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

Will I ever own a zettaflop?

geohot.github.io

Post reply on HN
Print view View original

31–40 of 98 posts

Re: Will I ever own a zettaflop?

#31

Post by jandrewrogers »

popol12 wrote: post #12
throw0101d wrote: post #7

Somewhat related, why the creators of Zettabyte File System (ZFS) decided to make it 128 bits (writing in 2004): > Some customers already have datasets on the order of a petabyte, or 2^50 bytes. Thus the 64-bit capacity limit of 2^64 bytes is only 14 doublings away. Moore's Law for storage predicts that capacity will continue to double every 9-12 months, which means we'll start to hit the 64-bit limit in about a deca…

Very interesting, could someone please do the same computation for filling 64 bit storage?

Storage densities can be extremely high. Filling 2^64 of storage is very doable and people have been doing it for a while. It all moves downstream; I remember when a 2^32 was an unimaginable amount of storage.

Many petabytes fit in a single rack and many data sources generate several petabytes per day. I'm aware of sources that in aggregate store exabytes per day. Most of which gets promptly deleted because platforms that can efficiently analyze data at that scale are severely lacking.

I've never heard of anyone actually storing zettabytes but it isn't beyond the realm of possibility in the not too distant future.

Re: Will I ever own a zettaflop?

#32

Post by jasonwatkinspdx »

jandrewrogers wrote: post #29
jasonwatkinspdx wrote: post #22

Earlier quoted context omitted.

Virtualization has pushed back the need for a while, but we are going to have to look at pointers larger than 64 bit at some point. It's also not just about the raw size of datasets, but how we get a lot of utility out of various memory mapping tricks, so we consume more address space than the strict minimum required by the dataset. Also if we move up to 128 bit a lot more security mitigations become possible.

By virtualization are you referring to virtual memory? We haven't even been able to mmap() the direct-attached storage on some AWS instances for years due to limitations on virtual memory. With larger virtual memory addresses there is still the issue that the ratio between storage and physical memory in large systems would be so high that cache replacement algorithms don't work for most applications. You can switch t…

> By virtualization are you referring to virtual memory?

No, I mean k8s style architecture, where you take physical boxes and slice them into smaller partitions, hence the dataset on each partition is smaller than the raw hardware capability. That reduces the pressure towards the limit.

Re: Will I ever own a zettaflop?

#33

Post by jandrewrogers »

jasonwatkinspdx wrote: post #32
jandrewrogers wrote: post #29

Earlier quoted context omitted.

By virtualization are you referring to virtual memory? We haven't even been able to mmap() the direct-attached storage on some AWS instances for years due to limitations on virtual memory. With larger virtual memory addresses there is still the issue that the ratio between storage and physical memory in large systems would be so high that cache replacement algorithms don't work for most applications. You can switch t…

> By virtualization are you referring to virtual memory? No, I mean k8s style architecture, where you take physical boxes and slice them into smaller partitions, hence the dataset on each partition is smaller than the raw hardware capability. That reduces the pressure towards the limit.

Ah yeah, that makes sense. With a good enough scheduler that starts to look a lot like a cache admission architecture.

Re: Will I ever own a zettaflop?

#34

Post by echelon »

ks2048 wrote: post #27
echelon wrote: post #5

There's no way we're not living in a historical simulation. This is all just such crazy coincidence. Everything is coming together so quickly.

What's a "a historical simulation" and why is it all such a coincidence? The "simulation argument" to me is ridiculous.

We don't know if life is rare, but intelligence seems rare given our sky surveys to date.

Life started on earth shortly after the solar system formed. It's a quarter the age of the universe that it took for intelligence and civilization to arise. A long, long, looooong time.

From a numbers perspective, our minds are all shockingly rare. The universe probably doesn't produce many of us through stellar and then biological evolution.

Taking that into consideration, contrast that with the flip side.

If high technological simulation exists that becomes indistinguishable from reality, it could simulate quintillions of minds. It seems like we're on the path to that technology, with a great degree of probability.

Given that, and the fact that historical simulation is likely easy for the future, it seems more probable to me that we're one of the quintillions of simulations rather than the origin timeline.

I'm half joking, but I'm half not. It's a fun thought experiment with absolutely no basis in science.

