Microsoft is restarting 3 mile island.A decade ago we tried to build a nuclear reactor and our government bankrupt our utlitity. We never got the reactor, and now we pay much more for electricity via Domninion.
Microsoft is restarting 3 mile island.A decade ago we tried to build a nuclear reactor and our government bankrupt our utlitity. We never got the reactor, and now we pay much more for electricity via Domninion.
Microsoft has one, the problem is corrosion and maintenance. You have enough of them some hardware would go bad and it is pain in the butt to lift a whole container from the ocean bottom to replace a stick of ram, then seal it up and sink it back down. We also haven't talked about powering something undersea and connecting fibers to it being a lot more expensive.May have been a podcast, but I recall that there's a company that builds mini-data centers inside shipping containers and then sinks them in the ocean. No need to water for cooling, no noise pollution, etc. I can't recall how they power them though (obviously they have a way).
We have long past the time of single point of failures. All large businesses have redundant regions for their data centers for disaster recovery. If you nuke their US site their backup would be up and running very quickly in Singapore, Australia, Europe, etc.These are the new missile silo aiming points for our enemies.
When I started doing inspections (audits) of facilities of pharmaceutical suppliers, I'd visit a lot of data centers. There were hidden away in plain site. Usually the basement of a building or what appeared the basement but was actually a cave underground. I'd go and visit the server used for my company and call it good. I'd mostly go because there data centers were such cool places to visit. There were about half the size of a football field. Nothing huge but still, that's a lot of computers in one place. Later, companies moved away from dedicated data centers and moved to AWS and the like and we didn't know where the data was stored. They had primaries servers with secondaries running full time. Plus tertiary backups. I don't know what they use now. It was a challenge for the pharmaceutical industry to not know the exact server where our data was stored. We've eventually gotten comfortable with it along with FDA.We have long past the time of single point of failures. All large businesses have redundant regions for their data centers for disaster recovery. If you nuke their US site their backup would be up and running very quickly in Singapore, Australia, Europe, etc.
You want to hear somehing scary? This is a tiny little baby data center.......trust me. You wouldn't believe what we (my company) has orders for, or what the maps look like.QTS built a huge data center in west midtown Atlanta. Phase 2 and 3 are under construction in this map but have since been completed and are right next to a residential neighborhood. The place is massive.
View attachment 334129
All three buildings together is considered small?You want to hear somehing scary? This is a tiny little baby data center.......trust me. You wouldn't believe what we (my company) has orders for, or what the maps look like.
My post was not clear; the idea is to put data centers out in the SF Bay for cost effective cooling using ocean water and less need for valuable land.Bay Area regions are pretty expensive last time I check on AWS US West. Oregon is a lot cheaper and unless you have some really low latency stuff like video conference server for Bay Area, you are better off starting a load in Oregon or Nevada (during day time only with cheap solar).
I drove through Central Expressway daily and the Vantage data centers never bother me one bit more than any other big companies office building. They blend in with fake glass windows and the gas turbine power plant on the streets powering them off grids are pretty quiet too.
Yep. I build my first real-time data base mirror in Oregon close to 20 years ago. Lotta earthquakes around these parts...We have long past the time of single point of failures. All large businesses have redundant regions for their data centers for disaster recovery. If you nuke their US site their backup would be up and running very quickly in Singapore, Australia, Europe, etc.
Not only that, but if anything gets nuked it’s going to be military bases first and by the time that happens, you wouldn’t want to be one of the remaining people aliveWe have long past the time of single point of failures. All large businesses have redundant regions for their data centers for disaster recovery. If you nuke their US site their backup would be up and running very quickly in Singapore, Australia, Europe, etc.
But do we need this huge increase in them?Using social media, i.e. BITOG, Facebook, etc - you're using data centers
Shopping online through Amazon, Walmart, etc, or browsing websites like Lowes, etc - you're using data centers
Searching the internet via any search engine - you're using data centers
Posting on the internet about disliking or being against data centers - you're using data centers.
Yup..........All three buildings together is considered small?
Your every post creates the need for more data storage. It’s like asking do we need to product more oil while you are actively using oil.But do we need this huge increase in them?
The podcast touched on this - they simply ignore it (shut that machine down). As you say, it's not practical to repair every single failure so they presumably wait until they reach some threshold before they access them.You have enough of them some hardware would go bad and it is pain in the butt to lift a whole container from the ocean bottom to replace a stick of ram, then seal it up and sink it back down.
You just used more resources of data center(s) by replying. I say "centers" because it's possible that whatever host BITOG uses may replicate data around North America or even the globe, depending on traffic, location of users, and so on.But do we need this huge increase in them?
What an odd thing to do. Unless they heatsink the processors directly to the wall of the container, they are flowing cool water through a radiator system inside the container anyway. Why sink it and not just pump cold water through the radiator system pumped from below the surface. The only thing that makes sense is that there is some regulation that doesn't allow you to pump water out of the ocean and onto a land-based structure. Doing this in international waters would be strange as you need to supply the power and have a robust data pipe, which would cost a lot to get that far out.The podcast touched on this - they simply ignore it (shut that machine down). As you say, it's not practical to repair every single failure so they presumably wait until they reach some threshold before they access them.
The sudden increase is all about AI.But do we need this huge increase in them?
My guess is most of the increase has to do with LNG export ban being lifted so we now sell those gas instead of using it within the US, then the AI, then Ukraine, then Iran.My electric bill is up 100% in the last 10 years with the biggest hikes in the last couple, and my gas delivery charges are also through the roof. They also suck up all the water, contribute absolutely nothing to the local economy, pen secretive deals with local utilities.... So no, I wish they'd all burn to the ground.