Wednesday, 13 November 2013

ITIL vs Dev-Ops (itSMF debate)

This month I participated in our local itSMF branch seminar. As it was the last quarter of the year, we were already a bit in the Christmas mood and as it is called the silly season we tried to make the content light and entertaining. To this regard there were 4 IGNITE session. This is where each speaker has a limited number of slides (in our case 20), which auto-forward (in our case after 30 seconds) so that the presenter has not control over the presentation behind him/her. And by asking the speaker to present not only in this format, but also in a more polarising way, we kept it quite entertaining.

But in addition we also introduced two DEBATEs. I’ve used this format in my training whereby the candidates are either in favour or against a statement (like ‘outsourcing will make all our problems go away’ or ‘you should always charge for a service’). This makes them think about all the arguments in favour, but also the weak spots of these statement and phrase there position and rebuttal accordingly.

For the itSMF seminar each team (for or against) presented an IGNITE (although the more traditional 15 slides and a 20 second auto-forward), and then a rebuttal. The debate topics were:
  •    Chicken or Egg, what comes first: process or tool
  •    Traditional Service Management vs. Dev-Ops.


I was on the latter debate, obviously defending traditional service management. My role was to present our case through the IGNITE, and I’ll provide a brief synopsis of that. To this effect I’ve shared my slidedeck HERE.

I first [slide 2] narrowed the debate down to ITIL vs. Agile and decided that a battle of this magnitude could only be properly debated, explained, reflected by using that other epic battle of good vs. evil: Star Wars!

No doubt that ITIL takes the role of the Jedi here [3]. The ancient, ruling order, steeped in mystic, historical processes but in essence a democratic society for the common good.
Agile on the other hand has more similarities with the Sith [4]: powerful, challenging the ruling order, using emotional, rash decisions (over rational, risk averse ones) and attractive to young padawans/developers.

A good example was the Deathstar [5], a somewhat rushed project that contained a fatal flaw, causing it to blow up with considerable damage to the customer (the republic).
So, what did they do: they build another Deathstar, with the same flaw and put it into production before it was even completed. No surprise then, it blew up again, with even more devastating customer impact (victory to the rebellion).

In my opinion, a lot of this has to do with perception [7, yes I skipped 6 which only served to further focus on an apparent lack of planning with Agile]. You see, Dev-Ops or Agile is currently at the height of the Gartner Hype-curve. In the minds of some people it can do no wrong, and should be used for everything and anything.
ITIL on the other hand is done the bottom of the curve, in the ‘through of disillusionment’. Many organisation have tried and perhaps not achieved the results they were hoping for. Now they are throwing out the baby with the bathwater, completely rejecting any of the common sense best practice IT Service Management approach that ITIL describes.

To a large degree I think this is because of the perceived bureaucracy that ITIL introduces [8]: processes, procedures, templates, workflows, … everything needs to be documented, reviewed & approved.
Perhaps so, but ITIL never actually identifies how complex, wordy, lengthy or formal these procedures need to be. Or even whether they should be written down at all.
More important is to understand WHY something needs to be done (rather than how) [9]. PRINCE2 (from the same best practice stable as ITIL) actually identifies tailoring as a must for every methodology and that tailoring is not about whether to use it or not but about how (formal) to use it. This is driven by factors of size, risk & complexity.

Change Management is a prime example [10]. This often becomes a bureaucratic bottleneck, because people focus too much on the formalities, paperwork, CAB meetings etc.
The ITSkeptic told at the Australian itSMF LeadIT conference, how one dev-ops engineer exclaimed that he could but a new version in production, ‘before you could get an Emergency CAB together’. That last statement doesn’t leave me with a lot of confidence, in particular if it was a mission-critical service/application we are talking about here.
Instead we should focus on the real goal of Change Management, which is to minimise impact. ‘Horses-for-courses’: sometimes you can go slow and secure, other times you need to go quick, but you better wear a helmet!

And Change Management accommodate the Dev-Ops concept of ‘failing fast’ [11] through defining different categories of change, each with their own level of documentation, assessment, approval and oversight.
  1.    The Emergency (or ‘oh-shit’) Change means we don’t throw caution to the wind, but instead apply a different, faster, less rigorous process (basically we decrease time, but increase risk)
  2.    The Standard Change allows a cookie-cutter approach to frequent, repeatable change, which will significantly reduce overheads
  3.    And the use of ‘minor’ Change categories will allow a less convoluted path to be followed, whilst still maintain the appropriate governance.


The key here is no to be afraid of failure, but also not to close your eyes to the risk [12]. Mistakes are part of everyday life, and in fact mistakes are essential for growth, improvement and innovation. I call this the concept of making ‘deliberate mistakes’ [13].

So in our battle of Dev-Ops vs. Service Management, we find that ITIL actually incorporates everything that Agile is trying to achieve. But then, it does much more, allowing for both fast and slow changes; and with a lifecycle that includes Strategy & Operations (as well as Design, Transition and Improvement) [14]

Thus there can be only one outcome, and the (ITIL) Jedi must win [15].
But perhaps rather than looking at Star Wars as the fight of Luke against the Empire (and his father) of the first (now the last) three episodes, perhaps we should encompass the whole (6 episode) series as the lifetime of Anakin Skywalker (from good to bad, and back to good): In Anakin we find good and bad, jedi and sith, passion and ration, similar to how Dev-Ops and Service Management attempt to achieve the same goal (-better- value of services) through different approaches [16].
As long as the good side wins in the end!
(May the ITSM force be with you)

the ITIL Zealot
November 2013