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

HP plans to release first memristor, alternative to flash and SSDs in 18 months

nextbigfuture.com

Post reply on HN
Print view View original

21–30 of 55 posts

Re: HP plans to release first memristor, alternative to flash and SSDs in 18 months

#21

Post by ericfrenkiel »

TheEzEzz wrote: post #15

What are the implications for back end development? Will this greatly reduce server complexity (need for redundancy)? Could AWS just give you a machine in the cloud, and you wouldn't need to worry about databases and database backups and so on, you could just keep all your data locally in natural data structures?

This would not necessarily affect back end development as you would still want high availability for the application by using an active-active setup across two availability zones in EC2.

With regard to your question on data structures, yes - they would have to change. MySQL and most other databases use b-trees since they were meant to live on disk. Just running MySQL on memsistor-backed storage would result in a considerable waste of CPU and storage capacity (b-trees are not very compact).

Running a database like MemSQL on memsistors would make the most sense since it uses data structures meant for DRAM.

Re: HP plans to release first memristor, alternative to flash and SSDs in 18 months

#22

Post by bluehavana »

plasma wrote: post #14

Watch HP research's Stanley Williams describe the memristor and what they are working towards in more detail on YouTube: http://www.youtube.com/watch?v=bKGhvKyjgLY&sns=em

The no more hard drive and persistent ram thing should be really interesting. Completely change how we do things.

Also, application specific processing... writing instruction sets for each application will be really interesting. Cell processing et al will be a thing of the past.

Re: HP plans to release first memristor, alternative to flash and SSDs in 18 months

#23

Post by modeless »

hornokplease wrote: post #20
uniclaude wrote: post #9

This article is mostly made of quotes from an EEtimes article[1], adds no information or value to it, and is (to me at least) less readable. [1]: http://eetimes.com/electronics-news/4229171/HP-Hynix-to-laun...

EE Times article submitted here: http://news.ycombinator.com/item?id=3080963

And here: http://news.ycombinator.com/item?id=3081273

Re: HP plans to release first memristor, alternative to flash and SSDs in 18 months

#26

Post by calebmpeterson »

So how would this affect what languages we use? How different data structures perform? etc. Will this amplify the end of the free lunch?

I'm especially curious how Rich Hickey's approach to state, time, and identity in a functional language fit in?

Re: HP plans to release first memristor, alternative to flash and SSDs in 18 months

#27

Post by montecarl »

TheEzEzz wrote: post #17
pjscott wrote: post #10

The really exciting part here, to me, is the idea of fabricating large amounts of nonvolatile memory on top of a CPU. Modern processors already spend a huge amount of their time waiting on memory, and a great amount of power trying to hide that memory latency. If these guys can lower memory latency dramatically -- and it looks like they can -- computers would get a lot faster.

This would be a huge boon for scientific computing. The data is often huge, but all of it must be constantly pushed through the CPU every iteration.

As the number of cores per socket is steadily increasing it is becoming more difficult to keep all of the cores on a single computer operating efficiently. A higher clock rate per core, or more cores per socket, is not very useful if they are all waiting on memory.

Re: HP plans to release first memristor, alternative to flash and SSDs in 18 months

#29

Post by zackattack »

viraptor wrote: post #16
dhughes wrote: post #4

I doubt HP itself will exist in 18 months.

It won't disappear. It could be split, close some sites, change leadership again, etc. but it won't simply cease to exist. The existing infrastructure itself is worth too much.

What would you say are HP's top 5 most valuable pieces of infrastructure?
Post reply on HN