This article caught my attention today
A story around how, of all organisations, LinkedIn marks the failure of 4 year programme (called Blueshift by LinkedIn) to migrate to Azure – going ‘all in’.
The reality is that this is a story that is being repeated across the Industry at the moment, what was seen as a route to speed and agility has quickly ended up
But why are people bringing stuff back – from my perspective I see three major drivers…
1) Cost
Very often one of the key drivers for Cloud Repatriation is cost, and if there’s something that you’ll quickly learn when you migrate to Cloud it’s that if you didn’t know what things were really costing you to run before, you will when you get a monthly bill for it. The ability of Cloud to provide you with on-demand scaling suits those first entrants well and is the reason that those first steps seem quick, cheap and easy – but as you approach scale, you reach a tipping point where the economic benefits fade away and eventually end up on the wrong side of that balance.
2) Governance and Regulations
Whilst it is very true that you can build yourself a great secure architecture in Cloud, and often you can ensure that you can pass whatever audit and regulatory requirements are in place for your industry in Cloud, very often the perception is that keeping to those strict regulations is just ‘easier’ if the data and systems are under your control. The risks of non-compliance if you’re in Financial Services or Healthcare are huge, and unless your Cloud Provider is going to take on the flow-down of risk – often it makes more sense to hold the keys to all of that yourself.
3) Data Gravity
In case you’ve not noticed, we’re generating data at an increasing rate every day – and storing that data is another one of those economies of scale that traditional clouds struggle with, in fact not just the storing the data but moving the data. around. Ingress and Egress charges for data into and out of clouds (in fact sometimes even between AZ’s of Cloud Providors) really hits you in the pocket. If you have a lot of data and want to be able to access that data without high charges – having that data in your own data center can be hugel beneficial. At the risk of jumping on the AI/ML bandwagon here as well, as soon as you want to point those tools at your data …. having it in one place makes a huge difference.
Conclusion – where does this leave us
This may come as a shock, but it turns out what really matters is understanding what you should run where and why – rather than trying to force-fit workloads into places they’re just not meant to fit. In just the same way that the rise of takeaway pizza didn’t kill off every Italian Pizza Restaurant, one way of consuming IT isn’t going to kill off every other one. More importantly, just as you woudn’t ever say ‘I will never go to a restaurant again, from now I’m all in on takeaway’ the same is true for Cloud.
Major migration programs like the one LinkedIn embarked on didn’t fail because the technology, people or process wasn’t there – they failed because they were a flawed concept from the start, and were always set up for failure.