Do you even really know if you were alive a second ago? Perhaps you were just instanced into existence with a set of memories - memories you are not randomly accessing. A very deliberate version of a Boltzmann Brain.

Again, pseudoscientific tomfoolery, but fun to ponder.

Re: Will I ever own a zettaflop?

#35

Post by defrost »

echelon wrote: post #34
ks2048 wrote: post #27

Earlier quoted context omitted.

What's a "a historical simulation" and why is it all such a coincidence? The "simulation argument" to me is ridiculous.

We don't know if life is rare, but intelligence seems rare given our sky surveys to date. Life started on earth shortly after the solar system formed. It's a quarter the age of the universe that it took for intelligence and civilization to arise. A long, long, looooong time. From a numbers perspective, our minds are all shockingly rare. The universe probably doesn't produce many of us through stellar and then biologi…

> but intelligence seems rare given our sky surveys to date.

Most of the universe, by a large percentage (or so I'm told), is further away than 200 light years.

What signs of intelligence would we see or detect in our own solar system from a distance of 200 light years should we scan it with all our latest and best tech?

Re: Will I ever own a zettaflop?

#36

Post by mysecretaccount »

Fun post, but I find the industry's obsession with compute to be rather vapid, and this is a good example:

> One million Claudes. To be able to search every book in history, solve math problems, write novels, read every comment, watch every reel, iterate over and over on a piece of code until it’s perfect – spend a human year in 10 minutes. 50,000 people working for you, all aligned with you, all answering as one.

We are already near the limits of what we can do if we throw compute at Claude without improving the underlying models, and it is not clear how we can get big improvements on the underlying models at this point. Surely geohot knows this, so I am surprised he thinks that "one million Claudes" would be able to e.g. write a better novel than one hundred Claudes, or even one Claude.

Re: Will I ever own a zettaflop?

#37

Post by latentframe »

Maybe the bottleneck is shifting from compute to energy and capital ; at some point it stops being a software problem and starts looking like infrastructure power land cooling... Just feels like the constraint is moving down the stack

Re: Will I ever own a zettaflop?

#38

Post by eru »

jasonwatkinspdx wrote: post #22
jandrewrogers wrote: post #16

Earlier quoted context omitted.

Single data sets surpassed 2^64 bytes over a decade ago. This creates fun challenges since just the metadata structures can't fit in the RAM of the largest machines we build today.

Virtualization has pushed back the need for a while, but we are going to have to look at pointers larger than 64 bit at some point. It's also not just about the raw size of datasets, but how we get a lot of utility out of various memory mapping tricks, so we consume more address space than the strict minimum required by the dataset. Also if we move up to 128 bit a lot more security mitigations become possible.

Please keep in mind that doubling isn't the only option. There's lots of numbers between 64 and 128.

Re: Will I ever own a zettaflop?

#39

Post by eru »

nl wrote: post #10

Firstly, True Names is an awesome read, and the real origin of cyberpunk. I much prefer it to Neuromancer or Diamond Age. Secondly, I recently tried to work out what year on the Top500 list[1] I could reasonably be for around US$5000. It's surprisingly difficult to work out mostly because they use 64 bit flops and few other systems quote that number. [1] https://top500.org/lists/top500/2025/11/

Isn't the Diamond Age something like post-cyberpunk already?

It came out three years after Snow Crash, which already ironically referenced "The sky above the port was the color of television, tuned to a dead channel".

I agree that Neuromancer wasn't a great novel, though it obviously had vibes that resonated with many people. The novel being otherwise a bit of a dud actually speaks to how strong the vibes were to overcome that.

Re: Will I ever own a zettaflop?

#40

Post by jasonwatkinspdx »

jandrewrogers wrote: post #33
jasonwatkinspdx wrote: post #32

Earlier quoted context omitted.

> By virtualization are you referring to virtual memory? No, I mean k8s style architecture, where you take physical boxes and slice them into smaller partitions, hence the dataset on each partition is smaller than the raw hardware capability. That reduces the pressure towards the limit.

Ah yeah, that makes sense. With a good enough scheduler that starts to look a lot like a cache admission architecture.

I'd never thought of it that way, and it's an interesting perspective.
Post reply on HN