Public cloud is top of mind for many folks in the IT industry, particularly with the recent Amazon Web Services (AWS) re:Invent show in Las Vegas. People attending the event and reading news about it were given the impression that "everyone" is moving to the public cloud. There is no choice. Even the location, Las Vegas, is meant to seduce you into this thinking.
However, let me beg to differ. Actually, I don't actually have to beg. I see choices, and you should too. In fact, it is common that, at the apex of hype of one direction, the pendulum is quietly starting to swing in the other direction, from public cloud to private cloud.
You might say, “sounds interesting, but give me an example”. Well, how about Dropbox? It started on top of AWS. Its service is a natural fit to AWS, a service on top of storage and networking that AWS is happy to provide on a worldwide basis. However, Dropbox recently invested significant engineering to move their service off of AWS, and now is largely controlling their own IT infrastructure.
What is reversing the pendulum?
Let's start with economics. If you are a startup, or if you need surge capacity, AWS can make perfect sense. You can quickly get data center capacity without a lot of investment or delay. However, if you are an ongoing enterprise with significant IT demands, the cost of AWS adds up so the CAPEX is higher than just buying the hardware yourself.
This is the same as with transportation. If you only occasionally drive, Uber makes perfect sense. However, if you are driving 24/7 like many IT operations do, owning your own infrastructure is far cheaper. Just do the math for your own organization.
But wait! Let's not forget about operational expense (OPEX). Amazon can amortize the cost of running AWS over millions of customers whereas you cannot. Well, if we dissect OPEX, almost all of the cost is people. The only way to get economy of scale with a people-intensive operation is automation. That's precisely what Amazon has done in AWS. At any reasonable scale of IT organization, most of the economy of scale Amazon has over you is by using automation. They automate the way servers are provisioned, they automate the way systems are monitored, and they automate portions of the network management. How did they do that? They hired a bunch of high-priced, smart software engineers to write the automation software.
Now, there is the rub. You cannot afford to, and probably do not have the tech-appeal to attract these folks even if you could. And, honestly, would you have the expertise to recognize who is good and who is not, even if they showed up on your doorstep — not that this is likely to happen?
And, internal development of this nature is fraught with dangers. I am aware of various internal efforts that have left companies to depend on a pile of inscrutable, unsupported PERL scripts. Locked into your own proprietary unsupported solution is not good. It is not the answer.
So, you might ask, how do you tackle this automation problem? AWS and none of the cloud vendors are going to open-source their automation software. It is their core technology and competitive advantage, when you think about it.
Well, a generation of companies are emerging that provide automation software for private clouds that is comparable if not superior to that developed by Amazon's pizza teams. Apstra is one of these, focused on data center network automation.
Apstra provides the Apstra Operating System (AOS), a data center distributed operating system, that "operates" your network for you. It allows your network folks to specify at a high-level their "intent", how they want the network to operate, and AOS takes care of the details of specifying configuration (including cabling) to all the network elements to achieve the intended behavior.
Moreover, it monitors the network in real-time to ensure that the network is operating as intended. Also, if you need to take a switch out of service, AOS takes all the steps required to free that switch up from an operating network without a glitch. Similarly, if you add a new switch in or extend a cluster with additional racks, AOS takes care of the gory details of configuration changes to allow you to perform these actions seamlessly and without the classic "operator error" problems.
AOS thereby provides a Self-Operating Network™ (SON) similar to the self-driving car. And, to be honest and realistic, as with the self-driving car, it allows you to take your hands off the wheel and eyes off the road to focus on other things, but still alerts you when human intervention is required. The end result is that your network engineers can focus on helping deliver on new business initiatives, providing agility while achieving even better network reliability compared to the conventional manual, home-grown approaches.
AOS is also vendor agnostic, so you can use any switch vendor's products in your network. Moreover, you can mix and match as appropriate based on your needs and best of breed products. Besides allowing you to build better networks with this choice, you can often negotiate better prices from vendors when you have this choice, and they know it. You can thereby further lower your CAPEX.
But even so, what is so frightening about AWS? Is it really fair to characterize it as a black hole? After all, shouldn’t your company just focus on its core competence and leave IT to the "public cloud"?
Well, first of all,"public cloud" is a complete misnomer. The reality is that AWS is Jeff Bezos's cloud and he is using it to enter more and more markets. Yours can be next.
For instance, Netflix has run on AWS for years. Now, Amazon is offering free video streaming running on their cloud to their prime customers. Their video streaming has a clear competitive advantage over Netflix because it does not pay the AWS margins, and it has inside information and influence on the future directions of AWS.
Would you turn your IT infrastructure over to your most dangerous competitor? Clearly no.
Even if Amazon does not enter the market with an internal effort, they can afford to buy countless startups that are hosted in AWS in any market, and give them the same Amazon advantages that Amazon streaming enjoys.
And, when you turn over your IT to Amazon, your IT expertise goes dark. Even if you admit to this trend, your best IT people are going to leave for growth opportunities, and your fate is preordained.
Moreover, you cannot afford to sit still and just run your own IT infrastructure in the conventional way. You simply cannot be competitive with AWS without going to automation. You cannot achieve the efficiency, the agility and reliability without it. Without serious automation of your IT infrastructure, your organization will not be able to resist this gravitational pull of AWS while staying competitive. What we are talking about is not a 10 percent or 20 percent improvement with automation; it is orders of magnitude.
Also, the time to start is now — right now.
Getting to real automation is not an overnight step. It requires adapting your organization, processes and infrastructure to automation while keeping it all running smoothly. It is this urgency that means you cannot afford to wait decades for standards and open-source solutions to emerge. AOS for data center network automation is available now.
You can start now, and you must.
I gave a talk at the recent October 2016 ONUG meeting in New York entitled "Incremental Automation or Rapid Annihilation". It may sound over-dramatic, but fundamentally I believe we are at a tipping point, and survival is on the table, which is why vendors are embracing Apstra to empower CIOs with data center automation to control their infrastructure and stay competitive. And, this data center automation is part of a larger trend toward automation. In this trend, as we have seen with Uber, the winners win big, and the losers do not survive.
So, I'd say with AWS re:Invent, as the saying goes, what happens in Las Vegas should stay in Las Vegas. Nice show, but not the path to stable existence. What happens in your IT organization should stay in your data center, in your own private cloud.
Let Apstra help you achieve this next generation IT while achieving all the agility, efficiency and reliability that allows you to control your IT destiny.
To learn more, check out the following resources:
- Cowen Group podcast with David Cheriton and Mansour Karam
- Mansour Karam's blog post on the "Self-Operating Network"
- Doug Gourlay's blog post SDN: What It Should Have Been
- David Cheriton’s white paper on the Distributed Systems Challenge in Data Center Automation