Does transitioning to cloud actually save you any money in the long run?
Sort by:
That makes a lot of sense <mention id="5d1c5a21d36e1d17d7d04ff9" displayname="Anthony McMahon"></mention>. Have you navigated a cloud migration at an organization before? If so, how did you shape your cloud migration strategy?
I still think, outside of the McKinseys and analysts in the world, no one's done a true total cost of ownership ROI on cloud versus on-prem and the velocity you get from cloud versus the non-velocity you get from on-prem. It's probably very sector-dependent.
I can tell you, for Facebook four years ago, anytime the fleet was over 100,000 machines it was economical to build it yourself and, for us, building it yourself was everything; from the data center all the way on up to the machines, the power distribution units, everything. But that's probably changed a little bit.
I can tell you for Yahoo, it wasn't. Server utilization was abysmal, so it depends on the rest of the organization.
The infrastructure starts to get complicated. Going to the cloud is not cheaper, it's absolutely not cheaper. If it was cheaper, then what you would see is cloud companies don't make very much money, except all of them make tons of money. So, where does that money come from? And it's opaque and they're now all moving, even the SaaS providers, to consumption-based pricing, and this is really, really naughty. The silver lining out of all this is it's going to be job opportunities for CIOs in that kind of role because it's not the marketing people that are going to figure this out.
We can spin up a virtual machine that is pretty close to free since we freed up quite a bit of capacity on-premise as we move more services to the cloud. Even though the majority of our production infrastructure's in the cloud, a lot of our corporate infrastructure is on premise because because the ROI is actually much better to keep them running on older hardware that is paid for. This might change when it is time to refresh the hardware, but we are setting servers lasting 5 to 6 years where in the past I would plan to replace every 3 years. We look at every workload and determine the location that makes the most sense.
Cloud migration is not cost savings; if you can get yourself out of that mindset, that's going to help. It's about what makes the most sense for you. Just buying all of that equipment, racking and stacking, and then supporting it and having people having to go and care and feed it, in some instances, still makes sense and in many it doesn't. It's a toe in the water kind of approach. There's no blueprint, but I think you have to have a strategy. As you start to retire systems then you can evaluate, "Okay. Does it make sense to have this here or does it make sense to have somewhere else?"
It will transform your costs, but it won't necessarily save them.
And if you don't actually understand what your costs are before you start then it is more likely to increase them.
Especially on the services you now need to consume/uplift because you went to cloud.