Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Kioxia and Dell cram 10 PB into slim 2RU server

blocksandfiles.com

Post reply on HN
Print view View original

91–100 of 116 posts

Re: Kioxia and Dell cram 10 PB into slim 2RU server

#91

Post by fancyfredbot »

rbanffy wrote: post #82
Pallav123 wrote: post #24

At current enterprise NVMe prices, the drives alone for this must easily push past the $500k to $1M mark. It's fascinating to see this level of density, but it’s strictly going to be hyperscaler or high-end defense/research budget territory for a long time.

I believe the rationale is that they are so much denser than they will compensate the price difference over the entire lifetime of the system. A 10 TB install is a full rack otherwise and this is 5% of the space. Colocation for HFT tends to be expensive and using 5% of the space for the same amount of data might make sense.

Your average HFT algorithm will not have time to access storage at all, let alone 10 PB of it.

Back testing an HFT algo could very plausibly need 10PB, but I don't see that being co-located with the exchange.

Re: Kioxia and Dell cram 10 PB into slim 2RU server

#92

Post by tomhow »

retired wrote: post #88
tomhow wrote: post #73

Earlier quoted context omitted.

Please don't sneer at imaginary people on HN. We're trying for something better than that here. https://news.ycombinator.com/newsguidelines.html

Guess it is time to retire this account then.

Its' fine to just observe the guidelines! They apply equally to everyone.

Re: Kioxia and Dell cram 10 PB into slim 2RU server

#93

Post by rbanffy »

fancyfredbot wrote: post #91
rbanffy wrote: post #82

Earlier quoted context omitted.

I believe the rationale is that they are so much denser than they will compensate the price difference over the entire lifetime of the system. A 10 TB install is a full rack otherwise and this is 5% of the space. Colocation for HFT tends to be expensive and using 5% of the space for the same amount of data might make sense.

Your average HFT algorithm will not have time to access storage at all, let alone 10 PB of it. Back testing an HFT algo could very plausibly need 10PB, but I don't see that being co-located with the exchange.

What would be good cases for data-heavy applications in expensive places?

Re: Kioxia and Dell cram 10 PB into slim 2RU server

#94

Post by rbanffy »

stingraycharles wrote: post #84
rbanffy wrote: post #79

Earlier quoted context omitted.

What is the designed lifespan of a Starlink satellite?

5 years. https://starlink.com/public-files/Starlink_Approach_to_Satel...

I don’t think the flash storage would keep data integrity for that long without a lot of redundancy. Also, processing capacity would be limited as well. Is there a problem that can be solved with a LEO server that can’t be better solved with the same amount of money sitting in a datacenter somewhere?

Re: Kioxia and Dell cram 10 PB into slim 2RU server

#95

Post by randomtoast »

dmurray wrote: post #35
evil-olive wrote: post #30

Earlier quoted context omitted.

no, absolutely not. orbital datacenters are never going to happen, it doesn't matter whether you try to frame them as compute or storage or whatever else. the extreme density of these SSDs is actually an anti-feature in the context of spacecraft hardware. the RAD750 CPU [0] for example uses a 150nm process node. its successor the RAD5500 [1] is down to 45nm. that's an order of magnitude larger than chips currently ma…

In the limit, packing transistors tighter should mean more radiation resistance, not less, because you can shield them with a smaller mass of water or lead or whatever.

The water cooling system then serves multiple functions, including acting as a radiation shield.

Re: Kioxia and Dell cram 10 PB into slim 2RU server

#96

Post by beAbU »

9dev wrote: post #52
tesdinger wrote: post #49

Earlier quoted context omitted.

For the sake of the generations that come after us, we really should not dump valuable material into space. I somehow doubt the electronics in space would be recovered and recycled properly.

Nothing is recycled properly. Recycling was a story told to ease consumers minds so they keep on consuming. The stuff you throw away ends up on a landfill, in the sea, or on a ship to someplace else where it gets burned and then buried. Sending it to space makes absolutely no difference.

At some point a massive landfill will become more economically viable to mine for rare or useful elements than an actual mine. There is a good case for piling rubbish into to a well-managed landfill for future generations to mine with better technologies.

Also, burning plastic waste in a power plant at least liberates the energy for use. One can argue it's just fossil fuels that spent time doing something useful before it became energy.

Taking rare elements to space only for it to stay up there or burn up in the atmosphere removes it from this whole equation.

Re: Kioxia and Dell cram 10 PB into slim 2RU server

#99

Post by dsr_ »

rbanffy wrote: post #94
stingraycharles wrote: post #84

Earlier quoted context omitted.

5 years. https://starlink.com/public-files/Starlink_Approach_to_Satel...

I don’t think the flash storage would keep data integrity for that long without a lot of redundancy. Also, processing capacity would be limited as well. Is there a problem that can be solved with a LEO server that can’t be better solved with the same amount of money sitting in a datacenter somewhere?

Pumping the SpaceX IPO.

Re: Kioxia and Dell cram 10 PB into slim 2RU server

#100

Post by sharts »

NitpickLawyer wrote: post #12

There's been a lot of talk about orbital DCs lately, but with these levels of density, orbital CDNs might be a more obvious usecase. It would be interesting to see if something like Starlink can use something like this to cache media content and reduce their overall data moving through the constellation. It could even be worth it to have some satellites in higher orbits (even GEO if the ground hw can reach it) dedica…

why pollute the skies?
Post reply on HN