futureofstorage

Future of Storage
Software defined storage at petabyte scale - how will storage change in the next five years?
Leo Leung
https://www.crowdcha... - with all this talk about businesses going to cloud, how big will the storage market be in 10 years?

Chris M Evans
How do you define "the storage market"?
Leo Leung
@shd_9 @storageDiva - this question is for you ^ ;)
Chris M Evans
...bearing mind storage is now dispersing back to the infrastructure.
Stuart Miniman
sounds like the changing hw, sw and service mix will be painful for the current big players $EMC $NTAP
John Furrier
I think that mkt research firms are having a hard time sizing the unified or converged market TAM imho; it's not shrinking that thesis is crazy
Leo Leung
@chrismevans - you mean like x86 volume and optimized servers? ;)
Jérôme Lecat
@chrismevans anything deployed on prem (storage systems, DAS, storage software, storage professional services).... *not* cloud not IaaS, not SaaS, not Box...
Scott Davis
Which question was for me? ;-)
Chris M Evans
I was thinking of new flash players, NVDIMM tech, disk industry, array vendors, hyperconverged players etc, all are selling "storage".
Leo Leung
@furrier - crazy is as crazy does - 3 points: consolidation, commoditization, culling the fat
Stuart Miniman
here's the Wikibon forecast on shifting share from traditional storage to non-array solutions (Server SAN) http://wikibon.org/w...
Jérôme Lecat
@furrier Why crazy ? When there is a value chain disruption industries often shrink: look VOIP vs PSTN, look Amazon vs Barnes and Nobles.
John Furrier
look how Cisco calc share for their UCS they count other non server tech to capture the true architecture but not apples to apples
Leo Leung
@shd_9 - what size is the storage market in 10 years, with Infinio in the world
Jérôme Lecat
@furrier There are no less phone calls, there are no less books being sold, but there is less margin for Voice at Telcos, or for selling books (ask Barnes & nobles)
John Furrier
@jlecat totally agree margin shifts to other value creation tech and innovations
Leo Leung
@furrier - margins erosion, margin dispersion. that is all.
Storage Godfather (HPEStorageGuy)
@furrier Agree - I know are server team is asking those that measure revenue to make the numbers apples to apples.
Scott Davis
storage is being disrupted on multiple fronts, its the aggregate of the disruptions and the myriad of solutions growing from it that is the key trend
Jérôme Lecat
@stu Wikibon's graph is convincing. Take out public cloud (Hyperscale in green on their 2nd graph), and it is flat.
Stuart Miniman
@HPStorageGuy @furrier funny, HP has been playing this for years with the mix of internal storage vs external. Slice that apple and find an orange inside ;)
Jérôme Lecat
@stu Scality thinks it is actually decreasing because in SDS architecture there is more pressure on margin throughout the value chain. Some of our customers do reverse auction on servers and HDD and buy the cheapest. The intelligence is in the software.
Scott Davis
disruptive ingredients: new medium - flash and memory DIMMS, Separate buying capacity and performance, scale out arch, colocation of app and storage, partitioning, object storage, availability/coherence as app responsibility
Storage Godfather (HPEStorageGuy)
@stu True but IDC breaks out internal vs external vs total. So, really not the same.
Chris M Evans
@shd_9 Some of these disruptions are just features though, not new paradigms or technology. What do you see as the real shifts?
Chris M Evans
@HPStorageGuy So how about some new categories? Can we really lump hyper-converged in with "internal"?
Scott Davis
@chrismevans I think the real shifts are: 1) decentralized storage, 2) massive scale and datasets, 3) separation of capacity from performance and 4) application-centric high availability.
Rob Peglar
@shd_9 well said, i would add app-centric D/R as well (as opposed to HA)
Scott Davis
@chrismevans My opinion on Hyper-converged as a category - http://talkingtechwi...
Software-Defined Storage & Hyper-Converged Infrastructure; Two Sides of the Same Coin?
When I talk to customers about Infinio, it often helps to explain the landscape of storage – where the industry is today.  Shortly after I joined Infinio I quickly formulated a 5 category taxonomy for the emerging storage landscape: Hybrid arrays, Al...
Storage Godfather (HPEStorageGuy)
@chrismevans Now that's getting complex. Glad I don't have to define and measure revenue. Technically, hyperconverged is shared internal. A new category?
