SSDs to be more economical solution than HDDs by 2016

[crossy]

I'm doubtful, these figures are based on SSD prices falling through the floor in 2016, while HDD prices remain the same per GB.

Spot on - and I would have thought it a pretty safe bet that the HDD manufacturers will continue to push capacity since that's the big plus over SSD's (at the moment).

I'd be surprised if there isn't even more expensive storage out there that's spinny disk based.

True, I'm certainly aware of highly fault-tolerant (claimed), high-performance SAN setups that have had eye-watering prices, and with narry an SSD in sight.

Of course, maybe Apple will sell you an "iSAN", that uses hand-built titanium/gold/carbon-fibre head arms and unicorn blood disk substrates. (Sorry, couldn't resist)

Here's my own prediction

Actually a good point - there's not that many NAND manufacturers, so losing one to fire/flood/earthquake/riot could put a real crimp in those predictions.

[abaxas]

Another thing to consider is that SSDs add another dimension (wear) to planning. IE another level of complexity.

[yeeeeman]

The same revolution is happening with the electric cars. They only need some sort of breakthrough in battery technology as the 3D was for the flash to really get the sales flying through the roof. I really do hope that will happen sooner than later.

[DanceswithUnix]

I don't get it.

We already have 3D flash, so prices due to that have already fallen through the floor.

Rotating media must at some point finally get HAMR technology. That will help their prices a *lot*.

Finally I don't get the thing about performance of rotating media. People are switching from SAN to NAS becuase performance on NAS is good enough and it is easier to use. That doesn't say to me that there is a problem.

Still, I do think that 15K rpm SAS drives will be replaced by SSD, but then I don't think that is really news.

[shaithis]

Still, I do think that 15K rpm SAS drives will be replaced by SSD, but then I don't think that is really news.

The problem with replacing SAS with SSD is the wear-levels of SSDs making them likely to fail simultaneously in a RAID. You would need staggered wear levels across the drives in the array.

It's the one thing about SSDs in the enterprise that makes me nervous.

[ik9000]

The problem with replacing SAS with SSD is the wear-levels of SSDs making them likely to fail simultaneously in a RAID. You would need staggered wear levels across the drives in the array.

It's the one thing about SSDs in the enterprise that makes me nervous.

that and data loss if you leave one disconnected for too long a period of time!

[DanceswithUnix]

The problem with replacing SAS with SSD is the wear-levels of SSDs making them likely to fail simultaneously in a RAID. You would need staggered wear levels across the drives in the array.

It's the one thing about SSDs in the enterprise that makes me nervous.

With spindles you replace them just before they hit their 5 year warranty/support limit.

With an SSD you replace them just before they hit their warranty/support limit, which I presume is going to be the same 5 years as an enterprise spinning media.

The risk of multiple drives failing at once is a risk if you run them to destruction like some home users would, but I don't see enterprise users doing that.

[crossy]

Finally I don't get the thing about performance of rotating media. People are switching from SAN to NAS becuase performance on NAS is good enough and it is easier to use. That doesn't say to me that there is a problem. Still, I do think that 15K rpm SAS drives will be replaced by SSD, but then I don't think that is really news.

Going to agree with you on that last bit - I'm actually seeing more and more bumf on hybrid arrays with a mix of slower HDDs and top speed SSD's where the SAN array firmware (? I'm not a storage guy) is able to recommend moving of higher hit data to the fast areas. I'm also not seeing a wholesale move from SAN to NAS, what I am seeing is more NAS's being deployed. But this seems to be more to do with people wanting shared storage - simple cloud storage as it were, (runs and hides before Saracen finds the tar and feathers)

The problem with replacing SAS with SSD is the wear-levels of SSDs making them likely to fail simultaneously in a RAID. You would need staggered wear levels across the drives in the array. It's the one thing about SSDs in the enterprise that makes me nervous.

Simple (host based) RAID you're right, but surely enterprise SSD's will invariably not be in someone's desktop, but more likely in some kind of storage array. And those, I know, have all kinds of stuff for predictive failure etc. Sure I saw something from IBM the other day that you can poke in an MTBF (or does it read something from the drive?) and it'll warn you when an SSD is approaching it's fail period. Also know that someone else does alerting based on block relocations (?). But I'm not a storage dude, so that stuff is all a bit "black box" to me.

[Spreadie]

Here's my own prediction

That did make me chuckle. Nice one.

Looks like the report is focussing on datacentre costs rather than home costs.
To put the difference is perspective, one of the arrays at my work cost £90,000 to add 32TB of 2TB 7200 RPM disks with licensing.

It cost them £90k for sixteen HDDs?

Did they buy them from PC World?

[mikerr]

Here's my own prediction

LOL - that's so true. We're still suffering from that HDD pricing con 4 years later.
Pre-flood I bought a 2TB drive for £45...

[DanceswithUnix]

I'm also not seeing a wholesale move from SAN to NAS, what I am seeing is more NAS's being deployed. But this seems to be more to do with people wanting shared storage - simple cloud storage as it were, (runs and hides before Saracen finds the tar and feathers)

I think it is a VMware thing. Rather than getting a VM to use a LUN on a SAN, you get it to use a file on an NFS mount.