The humble domestic goldfish is hardly an obvious icon of enterprise wide cloud transformation in financial services, but at wealth manager AMP it has become the stealth codeword for making hybrid infrastructure and SaaS pay off at scale.
The banking and investment stalwart this month lifted the covers on the progress of its six-year cloud platform journey, candidly revealing not just the breakthroughs of application migrations, but the character building pitfalls it confronted along the way.
It’s hardly a secret that AMP has been a relatively early local adopter of AWS’ public cloud in anger, as it moved from traditional on-premises systems in a bid to become customer centric with its services and experiences.
Much less known is how it developed a rapid and automated approach to tackle hybrid systems where the cloud vision frequently where falls from a dream into brutal reality.
“Today we have deployed more than 70 percent of our applications within our private and public clouds,” AMP’s head of technology for platform services, Kathryn Small, said in a presentation to the Dynatrace Perform user conference in Sydney.
“This has led to a reduction of infrastructure costs of over 30 percent.”
According to Small, a key element in that move has been determining what can easily be targeted for lift-and-shift cloud migration … and what’s less consistent and refuses to come along for the ride as smoothly.
Which is where AMP’s great love of goldfish comes in.
Having clearly endured countless presentations featuring the now endemic ‘pets vs cows’ analogy for characterising systems - where pets should be feared because they’re individualistic and difficult to upgrade, but cows are revered for the scale and interoperability of their herd members - there was clearly a necessity for a hybrid critter of compromise at AMP.
“Now what is unique to AMP is the concept of goldfish,” Small explained.
“It was through our Cloud 2 journey that we realised that not everything could be easily replaced. At the end of the day, state needs to be persistent somewhere.”
To put that in context, it’s meant that AMP’s key systems and applications - and their users - which couldn’t easily be lifted-and-shifted to public cloud don’t just get their plugs pulled in an all-or-nothing upgrade.
But they don’t languish as unloved legacy orphans either.
Small said this had meant deploying a “cloud 1.x framework which we think of as a continuous spectrum” which bridged private and public cloud spheres.
That still makes the goldfish a small pet; but an easily replaceable one that can work in multiples if necessary.
The key factor that prompted AMP to go down this path was “the establishment of a cloud centre of excellence”, Small said.
“This team’s sole purpose was to help applications migrate to the cloud leveraging native capabilities such as infrastructure as code, auto scaling and auto healing...This meant full stack automation, not just infrastructure; we had to automate our application and our code deployments.”
Small insisted that embedding flexibility was also key, especially when it came to proving worth and moving forward.
“How many times have you put together a business case only to realise that stranded costs meant that it didn’t stack up,” she said. “This is a stifler for innovation.”
In the meantime, as new projects are dialed up, what’s made the difference is the application of automation, artificial intelligence and root cause analysis to application performance management in the cloud.
“My personal favourite is the ability for self healing. I personally am very thankful that my teams are no longer called to join service restoration calls in the middle of the night when something goes wrong with one of our applications,” Small said.
Small said monitoring needed to be baked in as a priority to keep things running.
“I can’t stress enough that this simply should not be an afterthought. You’ll see from our own experiences why this should be top of mind for any project, not just moving to the cloud.”
The lessons from not sufficiently knowing what was happening on an enterprise platform were not always easy.
Having moved to a customer centric model, deployed mobile apps for customers and created a far more usable and richer user experience AMP became a brief victim of its own transformative success.
“From 2015 we saw unprecedented growth … more than 75 percent in the year that followed,” Small said.
But she cautioned although “this is a great problem to have for any organisation” things only went well if surges could be anticipated.
“Unfortunately for us, it wasn’t all smooth sailing and we experienced what I like to refer to as growing pains,” Small said.
“This was as a direct result of volume increasing which led to stability and performance issues for our customers. Our highly distributed architectures - which meant that experiences spanned many teams - led to prolonged outages.”
Not deeply embedding “resilience principles” often led to “cascading failures” and made trouble-shooting more difficult. Then there was the issue of keeping an eye on everything at once after changes.
“Visibility and monitoring of our critical customer journeys was lacking and in some instances this led to customers finding issues before we knew about them,” Small said.
“Worse still our chief customer officer was frequently finding issues first and telling us our mobile applications were not working. For a company that was trying to become customer centric this simply wasn’t good enough.”
The fix for that problem was the swift initiation of a resilience program.
“ [Its] sole purpose was to address these growing pains and help improve availability and performance for our customers,” Small said, adding this prompted an upgrade for AMP’s Dynatrace tools.
“In a single day we were able to implement synthetic monitoring,” Small said. “We had visibility, if something was about to break, we could respond.”