Hacker News Reader

Comment by crowcroft | Hacker News Reader

I guess the thesis is that Singapore is the primary business hub in SEA, but Singapore doesn't have a grid that can handle large scale data centers (especially with power hungry GPU demand). Logically Johor bordering Singapore is the next best option for data centers.

View on HN

Singapore is also very land-constrained. Google's latest DCs there are multi-story, which is obviously going to be more expensive to build that your standard one-story warehouse.

Long term, you'll want your datacenters to be roughly spherical to minimize cross-section path length latencies.

Volume increases to the cube, where as surface area only increases to square.

This results in large, perfectly spherical, data centre being more difficult to cool, especially when operating in a vacuum.

Agreed that surface cooling only for a spherical datacenter would be a bit silly; we're gonna need some exciting liquid plumbing to wend its way through, but that's already best practice for GPU cluster builds.

What about heat transfer? At least in a moderate, dry climate (whereas SP is hot/humid) there’s a natural benefit to having a high surface area to volume ratio, favoring wide/flat spaces, from the perspective of heat rejection. I’m not sure to what extent space cooling is actually a controlling factor in DC design, or what the typical strategies are (for instance I assume there are lots of new systems which make use of geothermal cooling). Anyways, just curious if you have any insight on that.

Passive heat rejection from the exterior of the datacenter was never going to be the way to go, even with a short, wide datacenter -- external evaporative cooling towers are the way to go. Liquid cooling for the core is already best practice for high power density racks, which gives you the ability to separate how you solve for cooling and how densely you can pack your gear.

Yeah cooling towers seems like the obvious choice, thanks for confirming! I’m sure there are some other more exciting designs that hook datacenters up to prosumer/mesh topology ambient-temp district heating/cooling loops - or there should be if there aren’t! - but what you described makes perfect sense.

Latency/Routing too. There are ISPs that would send you to Hong Kong or other places.

My understanding is that transit/peering in Southeast Asia is pretty awful in general, and that's why a lot of CDNs and service providers charge you extra for traffic exiting in that region. I'm not sure how much of that is a lack of investment in infrastructure and how much is unresolved corporate / geopolitical tensions - but either way, building more data centers isn't going to fix it.

I'm also curious as to why transit is so extortionately expensive in this region.

Lots of dominant telcos in their country (less competition).

Massive place requiring a lot of capital expensive submarine cables.

On a related note, in the last 1 to 2 years there has already begun a shift in terms of multi-national investment in APAC outside of Singapore.

Singapore used to be the obvious location for MNC offices in APAC. Especially since Hong Kong's return to China, Singapore has benefited and see an increase in head offices and employment.

However, over the last couple of decades surrounding countries have improved significantly. Malaysia, Indonesia, Thailand to name a few. Layer on top the ever increasing cost of living in Singapore and the latest trend is MNCs moving their offices outside of Singapore.

The benefit is a lower cost of operation including office rent and employee wages and it will be interesting to see how the trend continues. Not sure if that will help the process of locating datacenters in those countries as well, I guess we'll see.

While multinationals are investing elsewhere in SE Asia to serve SE Asian markets, I'm not seeing them move their HQs out of Singapore though, because they don't trust the Malaysian/Thai/Indonesian legal system and regulatory environment.

Cyberjaya (yes, you can cringe at the name) in Malaysia was built with the explicit goal of attracting them, but it's mostly call centers, not R&D. Even Malaysian success stories like Grab (the SE Asian Uber) prefer to operate out of Singapore.

Being able to trust the govt is a pretty important factor for businesses.

https://www.thevibes.com/articles/world/101600/singapore-los...

Some are moving HQ, some are just downsizing HQ and expanding offices in Malaysia or Thailand.

I agree it’s not a wholesale move out, but the trend is noticeable.

The cost of living is rising everywhere, not just Singapore, so I’m not sure how relevant that factor is. Do you have evidence that MNCs are net moving out of Singapore?

There was once a proposal to run an extension cord from Australia to Singapore [1] since Australia could have plentiful and cheap solar power. However, the project collapsed after the capricious billionaire backers ditched it.

[1] https://en.wikipedia.org/wiki/Australia-Asia_Power_Link

The SunCable project is still moving forward.

There was a disagreement between the co-founders. One wanted to use HVDC cable to transmit power, the other wanted to use electrolysis to produce Hydrogen, put that on ships, then burn it in Singapore to produce electricity.

To resolve the dispute, they both agreed to put the project into administration, then both made bids to take full control, with HVDC proponent Mike Cannon-Brooks (co-founder of Atlassian btw) eventually winning control.

I wouldn't be surprised if the Green hydrogen plan eventually makes a come back, as solar power from Northern Australia has a compellingly low LCOE.

Twiggy was also talking about producing hydrogen in Tasmania from largely unavailable hydropower, or so the argument goes. If I recall correctly then turn that in to green ammonia.

Many of those dams are full to excess, at least sometimes. And what Twiggy doesn’t waste turning falling water in to energy anaemic hydrogen gas, ammonia is handy though, we can keep selling to Victoria for a profit, for the Tasmanian government to, what, waste on a footy stadium no one wants? No no, to waste storing the new Spirit of Scotland ferries.

Cynical bugger.

> for the Tasmanian government to, what, waste on a footy stadium no one wants?

It’s a Utopia: https://youtube.com/watch?v=3XUn-EsThcE

(That video was uploaded a year ago but the episode the clip is from aired years before the new stadium was even proposed. And that’s not the only time that the TV show predicted a useless government project)

This is an insane idea, but would be epic if someone could figure it out. I don’t know much about how electricity grids work, but that seems fraught with so many challenges, not all of them technical.

I believe they have the technical challenges solved (at least on paper), and the Singaporean and Australian governments have both approved the project.

I'm a little surprised Singapore approved it, as it'll need to supply about 10% of Singapore's power in order to cover the cost of the cable, which seems like a big vulnerability to me.

Next question is, can they find enough buyers in Singapore for that much power.

Singapore's electricity is currently produced mostly from LNG piped in from Malaysia and Indonesia, so more diversification is welcome.

At least there’s a fairly direct route from those countries to Singapore. Australia is pretty far, and there are some countries in between that could be a menace at any point in the future.

Pretty big challenge for Singapore to figure out a way to guarantee security long term.

Why bother going all the way to Singapore when their neighbor (Java) has 100 million people?

Singapore rich. Indonesia poor.

Reply on news.ycombinator.com