News for hackers. Stuff that matters.
Re:The 80/20 rule for storage systems by nl on Monday December 15, 2014 @03:22AM (#8751091)
Can a 'perfect' algorithm actually give end-users the consistent feel of flash performance with 80% of data stuck on platters? I doubt it.

Of course not. That's not the point!

What may be possible is to give the "consistent feel of flash performance" some very high percent of the time for a dramatically lower price.

If it's possible to get flash-like performance 95%+ of the time for <50% of the price then I'd find that pretty compelling.

Re:The 80/20 rule for storage systems by zaroth on Tuesday December 16, 2014 @01:33AM (#8756102)
But flash will have higher performance and higher capacity, at least that's the projections. I don't know why flash shouldn't be able to win at price/capacity, longevity, total capacity, robustness, performance, etc. At some point, there just shouldn't be any reason to manufacture storage in any other form. In a few generations, isn't that the level of dominance of flash?
Re:The 80/20 rule for storage systems by rasz_pl on Tuesday December 16, 2014 @02:14AM (#8756163)
>I don't know why flash shouldn't be able to win at price/capacity, longevity, total capacity, robustness, performance, etc

yes, you dont, and this is why you think this: >But flash will have higher performance and higher capacity

longevity, robustness, endurance, it all goes down with every new generation of flash. Performance at silicon level is already as good as it gets, we only win by making wider busses and assessing more data in parallel. Price per gigabyte comes at a price (haha) or endurance. Speed comes at a price of more chips per disk (or more wafers/structures inside chips) = money.

Want that sweet new 1TB PCIE SSD able to write at 2GB/s? it will shit itself after ~24hours of continuous write.

There is no free lunch with SSDs, they are a dead alley.