The DXC Blogs – DevOps in GIS

23May17

Originally posted internally 12 Feb 2016, this was the first of what became a series of posts where I took an email reply to a broader audience. Global Infrastructure Services (GIS) was the half of CSC that I worked in before the creation of the Global Delivery Organisation (GDO) in DXC Technology:

JP Morgenthal sent out a note asking a bunch of people for their views on DevOps at DXC Technology. Here’s my reply to him:


In my IPexpo presentation some 16 months ago ‘What is DevOps, and why should infrastructure operations care?‘ I started out by saying ‘DevOps is an artefact of design for operations’, and highlighting the need for culture change (and only now do I see the typo in that deck).

With that in mind I’d say that GIS is at the start of the journey. We have a plan to redesign the organisation for operations, and we have the intent to change our culture (the way we do things around here), but activity on the ground is only just moving from planning to execution.

In my first sit down with Steve Hilton his list of priorities were: automation, automation, automation. Breaking this down a little:

1/ Operational Data Mining (ODM), which is the process by which we’re taking the data exhaust from the IT Service Management (and ancillary) systems that we look after and applying big data analysis tools to get insight into experiments that we should run to improve our operations across people, process and tools. It’s early days, but I’d say that so far we’ve been most successful in the people and process areas because we’ve found it hard to get out of our own way when it comes to tool deployment. This is a salutary reminder that DevOps is not (just) about a bunch of tools, because we’ve been able to have real impact without any tool changes (e.g. changing shift patterns to avoid a start of day overload has greatly improved SLA conformance without requiring any additional staff [or any other changes]).

2/ Becoming more responsive – this is the part where the Automation 1.5 programme kicks in, and the tools start to be meaningful. Some of it’s about streamlining how we do ITSM, and I’d note that there’s still a bunch of ITIL happening there. The rest is about ‘every good systems administrator should replace themselves with a script’, and providing the framework to create, curate, share and deploy those scripts. The SLAM.IO team in KH’s group have made great progress on wrapping triage scripts (that use Ansible) so that operators don’t have to yak shave their way through the same command line interactions every time they look at a box. This is a precursor to API driven integration, where the triage gets done before a human even sees the ticket.

3/ Becoming more proactive – once we step beyond incidents and problems we’re looking at deliberate change, and making that a push button repeatable operation rather than manually ploughing through runbooks. Agility takes centre stage here from a tooling perspective, but we also have Hanlon to help us deal with bare metal (before it becomes a cloud that Agility can work with) along with the output from DB’s Automation 2.0 team (which is in the process of being partly open sourced).

Upskilling our organisation will be a key part of the culture change, which is why I’ve been writing about the ‘Four Pillars of Modern Infrastructure‘ and encouraging people to learn Git/GitHub/Ansible/Docker and AWS. I’m now working with CK and GS on a DevOps bootcamp workshop that we’ll start delivering to the PODs next month. This follows on from work by GR, DE and others before I arrived.

We also need to be more data driven. ODM is a part of that, but more broadly I’m challenging people to ask ‘What would Google do?‘.

It’s also important to recognise that the shift to infrastructure as code doesn’t end with the code. Our code must be supported by great documentation, samples and examples to be truly effective, which is why I’ve been very clear about what I value.

This is of course a very high level view. You’re going to find some great examples of agile development and CI in the development centric parts of GIS, and I’ll leave it to the individual leaders in those areas to explain in more detail.

Retrospective

This was the first time I used ‘Design for Operations’ in the context of a relatively wide internal audience, and I’m super pleased that it’s since become a common part of how we talk internally and externally about what we do and how we’re changing. The post came shortly before the publication of The DevOps Handbook, so although I was familiar with the ‘3 DevOps ways’ of ‘Flow, Feedback and Continuous Learning by Experimentation’ from The Phoenix Project I didn’t call them out specifically. As I was later setting up the Operations Engineering (OE) group I took an approach of ‘All in on ODM’, making continuous learning by experimentation applied to operation constraints the tip of our change spear; and I generally think that improvements to flow and feedback come naturally from that. Automation 1.5 was a Death Star that we stopped building, and Automation 2.0 became part of the overall OE direction. 15 months later, and we now have a delivery organisation that’s no longer divided into Dev and Ops, so with respect to Conway’s Law things are now pointed in the right direction. I’ve also found some awesome (DevOps) engineers in our Newcastle Digital Transformation Centre who have a can do attitude, and the skills to make stuff happen.

Original Comments:

MN

This is exactly the type of aspirational view we need for the digital enterprise.

GIS solutions have to provide the foundation for these capabilities in order to elevate the value chain.

Many parts of the traditional role of GIS (equipment integration) are approaching utility if not already there.  If you look at AWS alone, they are attacking some very mature and integrated business models with their offerings, following plans that look very much like this evolution.

Very well put Chris.

 

MH

I blogged about Automation a while back Manual tasks of today should be the Automated tasks of tomorrow (link lost to the demise of C3). A couple of interesting takes on it.



2 Responses to “The DXC Blogs – DevOps in GIS”

  1. 1 Tim Coote

    Chris, isn’t the scope of DevOps and Design for Operations inclusive of the applications themselves? Or GIS still bound by Conway’s Law in this respect?

    • So GIS was bound by Conway’s law, which led to me saying ‘how do I do DevOps when all I have is Ops?’. Thankfully that’s a problem that’s now gone away with the new Delivery organisation we have at DXC, which spans everything.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s


%d bloggers like this: