Dog walkers, the San Andreas fault ... and the storage industry Mainstream storage arrays are being harassed by a wolf-pack of sleeker competitors. Looking at storage is like looking at a dog walker on the beach with a hundred howling dogs of all sizes and breeds on the leashes. Some are puppies, some young and sleek, some old and fat. They bark and growl and run in all directions …
Nice piece Chris. Just to amplify the point, I would note that the magnitude of the installed base under threat is not just billions, but a number of *tens of billions* -- if you figure overall industry revenue has been approximately $20-$25B/year over the last few years, and multiply that by ~3y expected service life -- well, it's a pretty big number. It might not be exactly $75B because, frankly, due to competition and technology improvement, customers get more for their money these days. Call it $50B+.
G
The problem is that you are thinking storage and not data. Storage is product, data is business problem and asset. Technologies and products come and go and you don't want to tie your data into any particular technology or product - at least if you plan to run your business for the long term.
Server storage, hyperconverged, even clouds are probably younger than your business and many of them are most likely gone before your business go. Server refresh cycle is totally different from data lifecycle - do you really want to tie them together and turn every server upgrade into data migration? Sure, stay with one vendor and it may work until the vendor goes south. Or if you are google and can write and manage your own.
It just makes sense to separate data from processing so you can treat them separately and avoid lock-in. There will always be point solutions that do something better and look fresh and sexy, but are they really solving data management issue or just trying to sell a new product?
Honest question: is moving your data once in a couple years really that big of a deal?
From where I sit, a lot of customers are forced to move portions of their data *much more frequently* due to issues of capacity, performance, availability (vendor bugs), correctness (vendor bugs), forklift upgrades (same vendor but cross-architectural boundaries), and more.
You are describing the problem perfectly. If you think products, not data, you end up buying into the latest hype and then waste your time on eternal changes and upgrades.
IT guys are paid to play with their gear so I'm sure data migration is not a problem for them and new toys are exciting anyway. Also keeps them employed until business gets weary writing checks and suffering service breaks and out-sources the whole mess.
You're joking, right? Even for small shops, moving data is very painful and expensive. It's not just the migration of the data itself, it the coordination of the event. Project management, application owners, sysadmins, project timeframes, maintenance windows, possible outages, and the evil that is Professional Services. I'd argue that in many cases, the OPEX cost of all the people and time involved in storage migrations is more expensive than the new storage itself. IT staff often don't think about that, but CIOs and CFOs do.
Then you're doing it wrong. That scenario follows naturally for small shops where you can normally buy 5+ years of storage when you install the new storage system. They learn all those lesson, then have 5 years to forget them.
The reality is you have to keep moving data storage. I have a friend who works in a place that should have known better, but they archived data off onto 3.5 inch floppies back in the day. Data was small (less than half a floppy), floppies were cheap, and everybody had always used floppy disk drives. Ten years later when they tried to read the data it was gone. Millions of dollars worth of research and they've lost the data. If you aren't constantly migrating your data to active, redundant storage, you will eventually lose it. Maybe Software Defined storage helps that. If it does, more power to it. But my experience is that every time you obscure the inner workings of something, you set yourself up for failure.
They hang out about the Ranch, and do their jobs as needed.
Kinda the same as my storage systems.
It's my job to archive all of that ... and I've been pretty good at it for these last forty years.
I don't think I need a third party to help me do what's obvious.
Thank you Chris, great article as always and for highlighting DataCore’s part in creating the category Software Defined Storage 17 years ago. Just to correct one point, of the 10,000+ installed, we do have a large enterprise base (not just SMB) installed in some of the world’s most demanding environments, securing in some cases, billions of transactions. Check out http://www.datacore.com/testimonials for lots of enterprise examples.
Also I have seen some of the comments about “Think data, not Storage” and I can tell you from our 17 year journey in the high tides of the storage seas that migration of data is the most demanding permanently feared kind of lurking monster hiding under sea level to eat you alive you if are no careful. We have the power of software only to allow our customers to move their data freely between storage types and vendors at any time without even noticing it at the application/user side. So this whole 5-to-5 year hassle of moving data from device A to device B becomes a no brainer and you look like a Dog-Walker that has a well-trained pack of dogs following him like a charm.
Regards,
Alex
Ugh, another meander through Mellor's alternate reality, where everything is seen as 'proof' that storage vendors will disappear any-day-now and that soon no one will want dedicated storage systems -- he's like one of those Jehovah's Witnesses that sees everything as proof of the immediacy of Armageddon and Judgment Day.
Dedicated storage systems exist because of the fundamental difference between the scale of storage that can be attached to a server, and the scale of data assets held by an organization -- the difference is not doubled or even a single order-of-magnitude, but rather many orders-of-magnitude. Sure, we're seeing talk of 16TB drives now, but they are still puny relative to data assets, which are transitioning from PB-to-EB scales.
Even the fastest storage available is spectacularly slow compared with CPU speeds; a sub-millisecond flash layer is still many orders-of-magnitude slower than modern CPUs approaching sub-picosecond timescales. That architectural delta has always existed, and there has always been a cascade of storage layers (of varying scales) needed to bridge the affordable to the practical. While all the technologies continue to evolve, that gap hasn't narrowed, and is not expected to narrow. Show me a forecast for economical sub-microsecond storage, I'll show you one for a sub-femtosecond CPU.
Then, apparently, Mellor's purple microdot kicked in: "...the future of networked primary data storage is going to be written in 3D Flash and NVMe fabric ink, slowly taking over from classic Fiber Channel and iSCSI" -- Wat? Although it is reasonable to expect 16Gb FC and 40Gb iSCSI to be properly eclipsed in time, it won't be 3D Flash or NVMe 'fabrics' that do it, neither is remotely positioned as the 'future' of networked primary data storage. It would be daft enough to declare 56Gb IB as 'The Future', but when Mellor wants to sound loopy he doesn't mess about, does he?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
