Sometimes, you see a phrase that makes you pause and think. I tripped over this one, courtesy of Justin Warren, who was commenting on the recent VMworld announcements.
The phrase made me think. Thinking is good. Thank you, Justin.
New ideas in the IT world are bright, shiny objects that initially capture our attention. They then become utterly familiar, and the world progresses to newer, brighter and shinier things.
VMware was founded in 1998. It was acquired by EMC over 12 years ago. Next week, EMC disappears and becomes part of Dell. Life moves on. VMware has been very successful in helping to define what "private cloud" means inside a data center.
Amazon Web Services was publicly launched in 2006, a decade ago. It too has been wildly successful, and has helped to define what "public cloud" means outside of data centers.
Both can reasonably be described as "legacy", if nothing else than through age and maturity alone. Both could be described as providing infrastructure as a service, or IaaS. They also can be described as two competing industry forces attempting to capture each other's territory.
When it comes to enterprise IT, I can't make the argument that either is "winning" the IaaS wars. It appears to be a standoff, with no clear road ahead.
So, what might be next?
So here's my list when it comes to enterprise IaaS:
Public Vs. Private? No, Public And Private
Public cloud IaaS services have their strong and weak points. On-premises private cloud IaaS has their strong and weak points. Attempting to hybridize them at scale means you've signed up to be in the systems integration business, as they use fundamentally different models.
Remember when we had IT silos around mainframe, UNIX and Windows? And nothing really worked together all that well? Same story, different chapter.
Both VMware and AWS are doing their level best to capture each other's territory. VMware shows demos of NSX interoperating with AWS. They add container and OpenStack support on top of their foundational hypervisor.
AWS positions itself to be as attractive as possible for cloud-native developers. Lambda, anyone? They promote a variety of migration tools and programs to ship those VMware VMs to their cloud.
So, if we're answering the question "what's next?", one part of the answer might be a single architecture and set of services at both ends of the wire (on-premises and off-premises), designed to be used together.
The only choice left would be the desired consumption model: opex or capex, on-prem or not.
Applications: Past, Present and Future
I always go on and on about how enterprise IT is different than other forms of computing.
One of its defining attributes is -- unfortunately -- a very broad application portfolio. If you've never patiently gone through a large enterprise's production application portfolio, you don't know the huge fun you're missing.
Some of it is prepackaged stuff from another era. Some of it is home grown, maybe with tools from the last decade. Some of it might be described as somewhat contemporary in architecture and tool set. Then there's the bright and shiny world of containers, microservices and "serverless" architectures.
And not all of it can conveniently run in an elastic x86 VM :)
Here's the problem: enterprise IT can't pick and choose which application workloads to support. Rationalization and rearchitecting is a very expensive (and sometimes risky) proposition.
So if all those enterprise application workloads are going to move to a cloud of some sort, that cloud will have to offer *very* broad support for multiple application models: past, present and future.
Lest you think that the future model will be dozens of clouds, optimized for different application models, that just brings up back to IT being in the systems integration business once again, just this time across dozens of cloud service providers.
Not an attractive scenario.
So, as we're contemplating "what's next?", another part of the answer might be having a cloud service with the ability to support a very wide breadth of infrastructure models.
A set of IaaS services that reflect the harsh reality of the typical enterprise application portfolio and its challenging diversity.
Control Planes Matter
Another key area where IT is different is their mandate to control the enterprise computing environment. Security management, application monitoring, resource consumption, financial management, and so on.
It doesn't matter where the computing is actually done, what does matter is that enterprise IT organizations are responsible for outcomes -- and being in control. To do this, they invest in people, process and tooling.
Now consider most public clouds: they have their own way of doing things, don't they? Once again, we're back to this unpleasant scenario of multiple, competing control planes as well as investing in significant systems integration to hopefully make it all work together again.
Back to our rumination about "what's next": how about the ability to easily extend current and future enterprise IT control planes with a public cloud service?
Workload Portability?
One thing that is endemic with many IT professionals is avoiding the feeling that they're locked in. Yes, there's always some friction involved with moving something from A to B and then C. But I would also argue that there's a clear opportunity for friction reduction.
On-premises virtualized workloads should be relatively easy to move from on-premises to IT's choice of public cloud provider. And back again, if desired.
Containerized applications should be able to be moved across public clouds, or be run on-premises if that's what is desired.
Once again, as we consider "what's next?", part of the answer could be around seriously reducing the friction associated with these tasks. I am a realist, though. Friction can be minimized, but not eliminated. And various forms of perceived lock-in will be with us for a very long time.
Moving Beyond Legacy Cloud
I don't think VMware will ever be a serious player in the public cloud, despite their desperate desire to do so.
They're a software company; running a cloud at scale necessitates a completely different business model. Heck, I'd like to be a long distance runner, I just don't have the body for it.
Maybe they can recruit folks like IBM to supply the hardware behind their software, but that's essentially a marriage of convenience, and certainly not a strategic play.
I don't think AWS understands enterprise IT, and what the world looks like through that lens. Given their growth rates elsewhere, maybe they don't need to.
AWS and their services can certainly be part of the answer, but I just can't see them being "the answer".
I believe that there's a clear gap in the market when it comes to IaaS and what's next.
Simply put, the world needs an IaaS that works the way enterprise IT does.
Warts and all.
--------------
Like this post? Why not subscribe via email?
Hi Chuck,
I've followed your blog for probably 2 years at this point. Probably my two favorite blog posts of yours have been the following, with the first one I think kind of / sort of similar to the tone of this current post.
http://chucksblog.typepad.com/chucks_blog/2014/06/dear.html
and
http://chucksblog.typepad.com/chucks_blog/2014/06/never-waste-a-good-crisis.html
I actually wrote a blog post, almost entirely fueled by being forced to make Azure work for our DR needs (it doesn't do the job very well.). Its here if you or anyone else is interested, and that's a rather tame view of my thoughts and struggles with public cloud http://www.ericcsinger.com/thinking-out-loud-the-cloud-iaas-delusion/
The public cloud right now isn't a good fit for enterprise unless the enterprise is willing to change their practices, architecture and everything else to fit the cloud world. Not to mention there are still plenty of businesses that prefer a capex model, and cloud doesn't fit that model well.
I also find arguments like Justin Warren's to make me roll my eye a bit. When folks "assume" that because we run vmware, we're a "point and click" group is just pure nonsense. Even implying that that is a bad thing BTW is also nonsense without full context (see my first post about the CLI vs. the GUI). I do so many things in PowerCLI and am pretty much able to do everything related to VM deployment via PowerCLI that I can do with an AWS or Azure API. The one thing I don't have is unlimited HW, I don't lack the ability to automate with VMware.
Posted by: Eric Singer | September 06, 2016 at 12:38 PM
Here is the arithmetic of clouds from three years ago:
https://blog.veoci.com/cloud-computing-its-smaller-than-you-think/
Posted by: Sukh Grewal | September 07, 2016 at 10:33 AM
Hi Chuck,
I love this post, and you bold statement that basically both VMware adn AWS are legacy! Wow. I am amazed at Werner Vogels' drive for innovation, but I totally see your point.
I also (selfishly) love the conclusion "Simply put, the world needs an IaaS that works the way enterprise IT does." Indeed this is the conclusion we came to at Scality, and our latest software is all about Enterprise features, like directories, encryption and how you manage rights and keys, and about "multiple" clouds, not public, not hybrid, but any cloud.
Jerome
Disclaimer: I am the CEO of www.scality.com
Posted by: Jerome Lecat | September 09, 2016 at 02:29 AM
Hi Jerome, always a pleasure to hear from you.
Bold statements? Me? Never :)
Seriously, though ... AWS grew fast by targeting developers, growing web companies and non-enterprise IT computing applications.
My argument is simple: if you want to gain share in the enterprise IT market, you have to deliver a product or service that is aimed at their needs.
VMware has the opposite problem in my opinion. They delivered a product that is used in almost every data center. But in a world where people don't want to own data centers any more, they don't have a good answer.
I hope all is well for you and the Scality team!
-- Chuck
Posted by: Chuck Hollis | September 09, 2016 at 09:33 AM