I want to thank some people for all the things we did in 2014: 1) my colleagues amongst the SE leadership at EMC;2) all my colleagues amongst the senior leadership across the Federation set of companies; 3) my great friends in the community - partners, competitors, all; 4) most of all - my brothers and sisters amongst the thousands of EMC SEs as a whole.
Beyond that - thank you for all your friendship through 2013. Heck, competitors - thank you for keeping us on our toes, and keeping the tango fun :-)
Things start to slow down about now. It's a more humane pace. While I love moving fast, I love learning furiously and start to get antsy when I'm standing still, I've always found that spending time with my friends and family, unplugging, reading a good ol' fashioned book is part of recycling the batteries.
Then, part of the transition for me is that starting on Jan 2nd I always dive deep into a new topic, a new technology - and that's the OTHER part of my formula for recharge and recycle.
People - it's a time for peace, a time for reflection, a time to consider how we can make the world better. Let's chill out.
Looking back on my predictions from 2014 is an interesting tour.
I think I would say I was right on 7.5 (#10, #8, the "Microsoft" half of #7 , #5, #4, #3, #2, #1) and wrong on 2.5 (#9 - lack of backlash is shocking to me, the "Oracle" half of #7, #6 - AFAs are having more of an impact than SDS).
Not bad! I hope that give me some (hah!) credibility. So - with that said, read on for my 2015 Predictions!
#1 - Software, specifically Open Source Software will continue to eat the world.
This isn't a prediction, but a statement of the blazingly obvious. It's been surprising to me to see how much and how quickly I'm seeing this in the enterprise customers I talk to. What's perhaps more interesting is that the "why" isn't clear to everyone (certainly not me). People talk about "lock in avoidance" - but that's doesn't hold up: picking a given distribution of code for enterprise support models has about the same "friction to move" as "closed" software. BTW - I also have come to hate the phrase "lock-in" as it's meaningless when you dig into it (@vmtyler has a great post on this topic I totally agree with here). I think that the real reason is actually the community itself that innovates and iterates more quickly than you tend to get around a "closed" product. This is why I think that 2015 will be the "year of Open Source" in EMC*.
*This is already overwhelmingly the case in parts of the Federation (Pivotal is the largest contributor to a very dynamic, very open Cloud Foundry Foundation, VMware has been one of the biggest contributors to swaths of Openstack).
We have some areas we are intense about where we need to create (or overwhelmingly contribute to) a community - and I think we're going to come to a rapid realization that we need to participate in Open Source projects far far more than we have in the past. I think this is true in the persistence elements in Openstack, in the hardware/software interface in Open Compute, and to the broad "Software Defined Storage" control and data plane ecosystem.
HOWTO Measure: The revenue of the top 10 OSS players (in software, support, or blended business models) will double in 2015.
#2 - People will deeply internalize that "Pets" are "Pets" and "Cattle" are "Cattle" - and ne'er the 'twain shall meet (because they are totally different and not twins at all!)
The discussion "de jour" with many customers is how they can apply the infrastructure architectures and operational models of "companies born in the digital age" (think Google) to their existing applications. The answer is that they CAN'T - not without changing the applications. "Pet" class apps simply depend on far too many specific demands in their "support ecosystem".
Smart customers are starting to realize that this Gartner "Bimodal IT" isn't a load of s@#t, but is a good characterization of the split persona that is required: 1) being a "rock" for "Pet" workloads; and 2) being dynamic and fluid for innovating quickly in support of "Cattle" workloads. Both sets of workloads will be on/off prem. Both sets of workloads are important.
Being intellectually rigorous starts to pay dividends. When you look at a "Pet" workload that has a transactional Oracle Database at near the bottom of the stack - the smart questions to ask are:
- "can I/should I replatform this using a modern app framework, a modern data fabric, and a 'design for failure' approach" OR..
