounding system at Celilo consists of 1,067 cast iron anodes buried in a two-foot trench of petroleum coke, which behaves as an electrode, arranged in a ring of 2.02 mi (3,250.87 m) circumference at Rice Flats (near Rice, Oregon), which is 6.6 mi (10.6 km) SSE of Celilo.

Incredible! I've always been fascinated when driving by the Sylmar station (many times in my life) and never knew that I was driving by connecting infrastructure when on PCH in that area.

And even more interesting is they are replacing/replaced it! https://www.circlingthenews.com/terminus-of-sylmar-ground-re...

Cool! Which led me further down the rabbit hole: https://en.m.wikipedia.org/wiki/High-voltage_direct_current

- HVDC uses voltages between 100 kV and 1,500 kV.

- HVDC can be cheaper for 500+ mile runs

- HVDC allows power transmission between unsynchronized AC transmission systems

- Recent advances make UHVDC feasible (800kV+)

Great excuse to get introduced to this wonderful engineering youtube channel (no affiliation) https://www.youtube.com/watch?v=DFQG9kuXSxg

What benefits? The downsides are massive resistive losses.

That depends entirely on the voltage.

AC won because transformers are relatively simple and can change AC voltage, but we can now build stuff to change AC voltage too.

Don't forget that three phase allows for three conductors to carry the power of three pairs of single phase conductors.

Alternatively, skin effect limits the amount of AC current you can throw over a single conductor. With DC, you can make your conductors arbitrarily thick and you'll use the whole thing.

Three-phase AC also allows for very electromechanically simple generators and motors. Industrial electrical loads have traditionally been (and often still are) rotating machinery, putting three-phase at an advantage.

That's not entirely true. Any load imbalance on the phases is resolved through the null, connected to earth at both ends of the line. You can do the same with three single phase lines, although you wound need a massive earthing system with low resistance.

Or you could build a 2 wire (or 4, 6 ...) DC line with complementary polarity that would carry more power due to skin effects. If the negative and positive loads are balanced, the currents though the earthing wire cancel out.

Arguably, it's now cheaper in raw material terms to up-step and down step DC with high frequency transformers and buck-boost. The problems are lower efficiency than a modern, high efficiency transformer, which leads to increased dissipation and lower life of the complex electronics involved, requiring active cooling. An AC transformer is a well understood technology that can operate for decades without any maintenance. DC-DC high power convertors are exotic beasts with large downtimes and expensive to repair and maintain.

We are a long way of seeing this setup used outside very special applications, AC might no longer have a natural advantage but the benefits of DC are too small to justify the massive investments and research required.

Though, it can certainly be argued that a single home-wide rectifier and smart DC wall sockets could improve efficiency and safety. You could have things like negotiated power up, ground insulation, cheaper and more efficient light bulbs and appliances etc.

People are already building 802.3af and 802.3ad based office LED lighting systems at 48VDC, since it's much cheaper to run cat5e around a big building.

Several are given in this Quora thread [1]. The main ones are:

• No frequency matching issues between systems.

• DC systems appears as an open circuit to AC systems, limiting the spread of faults.

• Less expensive towers and wires.

• No reactive losses.

• No line length limits. Apparently AC has problems with lines longer than half a wavelength (2500 km or 1500 miles).

[1] https://www.quora.com/When-and-why-is-DC-used-instead-of-AC-...

Edison made money off of selling DC infrastructure though.

"Brilliant" man, 1937 recipient of der Fuhrer's "The Merit Cross of the German Eagle with Star", granted to 'honor foreign nationals who made themselves deserving of the German Reich.' for selling IBM tabulation machines, technology which enabled genocide.

Reply on news.ycombinator.com