2019: The year all-flash finally goes lamestream – but you know we were into it before it was cool All-flash storage arrays are very fast, but their limited capacity and prohibitive cost has meant that, historically, they were not very popular in enterprise data centres. According to analysts, this is about to change. Market watcher IHS Markit has said 2019 is the year when all-flash arrays could finally overtake HDD-based …
I'm surprised that any HDD manufacturer is spending any cash on R&D. I just knocked (well, pulled as there are about 50 network cables in the way) a NAS drive off a shelf in my cellar, it fell down some steps. One of the two 6TB HDD is now not spinning up (I guess it couldn't park its heads in time). Bummer, need to move some CCTV cameras across to the other NAS.
Yeah, I need two NAS drives for CCTV because I have about 10 4k IP Cameras running. Yes, my cellar has a shark tank in it with a trap door above and yes, they've had lasers attached to their fricken heads since 1997!
I'm surprised that any HDD manufacturer is spending any cash on R&D.
Why?
It's obvious that the HDD is going to lose it's market for mid scale (~100GB to ~500GB) as the prices drop for a long time. However, that just means that HDD's get pushed back into a more niche market of larger storage, where they could survive for the foreseeable future due to the financial cost of a large SSD being massively higher than an HDD.
Nobody working at one of these companies wants to lose their jobs, hence they are going to do what they can to keep their jobs. Failure of one of the companies to survive means their market share transfers to their late competition, and cutting R&D first practically ensures that your the first one to die off. The point that R&D stops and you just thrash the remaining equipment is the very last bit of the endgame, which we are nowhere near reaching, i'd say that it's quite possibly several decades away yet.
If your not sure what I mean by that, have a look at the still going tape market. It started in the 1960's, was made pretty much thoroughly obsolete for most purposes by FDD's & HDD's by the 1980's but yet is still going strong for niche applications (ie, backups) in 2019, 40 odd years after pretty much everybody would have considered it "dead".
Tape has the difference that the media is dirt cheap compared to HDDs which offer similar capacities, and more importantly have significantly higher URE rates. The LTO5 drive I have at home is only 1.5TB and about $20 per tape on Amazon, a 1TB WD Blue HDD is $45. More importantly the LTO5 tape has 10^17 URE rates whereas the WD blue has 10^14. Since tape is relegated almost exclusively to backup storage, URE rates are a big deal. Sure you can find HDDs and Flash with similar URE rates but they are silly expensive.
You can easily compensate for UREs by just including some FEC codes. You'll need some more storage to hold that extra data, but it's manageable. Some backup software has the option baked-in. We're still quite some time away from the end of LTO tapes, but it won't hang around very long after spinning rust reaches price-parity.
it won't hang around very long after spinning rust reaches price-parity.
And the HDD won't hang around very long after the SSD reaches price parity. I just suggest that Tape has now survived 40 years beyond it's mass life purpose, and the HDD will probably be with us (albeit it increasingly a niche) for the same sort of time frame.
Well you're right, spinning media won't hang around for long when SSDs reach price parity with it.
And density parity.
And the wear out issues are solved.
Until all of that happens, spinning rust will remain mainstream and industrial SSD use cases will stay niche.
As for the forecast that SSD revenue will overtake spinning, well that sounds like cobblers to me. El Reg should have quoted Russell Grant, he's a professional fortune teller you know.
Arrays are cool and will always have a place in the data center, but most data centers are also increasing the role of distributed storage. Since the modern data center is basically just a private cloud now, the ability to just shove disks (SSD or spinning rust, doesn't matter) into all those compute nodes and then tie them together is a big deal.
Agreed, to a certain extent except that 'knowing' where and which copies of a particular file are on which nodes can present an interesting problem, perhaps critical when more than one box fails (e.g. power supply zorch!). Given that distributed has always been in use here it's a solved problem in my case and, yes, any real sysadmin worth their salt should already have that solved as well. Built-in actually with enterprise software, more problematic for SMB.
Meanwhile, my system of record is an array with defense in depth to "Cloud." I'm now making the transition from 3 GB WD Reds to 8 GB Reds although the old array will still be around for a while but not relied on as much. The first 3 GB has just died after a really good run since I bought them when that WD Red HDD was all shiny and with new HD smell ;-). Don't expect to see me transition anytime soon to SMR/HAMR/MAMR. I do realize that these are supposedly reliable but, as with SSD for the first couple of generations, I'm not buying into the model. I thoroughly understand the issues with each and am not sanguine. Yet.
"Analysts predict solid-state array revenue to overtake spinning rust and hybrid"
Why do I see the same articles year after year....
Spinning Rust is Dead!
Ah yes life at the office has been interesting since the fall of Email...
that's the next repetitive story line right Reg?
So bored of all these ads posing as Tech Reports...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