- .. "should I instead put it on 'maintenance mode' - where I use brute force like better kernel mode virtualization and all-flash arrays to incrementally improve and save some capex/opex?" OR.
- . "should I put this app on 'life support'/'EOL' to materially save some capex/opex?" AND..
- . if 2 or 3, how do I vector those capex/opex savings to innovating around some other app than this one that I CAN deploy using a "Cattle" set of tools and CI/CD methodologies?"
This is patently obvious if you don't force yourself into thinking that there is ONE WAY to do things (this is where many folks lose the plot).
HOWTO Measure: Difficult. We'll just know :-)
#3 - The Open Data Fabric world will really put the hurt on traditional players (Oracle).
This is a dangerous one, because Oracle keep proving me wrong :-) I just really struggle to see how they can keep dodging this simple bullet: if you want to build a really, really fast transactional app that is built like a tank and you are not constrained by a legacy app codebase. you don't put it on Oracle. You don't put it on a VMAX (or XtremIO) for that matter.
You would use a modern open source data fabric, in all likelyhood one that may be a in-memory database model. You would build resiliency far above the infrastructure and data fabric layers. Yes EMC (through Pivotal and EMC II) play here, but that's not the reason I'm making this point. It's one of those fundamental disruptive changes.
In 2015 (and even more in 2016/2017 with Intel Skylake really hitting it's stride) I expect this in-memory disruption to continue - and we will start to see the first real NVM models beyond NAND hit, AND the first use of "tiered memory architectures" in OS/in-memory software stacks. These are all accelerants to the core trend.
So - the core question is "can Oracle disrupt itself?" The answer of course is "yes" - which means the REAL core question is ".will they?" Everyone has to do this - we're navigating that at EMC II - with our "emerging" technology stack (Isilon, XtremIO, ScaleIO, ViPR, Data Domain) being roughly 60% of the revenue (and growing at 50% CAGR) that our most mature technology (VMAX) - but WOW - these transitions are TOUGH. Ironically - the more solid your "legacy" (and I would argue there's little as firmly embedded as Oracle), the harder it is to navigate the inevitable "innovators dilemma" inflection point.
HOWTO Measure: Oracle starts to miss materially in the back-half of 2015.
#4 - The All-Flash-Array market continues to be on fire.
While all the "sizzle" is in platform 3 and building apps in new ways - the all-flash arrays are Tylenol. They aren't curing cancer, they are instead an "easy button" for making platform 2 workloads that expect a solid persistence layer presented as a LUN (and less every day as a transactional-POSIX filesystem - an interesting reversal of a trend towards transactional NAS we saw several years back).
The number of apps that can't practically be re-platformed, but just need to run faster, more easily - well, that's enormous. For perspective, as noted here, the XtremIO business has grown faster than any other technology in the EMC federation.
While not noted on the chart, XtremIO has grown WAY faster than the leader in "Open PaaS", and that's coming from someone that thinks that Cloud Foundry is strategically critical.
And while 2014 was a breakout year for XtremIO, rapidly becoming the #1 player in the AFA category, the bigger backdrop is how quickly all AFAs are growing. They are still small relative to the Hybrid array market - but extrapolate out a couple years, and wowza - the market is huge. Anyone who isn't investing a ton of R&D organically or inorganically is going to be, well - hosed. Also in 2014 some in the industry realized that this storage world is non-trivial to enter (particularly in Type 2, Type 3, and Type 4 storage architectures - Type 1 architectures are relatively simple to get right, though people debate the desirability of that scaling approach)
HOWTO Measure: We'll know for sure if XtremIO crosses $1B in revenues (or is even close!) after 2 years of being in the market.
#5 : SDS will disrupt the storage industry (ultimately, even more than Flash).
I'm doubling down on this wager from 2014 where I was wrong, but will learn from my mistake, and moderate a little.
