Continuous Delivery Among the Donkeys

I was recently asked to give a brief overview of the mainstream market for continuous integration and continuous deployment and how companies are using it. The audience was a mixture of HeavyBit portfolio companies and their customers. So, it provided a nice mix of sellers and buyers. Since this topic comes up all the time in Pivotal conversations, I thought it’d be worth going over here as well.

Carrots on Heads, Please

I’m often fascinated—if agog in wonderment—at what the “elite” of the tech world are doing, those “unicorns.” There’s even “horses” out there—high performing organizations that are keeping up with those unicorns. But, I’m most interested in the normal folks, “donkeys,” as I like to think of them. I like seeing how the donkeys are doing when they strap a carrot to their head and try to keep up with the unicorns.

If you’re the type of person who wants the benefits of Cloud Foundry (making sure you’re shaving the right yak, focusing on application development instead of infrastructure development), you’ll want the benefits of continuous delivery, namely, speeding up your software delivery pipeline by automating as much as possible from builds, to tests, and even promoting builds to production.

These are outcomes of doing CD. The overarching goal is to reduce the cycle time to get new features into production, helping you establish a feedback loop that you then use to guide the perfection of your software. That last part is key, and often “money left on the table” when organizations stop short of changing their process.

So, let’s take a look at a quick survey of studies on how CD is doing out there in donkey-land.

Yeah. Everyone Wants a Pony. Astute Question, Boss

When I was at 451 Research, we did two studies of the DevOps market, from a donkey perspective. The first study had 201 participants, the second 300. We achieved a good cross-industry mix, not just people in the technology world. There were plenty of horses and donkeys in there. One of the first questions we asked was around each company’s desires to speed up software deployment. That is a core benefit of DevOps (perhaps the core benefit), and certainly a core benefit of continuous delivery.

In aggregate, the answer was painfully obvious—of course people wanted to go faster. Everyone wants a pony! Sliced up by industry, things got a little more interesting:

There are obvious ones who want to speed up like retail, entertainment, and SaaS companies. Industries like transportation can seem weird until you realize that IT-driven companies like Fedex, UPS, or even Uber are in that category (which is to say, there’s tremendous competitive pressure for IT to move faster). On the low-end, it’s sort of depressing that health care is less interested (I know I’d certainly like more, interesting uses of software when I visit the doctor), and perhaps things like construction are not so surprising.

When I look at this chart it helps me triangulate what types of industries are most interested in using software to change how they do their business. Those industries on the left side are certainly high on the list of people who’d be interested in continuous delivery, one would assume.

The Job To Be Done: A Platform for Speedy Software Delivery

With the desire to speed up software delivery frequency, we’ve found the job to be done. Now, a job to be done is a fun way of describing the problem a product or service solves—what “job” does a customer “hire” a business to do? I hire a hamburger to (a.) fill me up, and, (b.) make me happy. In contrast, when I’m on a long road-trip, I might just “hire” a gas station hot dog drowning in pump-chilli to fill me up, but not really make me happy.

In the realm of continuous integration and continuous delivery, there’s a clear job to be done—creating the “pipeline” for packaging up, verifying, and then deploying software into production. That is, everything between writing code and operating it. Someone once suggested to me that DevOps was simply the evolution of continuous delivery, which, while not entirely accurate, is useful framing. In that sense, I think it’s good to use one of the older but still useful process studies from the DevOps world, the software delivery platform:

Originally published in 2012, it’s stood the test of time and even popped up in one of the more useful cloud books of last year, The Practice of Cloud System Administration. CI/CD plays a huge, if not necessary, role in this process. One key point is to think of structured platform—moving code through this “pipeline” requires a lot of standardization and discipline. The alternative of re-inventing the packaging and infrastructure layers each time is the wrong kind of chaos.

There’s another vital part missing from the diagram, though, usually not from the conversation around it—a feedback loop. In addition to just getting software out the door more frequently, the primary business benefit of continuous delivery and DevOps (you see how those two so neatly dance around with each other?) is access to oodles of feedback about how customers are using your software. This is used to hone and modify your software to better fit what your customers want, and one presumes <hopes>, that it leads to making more money from them.

Adapting to this feedback loop is the “money on the table” when it comes to process. I see very few donkeys taking advantage of those feedback loops, and the horses and unicorns even struggle with changing their process. Surveys and anecdotes alike show that people feel like things are often going wrong with their modern IT projects and often blame doing too little when it comes to change.

Speaking of Failure: CI/CD Use is Low in Donkey-land

Getting down to actual tool-usage, surveys from 451 and others, show a consistently lower use of CI/CD tools than you’d expect:

Here, you see the sad donut and some confirming, but encouraging survey data from DZone.

