Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror

Comment Re:TLC NAND = unstable? (Score 1) 42 42

Well, it won't be the same price - it requires a more complex fab process - but yeah. Consumer MLC drives have proven themselves to be robust and reliable, for the most part. TLC still seems to be a bridge a little too far.

I'd like to see Tech Report re-run their endurance test with current drive models. The only "problem" is that drives are so good now that by the time the best model fails and we get the final score, none of them will be on the market any more.

Comment Re:The butting edge (Score 1) 42 42

A huge proportion of computing is moving to the cloud. Conventional disk storage is a nightmare for cloud services because there's such a huge disparity between sequential I/O and random I/O performance. CPU, memory, and network bandwidth all divide up nicely, but as soon as you have contention for disk I/O, it all falls apart.

This is known as the "noisy neighbour" problem. You might be happy and fine on your cheap VPS for months, and then the next day it collapses in a heap, even though you're getting the same allocation of resources you always have. It takes a lot of complex engineering and expert management to keep noisy neighbours at bay.

SSDs eliminate this, because SSD performance on multi-threaded random I/O is not far short of even their best-case sequential performance. Which means that every cloud provider wants to move to pure SSD. Exabytes worth of it. I work for a relatively small company, and we have 1.5PB of disk that we'd love to convert to SSD.

That SSD needs to be dense and reliable. Smaller process nodes improved density at a cost of write cycles. 3D flash gives us improved density and increased write cycles at a cost of more complex fabrication.

The market is there, all right.

There are running jobs. Why don't you go chase them?

Working...