First - what is the same as my 2014 prediction? A: core premise. While there is a ton of disruption in storage land, the big impact is not from the AFA and flash media transition (though that is huge), it's around new architectural models. By that I mean software only data planes - across use cases of transactional hot data (think ScaleIO and in 100% VMware use cases VSAN), software-only NAS, software only object stores that are behind the huge content depots and new applications. But, why, if the pure SDS market is smaller than the AFA market, why do I think that SDS models are even more disruptive?
Because they disrupt the core BUSINESS - not just the tech. They have a different economic model (high margin software at lower top-line revenues, with a "bring your own COTS hardware" element). They mean that new architectures (hyper-converged) are possible. They mean that people who are big players - but not storage vendors - can try their hand in the game.
Second - what is DIFFERENT than my 2014 prediction? A: timing and magnitude. While VSAN and ScaleIO were in the market for almost all of 2014, it was only near the end of the year where I saw the volume of activity and deployments ramp up furiously. It takes time for people to adopt something. The other thing that surprised me was how many people turned right around after saying they wanted "software only", and asked for software + commodity hardware bundles (which in effect, is the SAME as the appliance model) once they realized they wanted a clean, clear support model.
HOWTO Measure: we'll have enterprise customers who have more than 1PB on VSAN and or ScaleIO deployed, and we'll see the SDS "appliance" business be north of $1B.
#6 : Certain debates will move into the "idiot detector" category.
Every year, I find that some people are having the same silly debates over and over again. In 2009 it was "NFS vs. VMFS". In 2012 it was "Public Cloud vs. Private Cloud". In 2013 it was "openstack vs. VMware" (and near the end of 2014, "kernel mode virtualization vs. containers"). These invariably are silly, polemic arguments - often led by either: a) press trying to drive eyeballs; b) analysts trying to get people to pay for opinions - and a "it depends" is a struggle to get people to pay for; c) investment analysts trying to drive speculation, which drives equity value up/down; d) vendors driven by either "narrow mindedness" or "an agenda"; e) the honestly curious, but slightly stupid - wanting the delightful simplicity of a choice being right or wrong.
Let me put out the list that to me are starting to auto-group people into a/b/c/d/e:
- "VMs vs. Containers." - the answer is so obviously an "and" (along with cluster management options!) that I feel silly even spelling it out. Umm - do you have any workloads that aren't using Linux? If yes, then the answer MUST be an "and". Beyond that hypervisors vs. containers each bring something different to the table. The one thing that I do tend to believe is that most, but not all cases that need a kernel-mode hypervisor abstraction don't ALSO need a container abstraction (and vice versa) - so that for a given app and workload, VMs and Containers are an "OR".
- "Public/Private" - the fact that anyone is still having this debate befuddles me. It's an AND. The question to debate in 2015 is "do you define cloud as IaaS + PaaS + SaaS?" (you should) and "do you abstract IaaS with an multi-cloud CMP or with an open PaaS model?" (or both).
- "Pets vs. Cattle" or "P2 vs. P3". People tend to get religious about "what we do in P2 matters!" and "P3 is the only way to go!". Breathe deeply. P3 is how you should build any app/infra stack where you CAN (and don't have baggage). P2 optimization (things like kernel mode virtualization, automation, flash, SDS, SDN) is how you PAY for all that cool P3 stuff :-)
- "Converged vs. Hyper-Converged". Both have a fit. Both scale differently. Both have places of strength, and places of weakness. Neither are webscale/rack-scale/hyper-scale, though that's a 3rd "branch" aka "phylum" in the "converged infrastructure" domain. Expect lots of debates on this topic, but there's only one real debate (IMO): "are you crazy enough to think that you drive value by hyper-engineering your own CI stack vs. doing something more valuable to your business (whatever that is)?" <-that's an argument that is loosing more and more every day.
HOWTO Measure: Chad's blood pressure goes down :-)
#7.Solutions construction looks more like "The Avengers" than like "Iron Man".