The sad donut shows that somewhere between 25–30% of the respondents are using CD tools like Jenkins, Bamboo, or hosted CI/CD services. There is a seemingly positive 35%, let’s call it, who are creating their own CD platforms. In past years, “rolling your own” when it came to service delivery platforms was needed and there was no other option. But, the technology is mature at this point, calling into question the strategic worth of DIY’ing your continious delivery platform. Effort spent creating your own CD tool is probably better spent on, you know, the actual software your customers use. The most distressing part of the donut is the near 30% of respondents who are doing nothing. I call this “distressing,” because, one, I like to be hyperbolic to keep myself awake, and, two, because CD as an idea and technology has been around a long time, well over five years. It is hard for me to imagine a software development team that wouldn’t benefit from it.

Now that I don’t work at an analyst shop, I can freely mix and match research data. So, I wanted to pull in some additional survey results on CD. On the right, you can see a year over year comparison of a DZone study, first listing companies who believed they did CD and then, according to a pretty good criteria based on orginization indicators laid out by Martin Fowler, judging which of those respondents are actually doing CD. Again, the results are somewhat shocking, but help triangulate the sad donut.

The Pipeline Is Your Platform, Perfect It!

For ease-of-thinking and strategic application, I wanted to simplify the software delivery platform diagram a bit. So, here it is with Pivotal colors!

There are a few things to think about:

1. The most valuable part of this process is actually a handful of pixels—the first few steps where you’re creating the software used to help run your business. The feedback loop is clearly key to getting the right specifications in place and making sure you’re writing software that is actually effective. I believe these first two boxes are where most companies—most donkeys to be sure—should spend the majority of their effort (and yes, one should bundle test/verify in there, that’s always nice…).
2. Over recent years, we’ve spent most of our time as an industry focused on the rest of the boxes, especially the infrastructure platform and production concerns boxes. Topics like cloud, OpenStack, and containers have made this area a churning pool of excitement. When I was an analyst and doing cloud strategy, I came across enterprises all the time who wanted to build their own platforms out of piece parts. For some—often the unicorns—it makes sense, or at least it used to. For others, it’s probably gold-plating and a mis-application of time and money. Unlike in recent years, there’s are so many “off the web” (to morph the old OTC idea) infrastructure parts that you’re likely better just using one of those rather than letting your people go crazy with the always entertaining task of building a platform.
3. You also need to be on the look-out for fat boy scouts in this pipeline. If you lead an exciting enough life that you’ve read The Goal (a “business novel”), you’ll hopefully recall the lesson of the fat boy scout—your line of marching boy scouts will only be as fast as the slowest marcher. It follows that optimizing anything else before addressing bottlenecks is a waste of time. Focus on analyzing the whole process (i.e., the pipeline) and ruthlessly remove the bottlenecks. If you want an IT nerd version of The Goal, check out the more recent The Phoenix Project.

All of this raises a larger point. While we might think of all of this as Agile and the sort of cowboy codery that’s often associated with it…there’s actually a tremendous amount of discipline, process, and careful work involved in running a business with a continuous delivery engine. It’s much more than just using a tool. It relies on building out and perfecting the entire pipeline. The rewards are high, though—getting software out the door faster, making customers happier, which hopefully leads to more profits.

Let’s Fix the Sad Donut

From the anemic, but slowly beefing up, usage of CD we’re seeing out there it’s clear that we can do better as an industry. Almost five years after the publication of the Continuous Delivery book, there’s a lot of that half full glass to fill. In the meantime, the growing interest in developing mobile applications, cloud native applications,12 factor style applications, DevOps, and cloud in general has sparked a renewed interest in the otherwise sleepy,but always valuable, corner of the IT world, software development. Software development is suddenly very interesting and the focus of lots of great work and innovation from companies large and small. Indeed, HeavyBit’s investment thesis is that development tools as a market is fast becoming a big deal again; a strategic point that Pivotal obviously believes as well. Part of the goal of Pivotal Cloud Foundry is to provide as much as possible in the nature of structured platforms to slim up all those fat boy scouts in the pipeline. While the label “Platform-as-a-Service” has been and is certainly one that applies, we prefer to think of what we offer as a set of tools that helps our customer perfect the software delivery pipeline, that is: a platform.

Recommended Reading:

• Elisabeth Hendrickson on Using PaaS for Continuous Delivery at the Cloud Foundry Summit
• How to Continuously Deploy Code with CloudBees’ Jenkins for Pivotal Cloud Foundry
• Check out the Pivotal Cloud Foundry blog
• Check out the recording of the actual talk, which includes a bonus section on how containers have zoomed in to effect all this.

About the Author

Biography