Jay Cuthrell jay@cuthrell.com +1-415-763-8343
ask any analyst or pundit what % of their practitioner contacts can _accurately_ provide total spindle counts (then ask if they actually 'care')
Jérôme Lecat
@shd_9 Yes, and I would add 5) decoupling of HW and SW and 6) application-centric provisinning
Stuart Miniman
Storage today is over provisioned and underutilized - how do solutions like cloud and SDS attack this issue
Chris M Evans
It's an apparently perennial problem. :-)
Chris M Evans
So let's start with cloud; the utilisation issue simply becomes the provider's responsibility; now the have to deal with the efficiency.
Chris M Evans
Nice for the customer as they pass off the problem. I don't see SDS as really attacking this problem unless the architecture of the platform is capable of making provisioning easier.
Leo Leung
@chrismevans - i was waiting so patiently for someone to bring up architecture. Yes, there's a new architecture in town...
Chris M Evans
So, why should we think that SaaS or IaaS providers could be more efficient at managing storage than customers? :-)
Jérôme Lecat
Two cause two response in 1 solution
Leo Leung
@chrismevans - Some don't have to be, that's for sure
Chris M Evans
@lleung - Architecture still matters and the devil is in the detail, depending on what you're looking to achieve.
Jérôme Lecat
1) Silo effect => solution SDS is scale-out and multipurpose on 1 platform
Chris M Evans
@lleung In many cases I wonder how many customers understand the nuances of a platform until they're too far down the line with a deployment...
Stuart Miniman
@chrismevans the hyperscale guys don't buy storage arrays. Genesis of the hyperconverged/Server SAN discussion.
Rob Peglar
@chrismevans @lleung not new at all. efficient scale-out has been around quite some time now.
Jérôme Lecat
2) 15K RPM cannot deliver enough IOPS to fill capacity => mixing flash (IOPS) and HDD (capacity) in SDS solution solves it
Jérôme Lecat
With SDS at large scale you get virtualy infinite IOPS and infinite throughput for thousands of billions of "pieces of data" (files, object, meta data, blocks)
Leo Leung
@peglarr true enough - many concepts are not new - implementation is
Chris M Evans
Have we seen enough hyper-converged to be happy those solutions scale? Have we seen any real horror stories yet?
Chris M Evans
@lleung Yeah, everything derives from the mainframe... :-)
Stuart Miniman
hey @jbgeorge - any commentary on cloud storage efficiency?
Rob Peglar
@lleung there will always be new implementations. Which is great. Then there are innovative implementations which began long ago & constantly renew themselves as technology/demands change
Leo Leung
@chrismevans - would love to hear positives and negatives about hyperconverged
Chris M Evans
I wonder if the cloud providers are efficient - that's pretty hidden from the customers.
Rob Peglar
@chrismevans many are not. they have traded tech efficiency for economic efficiency. This is a constantly swinging pendulum
Andrew Reichman
@chrismevans cloud providers have to have enough capacity to handle customer spikes anytime, which makes them pretty low utilization any given moment
Chris M Evans
@peglarr I like it when I see a technology that's been re-invented. makes it easier for me to understand!
Leo Leung
@peglarr it's a fun market we live in, right?
Rob Peglar
@lleung still fun after all these years. 0x26 to be precise
Andrew Reichman
@chrismevans but they are very efficient in terms of staff and compression ratios, and they buy at super low price because of enormous volume
Andrew Reichman
@chrismevans their customers CAN be much more efficient because they can turn off resources, but only if they read and understand key metrics and understand the tools well
Leo Leung
@reichmanIT - spoken like a former AWS guy
Chris M Evans
@reichmanIT Absolutely agree - does the customer care as long as the price is right?
Andrew Reichman
@chrismevans exactly- the customer doesn't care if they get good pricing, unlimited access to capacity, reliable service and the tools they need to see how much capacity and perf they have and need
Rob Peglar
@chrismevans for now, probably not. but in the future the enduser consumer will care, because of the implications in terms of inputs used to 'generate' that cheap price.
Andrew Reichman
@chrismevans with buying gear, you're stuck with it when you bought it, so metrics don't matter until refresh time and people over-provision to avoid an outage at any cost
Rob Peglar
@reichmanIT @chrismevans not always true. depends strictly on time-to-deploy. this is one of many reasons why scale-out is optimal
Chris M Evans
@peglarr That becomes the "steady state" cloud scenario rather than the uncontrolled land grab we see now (although cloud will probably follow the ever expanding universe model).
Chris M Evans
@peglarr @reichmanIT - I think there's an operational excellence factor there too; however many problems occur from poor budgeting.