I think I **might** be too close to this one, but I constantly struggle where people want simple, clear organization models (a rep/SE pair, or a product team hyper-focused on one thing only). Increasingly, getting ANYTHING done requires a nimble, autonomous, highly motivated group of people - each with their own super power. When I look at when a customer really partners up with us, and deconstruct that success - it's a crazy cross-functional team effort. If you're a rep saying "I need MY SE" - you might just be a dinosaur. If you don't trust that people who don't report to you are doing good things (even though their motivations may not be 100% aligned with you) - you might just be a dinosaur.
Yup - focus and "being a one man hero" like Iron Man is nice, but man - it pales compared to what you get when you call out "Avengers: Assemble!", and bring together a group of rock stars to accomplish something together.
*BTW - I still want to be Iron Man in any team. Tony Stark, you're my hero :-)
HOWTO Measure: EMC will have more specialist roles (including those that focus on solutions, or specific vertical markets or GTM like partners, SPs, etc) proportionally at the end of 2015 than entering 2015. I would wager the same will be true of other complex, portfolio/solutions companies.
#8. Federation will need to transform (frankly so will everyone)
Continual transformation is the nature of reality. Call it evolution :-) If you don't keep moving, changing you by DEFINITION will eventually be obsolete. There's a ton of speculation about the Federation model, and I'm not going to add to it here.
I'll tell you what I hear - customers like the idea of best of breed. Customers like the idea of choice. These are built into our highly autonomous Federation structure - and yes, it's awesome in a couple dimensions. It breaks apart work which increases parallelism across Pivotal, EMC, VMware. It creates a built-in "failure to self-disrupt" safety net (very unlikely that all 3 would fail to self-disrupt in parallel). It creates a lot of innovation as each entity is largely autonomous. It also gives each part focus - and yet can come together (see my #7)
That said - customers are also asking us more and more for outcomes vs. parts. They are asking for big solutions, not component elements. This applies pressure to be "more together".
This works if there is a joint strategy ultimately - and there is, but there are some places where we increasingly have a slightly different view of the world. People will jump to the conclusion that I'm talking about the SDS domain (VSAN/ScaleIO, when/where SPBM vs. ViPR controller) - which would be wrong.
The hairier question is how we see future app stacks being built (and the infrastructure layer that supports it). We'll navigate this together as a group - I'm looking forward to the Federation leadership meeting in January where I will see my VMware, Pivotal, and EMC brothers and sisters from around the globe.
HOWTO Measure: Federation structure will be different in some way by the end of 2015. Could be closer, could be continued "partner model" with elements changed, could be more loosely coupled.
#9. There will be a big flame-out in 2015.
I love the Speaking in Tech podcast (can't wait to hear the new format guys!) and they do a great "annual predictions" episode - but I was grumpy when one of the guests said this one :-) That said, I still will put it in my list. I think people don't know how much pressure there is in the IT ecosystem. Assets that use to be worth a lot are rapidly decaying, and new stuff is appearing everywhere. Companies everywhere are having their fundamental business models are under crazy pressure. Pressure either creates diamonds (perfectly handled, or with millions of years), or creates explosions (when slightly less well handled).
I have **NO IDEA** who this will be (and will be working my tushie off to make sure it's not EMC - that we instead make diamonds) - but I think we'll see some "Crazy Ivans" indeed.
HOWTO Measure: This one is obvious - a big, obvious flame-out.
#10. We will see amazing things that no one expects - including me!
This is always my favorite part of ringing in the new year (except seeing my family!). Whenever I toast au revoir to the last year, and bonjour to the new one - I think to myself of all the great surprises, learning, growth that were UNEXPECTED in the previous year. In many cases, things that looked like grey thunderclouds had silver linings. This is a prediction I make every year. The unanticipated, the unplanned, the suprises - they are priceless, and one of the best (scariest!) part of our shared human experience!
With that, THANK YOU for your readership, THANK YOU for your friendship, THANK YOU for your dialog/comments/arguments, and see YOU in 2015!