futureofstorage

The amount of 100 petabyte consumer cloud conversations I've heard lately is ridiculous

And I bet SMB will leverage services like Salesforce, Office 365, and Box rather than AWS. @reichmanIT any thought ?

3 Votes Vote

do you have a thought on the demarcation line for owning most of IT vs pushing it top the cloud (obviously most use both today)?

@stu I think that when you spend more than $1M/year in storage you will think of operating your own IT, when it is bigger than $10M definitely.

I think you will see the majority of SMB's going cloud, after all, they can already pick up many of their SaaS services through the cloud and quality continues to improve.

business size is definitely one. newer businesses don't even think about on-premises

@jlecat customers always want SaaS instead of IaaS if it meets their needs and is secure- with AWS you trade servers for cloud servers, but still have to manage them- with SaaS you solve business problems

@jlecat just look at #DF14 - 140k people attending w huge cloud ecosystem. SaaS takes over 1 app at a time.

There's also the question of trust; as SMBs start to use and trust cloud services, the adoption increases.

3 Votes Vote

John Furrier

cloud was born for SMB with the elimination of the datacenter. We have no datacenter at all no servers or other software that needs attention; lowers cost for SMBs

it's not just SMB either- enterprises are happy to get SaaS as well if it scales- how many fortune500 companies use Salesforce- how many manage their own payroll?

So who do the SaaS players buy from then? ;) They can't all come from Carnegie Mellon...

Once you go cloud, it's hard to come back. Sort of like Hawaii...

@reichmanIT - I think there is violent agreement about SaaS and Cloud here. Game over in SMB.

@jlecat don't those with the biggest IT budgets have the most reason to consider moving off legacy environments to gain new economies? paradox that the smaller guys do it, but those with most $ don't?

1) the biggest DIY and buy from HDD manufacturers, that's a lot of margin going away from the Storage Systems industry

AWS is super eager to find and sell service to SaaS players, and do w dropbox, heroku, icloud etc.

SaaS startups are crazy to build datacenters, but when they reach scale they think they can do it cheaper in house. Some can, but most are deceiving themselves about that and should be shopping IaaS

@jlecat That leads to next question - What does the storage industry look like when that happens en masse?

2) the others - most of them - are like the rest of us. They need to find suppliers... (We're here :-))

@reichmanIT - Sure, until the SaaS player grows out of their toddler clothes

@reichmanIT 2) and yes Andrew, they can also outsource to AWS....

storage industry sels converged software & hardware to solve mostly OLTP - virtualization / SDX allows buyers to pair targeted software w cheap hardware and solve problems cheaper-

but building a custom solution is harder and enterprises that struggle to hire geniuses are probably better off waiting to buy a vendor packaged solution aimed at solving their problem

@reichmanIT - if you talk distributed systems, there really are only like 1,000 geniuses in that world

@reichmanIT - maybe fewer

agree- so that points to only a few players that can attract talent building software/hardware and the rest buying products and services from them

@reichmanIT I think I can name them

@StorageOlogist says that cost of flash figure assumes you fo not have the same data reduction on traditional drives. Not a real compare

there will probably be some small smart innovators that contribute, but most will get bought by the big fish we all know very well

@stu the deduping all flash guys say that they can trade some of their perf boost to power dedupe without too much latency that you would get with disk systems running dedupe on everything- not clear yet if true

@reichmanIT That really depends on architecture. Dedupe with ASIC with 3PAR doesn't have much if any tradeoff. But there's work for us to do for dedupe of HDD.

@HPStorageGuy it's tough to retrofit dedupe- big question is whether you use analytics to tier among flash and disk OR use all flash and dedupe everything- always tradeoffs

@reichmanIT I meant there isn't a tradeoff with 3PAR when we dedupe. Some that dedupe inline have to turn it off when IO load is high. But yes, everything has tradeoffs, except 3PAR (sarcasm).

@HPStorageGuy no tradeoffs? perpetual motion? cold fusion? low calorie donuts? KC Royals in the World Series?

@reichmanIT LOL - well, I know one tradeoff there was Angels aren't in the World Series.

Stuart Miniman34

At #vmworld the storage discussion is flash-centric, what is the balance of flash/capacity today and going forward?

[+] Show Hidden Comments

John Furrier

the balance is about latency, throughput, and cost per gig and knowing the workload

I think at least 25% of the show showed flash related products

75% in booth square footage

We had flash as a feature (caching), we had flash as a product (SolidFire etc). Then of course we've all flash versions of of hyperconverged.

but that doesn't mean usage

it will be variable. enough flash to enable latency-sensitive apps, enough capacity to enable the biz competitive advantage

@chrismevans it's at least 50% of the buzz, along with hyperconverged and SDS. Flash is important, but has it gone too far?

according to IDC 1.8% of Storage Systems capacity is flash.

It's a feature!! Until the next greatest thing comes along.

@peglarr - agree - flash is an enabling technology... only

- big buzz, low actual capacity shipped, let's move on ;)

Jerome, that's a fair point - lots of hype/noise around flash; not necessarily the area with the most sales.

there is also the biz choices between ephemeral store and persistent store. this is a biz decision that tech will (always) evolve and enable the choice

flash accelerates perf, but is expensive- most data sits idle and doesn't need it- accelerating with the smallest amount of flash, or shrinking footprint w/ all flash remains a challenge

Scaity deployments typically have 1% of capacity in flash, for a 10x increase of performance

@reichmanIT indeed expensive, the 1.8% of capacity is 20% of the cost, again according IDC

CIOs must stop thinking cost/TB and start thinking investment/biz outcome. Data and its analysis enable biz outcomes to be optimal

Our hypothesis - 20% flash / 80% SDS capacity oriented going forward

@jlecat as with any expensive accelerating technology, the key is analytics that allows you to use it where it wld add the biggest benefit and archive the rest

ratio will vary wildly depending on vertical and use cases. Difficult to generalize.

@lleung 100 PB is small in some cases. I have had several 1EB conversations over the past 12-18 months

for my $$, the analytics described by datagravity has more transformative potential than noise around flash, SDS & hyperconverged, but is a harder problem as indexengines & Kazeon showed

@peglarr indeed - exabyte is the new petabyte

@lleung it's the Double Olympic phenomenon. Every 8 years the capacity of a single system rises 1,000x.

@peglarr - point is - too much conversation about flash, not enough about the bigger shift in storage

@lleung in 2000 the industry was designing 1TB systems. In 2008, 1 PB. In 2016 it will be 1EB

Not true. _Some_ storage - depends on architecture - suffers from OP/UU. Others do not. Not having to pre-allocate has been around for decades, not a new idea.

With dedupe ratios, efficiency, etc, cost of flash is at less than 15K drives (at least with HP 3PAR it is). Hearing customers switching from HDD to SDD because of this.

@HPStorageGuy Then we will have tiered flash within the same array and start the cycle again.

Do you mean we'll tier flash using different ($/GB) flash drives? I can see that coming.