futureofstorage

That's a big number. I'd like to see us get to a better definition of SDS - too many non-SDS solutions out there.

@chrismevans - we're talking just the software, not the gigantic Taiwanese industry that is emerging from the shadows.

@chrismevans agreed it's overused, but @lleung's descrip of SDS as "managing data like software locked inside the array" is pretty good

and yet... $5B seems low if you consider TAM across effective lifecycle of existing storage markets and projections for scale requirements going forward

If SDS also shifts the hw to white box, then the $5B it is likely shifting/replacing at least $10-15B of traditional storage

@stu I don't think it all has to be white box; just decouple hardware performance from storage "object" performance.

@chrismevans "I'd like to see us get to a better definition of SDS" --> +1 seems each company has their own definition of SDS

OK - what do we need for "real" SDS? sw that is independent of hw, simple deployment, 1-click scalability, what else?

I feel better placed than say 2 years ago to create a definition, I see more substance around what performance decoupling means and how we should expect storage to be "driven".

@stu abstraction from the underlying hardware for massive scale effects like protection, performance, self-healing, continuous uptime.

@chrismevans SDS definition 1) storage infrastructure can be configured via API AND 2) it is delivered in SOFTWARE which can be deployed on STANDARD servers.

@stu Not quite - storage object performance that is separated from the hardware e.g. LUN/volume etc gets 1000 IOPS irrespective of hardware speed etc.

@jlecat 100% on the API piece. Management (provisioning etc) driven by API not a sysadmin. But that API needs to be (a) rich (b) native, not a false wrapper around an old product.

@chrismevans The $5B is for the software portion, if you ask how much value of the storage industry is going to shift, it is more $30B or $40B, moving from dedicated HW to standard servers and margin pool reduction (means cheaper for customers!)

kasia @chrismevans Vendors making up their own definition for a storage trend is nothing new: storage virtualization, ILM (remember that one), SDS. BTW, what happened to the ILM company?

@chrismevans - FOP - it's any hardware - super dense if you like (approaching 100 drives in a 4U), super fast if you like, super cheap if you like

@chrismevans Yes Chris on API, my point is SDS needs to be SOFTWARE. HW product that call themselves SDS are just SDS-washing....

@jlecat if $5B shifts $30-40B, is there any way that the current large players can "win" at such different economics?

@chrismevans - false wrapper around old product - stop talking about JJ Abrams like that

@HPStorageGuy - ILM... wasn't that Quantum?

@stu Some large players will adpat quickly ("elephant can dance") others will just die (in the form of merger probably)

I think every storage vendor talked about it but the guys in Hopkinton made it their corporate slogan; can't remember if it was alongside "Where information lives" or before it.

And much to my embarrassment, I was marketing employee number 2 at HP. I learned a lot from that experience.

@HPStorageGuy I seem to remember Veritas talking ILM around 1996.

@jlecat @stu The impact of these disruptions in storage market of all these trends will be as large as the VMware/virtualization driven disruption of server & processor vendors.

@stu - it's like taxis vs. uber, kmart vs. amazon, nokia vs. android, ...

@shd_9 Agreed.

@chrismevans They started with HSM. They may have been the first to use it, don't remember now. I know Compaq was using some other acronym and we changed it to ILM post merger.

Totally off topic, but some taxi twitter account in Barcelona tweeted me that Uber is illegal in Spain. So don't take an Uber ride in Spain or risk jail time! LOL

bigdata and #IoT will drive more storage and cloud usage the consumptions patterns are so clear

no question whatsoever. prime driver of the capacity side. Machine, sensor-generated data - exploding, with good reason.

- ok, fine, I'll say it. object storage is also an enabler, not a market.

the smart guys are all using object under the covers. they just don't talk about it. they talk about the problems they solve.

the only 'eventual consistency' is that of the market -- the technology consideration itself is secondary

Openstack is going thru it's growing pains of consolidation

@lleung yeah, the buyer market doesn't seem to have interest in understanding object storage as a market, but does want the attributes it can provide

@reichmanIT - folks can keep talking about object storage until they're blue in the face. I pass.

@lleung object is too broad a term. Really a simple taxonomy: one type of persistent store is a fixed identified (an LBA) mapping to a fixed collection of bytes (a block)

@reichmanIT Perhaps there's a consumption issue; customers struggle to understand how to consume object when application/interface/programming work is involved.

@lleung Object storage - IMHO real underlying innovation is monolithic Get/Put access interface vs. Posix R/W interface of days past.

@lleung the other type is a variable identifier (name, URI) mapped to a variable collection of bytes (an object)

@lleung file is really a subset of object, with certain app-facing behavior (e.g. POSIX semantics) and persistency behavior (immediate)

@shd_9 - somewhere seagate kinetic guys are smiling

@chrismevans totally - when that isn't abstracted from the customer, the complexity is overwhelming and they just go buy another ultra-dense array.

@peglarr - it takes a file guy to really talk object ;)

@peglarr and perhaps other things like locking, security etc

@chrismevans storage buyers don't typically get down with heavy programming, and programmers don't really think about infrastructure- devops MIGHT be the people shift that allows object systems to get customization they need and increase adoption

@chrismevans of course. but all the bloviating over 'object is new & cool' is silly. It has always been cool, but it is not new. How one implements it - as you said, devil in the details.

@peglarr and the combo of object storage with content addressability is where other value comes out - scalability, load balancing, etc.

@peglarr object goes back to lustre days (and i'm sure you'll tell me even earlier). but it's still not a market.

@lleung I will tell you that, yes :-) Way further back than that. 1960's. Channel-attached h/w with object-oriented access methods.

@peglarr - we need to have coffee sometime #storagenerdsunite