In this week’s episode, Richard and I talk with Dino about the work Pivotal does to help companies quickly start migrating applications to Pivotal Cloud Foundry. Check it out, and subscribe if you haven’t already.
Reda Hmeid over at InfoQ does a good job at trying to define “digital”:
Being Digital is the re-imagining of business processes to be by default a fully online, fully automated process from end user interaction to back office processing, with no need for human intervention.
…down to several characteristics of such an approach like immediate process loops for end-users (“real time”), automating most everything, being online, and well designed.
Like reading about doing agile, DevOps, and “cloud native” in the real world? Help me finish up my current booklet on that topic by reviewing my almost finished draft.
I’ve been working on this since around August of this year. I’m almost done! There’s some of my content you might have seen around the web here and there, but most of it is new. I’ve tried to wrap up all the common topics I talk about with large organizations and put in as many cases and anecdotes – proof and data! – from “donkey” organizations as possible.
Help me get more eyes on this, and also read an “early edition” before you have to get your boy Johnny Leadgen on the case.
Gap’s Philip Glebow goes over their use of Pivotal Cloud Foundry, including things that worked well and need improvement. His list of the supporting tools they use – like APM and data virtualization – is handy as well.
- (~2:30) Fast deploys: “We can deploy changes faster than people can really consume them.”
- (~18:00) Developer morale: “We could really push something in five minutes… and developers love it, you click the commit button and there you go.”
- (~19:40) [Poor transcription by me] On the danger of changing too fast: …generally we want to have a little bit of control into what goes into that production environment… but we don’t want to change so rapidly so that users are confused… There’s also a little bit of cultural change that we need to go through… ((too rapid of change is jarring)) …and as we bring that capability forward, we want to be sensitive to those concerns.
- (~23:58) Overview of their pipeline and testing.
(~26:29) [Poor transcription by me] Typically we’ve organized out teams around sort of domain concepts – so we have a pricing team – then there’s several squads, then that squad is responsible for optimization – price, packing the stuff, etc. That’s how we’ve organized the teams, two pizza teams, we’ve tried to that. Also, distributed teams… sometimes that’s a little bit complicated.
By changing its development practices and investing in a private cloud platform as a service, there have been clear benefits to the business. “Historically it would take two or three days for a deployment to go to production, with lots of manual production. Now with the apps in the garages we can do it on the basis of Cloud Foundry within minutes.”
From Gary Gruver, one of the better “how to do agile and DevOps stuff in large organizations” authors:
For these organizations implementing DevOps principles (the ability to release code to the customer on a more frequent basis while maintaining or improving stability and quality) is more about creating a well-designed deployment pipeline that builds up a more stable enterprise systems on a regular basis so it is much easier to release the code on a more frequent basis. This is done by creating a deployment pipeline that integrates the code across the enterprise system on a much more frequent basis with automated testing to ensure that new functionality is not breaking existing code and the code quality is kept much closer to release quality.
From my perspective this approach to DevOps is rather different from the more unicorn type approach described in this article. It though does address the biggest opportunity for improvement that does exist in more large traditional organizations which is coordinating the work across teams. In these cases the working code in the deployment pipeline is the forcing function used to coordinate work and ensure alignment across the organization. If the code from different teams won’t work together or it won’t work in production, the organization is forced to fix those issues immediately before too much code is written that will not work together in a production environment. Addressing these issues early and often in a deployment pipeline is one of the most important things large traditional organizations can and should be doing to improve the effectiveness of their development and deployment processes.
I get all ants-in-pants about this whole bi-modal discussion because I feel like it’s a lot of energy spent talking about the wrong things.
This came up recently when I was asked about “MVP”, in a way that basically was saying “our stuff is dangerous [oil drilling], so ‘minimal’ sounds like it’d be less safe.” I tried to focus them on the “V” and figure out what “viable” was for their situation. The goal was to re-enforce that the point of all this mode 2/small batch/DevOps/PDCA/cloud native/OODA nonsense is to keep iterating to get to the right code.
Part of the continual consternation around bi-modal IT – sad/awesome mode – is misalignment around that “viability” and scoping on “enterprise” projects. This is just one seam-line around the splits of the discussion being unhelpful
The awesome mode people are like:
You should divide the work into small chunks that you release to production as soon as possible – DevOps, Agile, MVP, CI/CD – POW! You have no idea what or how you should implement these features so you need to iteratively do it cf. projectcartoon.com
And the sad mode folks are like:
Yes, but we have to implement all this stuff all at once and can’t do it in small slices. Plus, mainframes and ITIL.
Despite often coming off as a sad mode apologist, I don’t even know what the sad mode people are thinking. There’s this process hugger syndrome that, well on both sides really, creates strawpeople. The goal of both methods is putting out software that makes users more productive, including having it actually work, and not overpaying for the whole thing.
