DevOps: Delivering cloud reliability at all costs

How DevOps is helping unify development and operations teams and building a better environment for application delivery

Cloud technologies have been fundamentally transformational for financial software provider MYOB, which like most legacy software providers has undergone a massive period of change in recent years as it shifted its architecture to the cloud.

DevOps has been internally recognised as an integral part of MYOB's transformation for nearly four years now, providing invaluable support for an overhauled, Agile software development process that has become crucial to ensuring that MYOB software is developed iteratively and delivered efficiently.

Read part 1 of this series DevOps: Keeping your cloud transformation under control

“There are a thousand interpretations of what DevOps means and we've probably gone through them all,” says business division product delivery manager John Sullivan, who sees DevOps as managing the convergence of design, development, user experience (UX) and other considerations that all affect how a service is delivered to customers.

“The days are gone when you would build software and look at it from both a functional and non-functional perspective,” he says.

“The UX becomes the functional side of things in some respects, and the operational bits become the system metrics: 'We will cope with load up to this level and still respond to the customer within 300ms', for example.”

MYOB has integrated the DevOps concept into its operations so deeply that operational staff physically sit next to developers to ensure rapid shadowing of development requirements with operational capacity.

Infrastructure teams used to respond to these design and performance goals by commissioning additional physical infrastructure – a long process that limited responsiveness to rapid surges in customer demand. Worse still, this process often resulted in physical servers configured with software stacks that weren't correctly configured to support the new software.

Yet with the transition to the cloud pushing these considerations more closely together, MYOB has integrated the DevOps concept into its operations so deeply that operational staff physically sit next to developers to ensure rapid shadowing of development requirements with operational capacity.

Careful scripting allows new virtual servers to be spun up from templates within moments and decommissioned when demand subsides.

“Those teams used to be completely separate and managed separately,” Sullivan explains.

“We've been transitioning to a model where the operations people build cloud infrastructure that they can easily bring online; they sit with development teams and say, 'If that's what you need, this is the best infrastructure for you. And if you write your code and build your system in this way, it's easier for us to manage operationally, and to manage, and to get central logging.' The DevOps guys actually tell the developers how to write their code.”

By promoting teamwork between developers and operational teams, MYOB has been able to keep its cloud operations – predominantly focused on Amazon Web Services and Microsoft Azure – in step with both developer and operational requirements. The net result has been to deliver a robust business over reliable infrastructure that – as a direct result of DevOps – hasn't had even a momentary outage in more than four months.

“If you're a SaaS provider like we are the customer experiencing problems because of your infrastructure is a really bad thing,” Sullivan says. “We can now build things, test when they're wrong and make them scale better.

“Infrastructure should never fail now because cloud gives you the opportunity to scale infinitely. And even if incidents might occur internally, I never want an incident externally – and I don't care if it costs me $1000 a minute to make sure 1000 customers don't experience pain. I'm just going to pay that.”

Powering the collaboration imperative

Just as the DevOps philosophy helped unify the previously disparate development and operations teams at MYOB, it has helped coordinate development work across a number of physically separate parties as they worked together to build the new Australian Tourism Data Warehouse (ATDW).

Designed as common aggregation platform for operators of tourism operators, government organisations, public attractions and landmarks across Australia, the ATDW replaces eight different legacy systems and has been built by development house SapientNitro in partnership with hosting provider RackSpace – which, like many cloud-infrastructure providers, has increasingly built DevOps capabilities into its operational infrastructure.

RackSpace's DevOps team “worked hand in hand with the development teams at SapientNitro,” Rackspace senior director and general manager Angus Dorney explains, noting that the close partnership “helped them deliver their environments in the cloud and in a way that's very dynamic, automated, and scalable on demand.”

Regular meetings brought together project staff, developers, and tourism stakeholders from eight jurisdictions to regularly review constantly-changing prototypes that were shaped based on customer feedback.

“The whole discovery phase was about the user journey,” says SapientNitro business lead Stephen Forth. “We worked in a very collaborative hybrid sense to understand user personas and to develop custom journeys involving everyone. And we looked at what the infrastructure should look like. That's the beauty of a hybrid team that fosters natural collaboration between the involved parties.”

Ultimately, Dorney says, RackSpace's work on the ATDW project fed further expertise to a DevOps team that “continues to learn from every project. The continuous learning experience is really important, and I think it's all part of a culture where prototyping, and being able to compare new versions quickly, is also important. That's really where that dynamic DevOps infrastructure in the cloud is helpful for this type of project.”

Not every organisation makes the DevOps transition unscathed, however: as with any orgnisational change, there are risks that unforeseen problems may compromise efforts to integrate DevOps into the development culture.

This was the case at now-defunct electronics retailer Dick Smith, whose CIO Paul Keen reported in 2014 that DevOps had been “http://www.cio.com.au/article/559672/devops-an-uphill-battle-dick-smith-it-boss/" target="_blank">an uphill battle” and “a really painful journey” that had proven difficult both by the relatively immature tools available to support the change, and by the desire to drive too much change through the organisation, too quickly.

“It's a thing you get a little bit better at each time,” he said.

Read more: Local Lenovo boss sees big potential in hyperconverged

With tools maturing and definitions of DevOps coalescing, the intervening two years of DevOps utilisation have boosted the chances of most organisations being able to make the change.

DevOps has also inspired other parts of the business to cosy up to operations staff, borrowing from its match between infrastructure and capability to build new practices such as SecOps – the alignment of information-security practitioners and operations staff.

“Having a security architect space separate from the overall architecture team provides that little competitive tension and healthy debate,” ME Bank CISO Samantha MacLeod said at a recent FST Media conference on financial-services security practices.

As part of the bank's emerging SecOps philosophy, she said, the security team “is working directly with the Project Management Office and change organisation, building blueprints on different topics in the end-to-end lifecycle. We're telling them to start thinking about these issues upfront, and providing a checklist on security that they should have.”

Formalising the security team's expectations of the rest of the organisation provided opportunities for security specialists to intervene quickly when it became necessary: “If there is something amiss, they can come talk to us and we'll help them secure it,” MacLeod said.

“But if there is something really significant that we think is so disruptive that it presents a security challenge for us, we're going to embed ourselves in that program of work and have architects involved every step of the way until we figure it out.”

Join the Computerworld newsletter!

Error: Please check your email address.

Tags software developmentinfrastructureMYOBCloudDevopsinfrastructure managementcloud computingrackspace

More about AgileAmazon Web ServicesME BankMicrosoftMYOBRackspaceSapientNitro

Show Comments