Listen, I’ll acknowledge right up front that the premise of the title of this post is ridiculous. But considering some of the other more outlandish and dogmatic arguments I’ve seen around the topic of AWS and it’s position in the IaaS market lately, maybe it’s not that out of place. Let’s all take a deep breath and see what’s going on here…
Part of why I love Twitter is because it’s an organic place where a conversation can unfold, either in real-time or after the fact. You can catch a snippet of a thread, dig through the multiple conversations that got the discussion to that point and watch the multiple players as they respond. Last night, somehow, the topic was how every IaaS provider and customer was somehow competing with AWS, whether they liked it or not, and as part of that thread the following statement was made by George Reese, CTO of enStratus:
While this was one of the more sound-byte worth comments in the thread, the basic premise (as I understand it) was that AWS has become the defacto standard for all IaaS deployments anywhere, and in achieving this milestone they have set the bar for business model, features, price and functionality in many ways. Here’s another good quote from the thread from Randy Bias over at Cloudscaling:
Now, I’ve never met Randy (and I hope to), but from what I’ve seen he’s possibly the biggest AWS fan on the face of the planet, at least that I’ve ever met. Whether it’s calling them a “runaway train” and predicting $16B in revenues or calling them a “game changer” and “the primary measuring stick for all IaaS”, Randy obviously likes what he sees. Every option, has at it’s core an assumption (or many), and I think that’s where my thinking about AWS and IaaS in general diverges from Sirs Reese and Bias. This notion that the buyer of technology services is shifting from IT to application developers just isn’t one that I see regularly as I talk to both enterprises and service providers. I even forgive Randy and George for picking on VCE and the Vblock specifically, since I’ve learned that’s just what happens when you are the most recognizable and successful entry into any marketplace.
In every one of these discussions, I see the same names brought up as proof-points: Amazon, Google, Netflix, Salesforce.com, Facebook… What I don’t understand is how some of these companies even end up being part of the discussion, especially when the topic is a comparison to enterprise IT. How many times do we have to rehash the fact that Facebook isn’t representative of IT, at least not it’s massive, public-facing side? It’s a single application, scaled to incredible size, ditto for Salesforce.com and Netflix. That’s not IT, where there can be dozens of applications sitting side-by-side, some developed in-house, some bought off the shelf, each one having it’s own development and maintenance cycle. You can’t, I repeat, CAN’T, put that kind of workload footprint in the cloud. And for anyone who says “legacy is shit, you need to start developing your way out of that hole” I tell you to wake up and look at the reality of today.
Would it be great if every enterprise had every application they needed developed in-house and/or on a common platform with all of the resiliency, redundancy and availability built in? Abso-frickin-loutely. THAT would be a game changer. It would change the basic premise under which the concept of application delivery would run, and (finally) put all of the focus on the apps and the customer experience, where it belongs.
HOWEVER, even that drastic a departure from reality wouldn’t change either the overall IaaS market, or the opportunities that exist for the service provider community. Some people are going to want/need to run the infrastructure inside their own firewall, some people are going to want to outsource it. Some people will need to build data centers, some people won’t. Hell, some people will be good at building enterprise-class, “cloud-ready apps” and some won’t, and those that won’t will want help. Hopefully, most enterprises will stay as close to their core competency as they can, and give away those functions that don’t need to be part of the business model.
So with all of that said, could one of these magic unicorn clouds, internal or external be hosted on a Vblock (or any other converged infrastructure stack)? Of course it could. Would a service provider choose to host their magic unicorn service on a Vblock? Of course they could. How is this possible? Because I dispute the basis of the argument that AWS and their disciples put out: limiting functionality, putting the onus for availability and driving the cost of resources to the lowest possible point by using commodity hardware with a high rate of failure isn’t what the enterprise market demands. It just isn’t. Yet. Now, you could (and should as a good skeptic) argue that Randy and George are pushing a business model that ultimately meshes with their respective companies, and that I am pushing one that meshes with mine, and that would be very fair. After all, if you don’t know any of us how do you trust our motivations? In return I’d ask you this: what are YOU seeing in YOUR enterprise? What are you hearing about in others? What are you seeing out of the Federal government space? What are you hearing from your security and compliance teams? At the end of the day, however unwelcome it may be, reality is what it is, and the reality where IT is even MOST of that way towards a completely programmatic consumption mode is just fiction. Chris Hoff with Juniper Networks summed it up well later in the conversation when he said:
I certainly agree, AWS has had an impact on the market mostly because they introduced the world to a new way of consumption, one that didn’t necessarily exist before. For those enterprises who needed it, AWS was a great new answer. You can’t ignore their growth or their level of innovation, and I can’t help but be impressed by both.
Of course, in my heart of hearts, I lament the fact that we live in this reality. Having been an end user, service provider, and infrastructure vendor, I hate that we have to abstract and virtualize our way around the legacy limitations of our collective IT past. No application user has ever cared about the brand of fabric switch or server that was used. They want a good experience with the applications they need, and that’s it. Everything else is secondary to them, but unfortunately it will take some time for the enterprise to catch up and be able to focus the lion’s share of their attention and resources in that direction. Reality is what it is, not what we wish it to be.
What do you think? On a continuum from legacy to fully public cloud where is your company? Where are you moving to be? Does the term IaaS need to have some additional context to it so that we don’t conflate one kind of consumption model with another? Disclosure of affiliations and common courtesy is always appreciated in the comments!