The Enemy is finding any activity that doesn’t support those goals and eliminated it as much as possible. In this, there was some good scrabbling from the happy mode people laughing at ITSM think early on, but at this point, the sad people have gotten the message, have been reminded of their original goal, and are now trying to adapt. In fact, I think there’s a lot the “sad mode” people could bring to the table.
To play some lexical hopscotch, I don’t think there is a “mode 1.” I think there’s just people doing a less than awesome job and hiding behind a process-curtain. Sure, it may to be their choice and, thus, not their fault. “Shitty jobs are being done,” if you prefer the blamelesss-veil of passive voice.
Fix your shit
When I hear objections to fixing this situation, I try to b nice and helpful. After all, I’m usually there as part of an elaborate process to get money from these folks in exchange for helping them. When they go all Eeyore on me, I have to reframe the room’s thinking a little bit without getting too tough love-y.
“When I put these lithium batteries in this gas car, it doesn’t seem to work. So electric cars are stupid, right?
You want to walk people to asking “how do we plan out the transition from The Old Way That Worked At Some Point to The New Way That Sucks Less?” They might object with a sort of “we don’t need to change” or the even more snaggly “change is too hard” counter-point.
I’m not sure there are systems that can just be frozen in place and resist the need to change. One day, in the future, any system (even the IRS’!) will likely need to change and if you don’t already have it setup to change easily (awesome mode), you’re going to be in a world of hurt.
The fact that we discuss how hard it is to apply awesome mode to legacy IT is evidence that that moment will come sooner than you think.
(Insert, you know, “where’s my mobile app, Nowakowski?” anecdote of flat-footedness here.)
ITIL end in tears(tm)
The royal books of process, ITIL, are another frequent strawperson that frothy mouthed agents of change like to light up. Few things are more frustrating than a library of books that cost £100 each. There’s a whole lot in there, and the argument that the vendors screw it all up is certainly appetizing. Something like ITIL, though, even poorly implemented falls under the “at least it’s an ethos” category.
I’m no IT Skeptic or Charles T. Betz, but I did work at BMC once: as with “bi-modal,” and I really don’t want to go re-read my ITIL books (I only have the v2 version, can someone spare a few £100’s to read v3/4?), but I’m pretty sure you could “do DevOps” in a ITIL context. You’d just have to take out the time consuming implementation of it (service desks, silo’d orgs, etc.).
Most of ITIL could probably be done with the metaphoric (or literal!) post-it notes, retrospectives, and automated audit-log stuff that you’d see in DevOps. For certain, it might be a bunch of process gold-plating, but I’m pretty sure there’s no unmovable peas under all those layers of books that would upset any slumbering DevOps princes and princesses too bad.
Indeed, my recollection of ITIL is that it merely specifies that people should talk with each other and avoid doing dumb shit, while trying to improve and make sure they know the purpose/goals of and “service” that’s deployed. They just made a lot of flow charts and check lists to go with it. (And, yeah: vendors! #AmIrightohwaitglasshouse.)
Instead of increasing the volume, help spray away the shit
That gets us back to the people. The meatware is what’s rotting. Most people know they’re sad, and in their objections to happiness, you can find the handholds to start helping:
Yes, awesome mode people, that sounds wonderful, just wonderful. But, I have 5,000 applications here at REALLYSADMODECOGLOBAL, Inc. – I have resources to fix 50 of them this year. YOUR MOVE, CREEP!
Which is to say, awesome mode is awesome: now how do we get started in applying it at large orginizations that are several fathoms under the seas of sad?
The answer can’t be “all the applications,” because then we’ll just end up with 5,000 different awesome modes (OK, maybe more like 503?) – like, do we all use Jenkins, or CircleCI, or Travis? PCF, Docker, BlueMix, OpenShift, AWS, Heroku, that thing Bob in IT wrote in his spare time, etc.
Thus far, I haven’t seen a lot of commentary on planning out and staging the application of mode 2. Gartner, of course, has advice here. But it’d be great to see more from the awesome mode folks. There’s got to be something more helpful than just “AWESOME ALL THE THINGS!”
We’ve seen a goodly spate of news in the container space recently which we cover in the episode. In the second half, we talk with Kevin Hoffman about the .NET world, Steel Toe, and his book, Beyond the Twelve-Factor App. A recent survey from the Cloud Foundry Foundation is widening the framing around container management, adding in the use of Platform-as-a-Service into the usual container orchestration mix. The survey also shows some interesting results around adoption, e.g., managing containers in production ends up being more difficult than people predict during evaluations. Also since our last episode, DockerCon brought a bevy of announcements in the container ecosystem which we cover briefly. And highly relevant to our guest, Kevin Hoffman, .NET Core 1.0 was officially released, as open source. In the second half we talk about the recent history of .NET and how it’s being used to create microservices. We also talk about the three extra “factors” Kevin’s book adds to the 12 factor app and typical experiences when migrating to 12 factor apps.
Full show notes: pivotal.io/podcast