We have all seen the headlines: company X saves $Y by building its own infrastructure (as opposed to using a cloud provider).
Dropbox revealed in its S-1 filing that moving storage and applications from AWS to its own datacenters resulted in significant cost reduction. Bloomberg just published a nice analysis of the move here.
Having used major cloud providers (Amazon Web Services, Google Cloud, Microsoft Azure, IBM/SoftLayer) and having built my own at Advection, it's clear to me that cloud services are here to stay. They will evolve, but the basic premise of being able to access capacity "on tap" will remain a strong argument.
I have also witnessed a number of major cloud deals that were won or lost, and there are patterns. So, when to use cloud services, and when to avoid them? Here's a simplified view:
DO use cloud services
- you're starting a new venture, and your requirements are uncertain
- you do not have network engineers on staff
- you are a large organization that can balance multiple cloud providers and negotiate significant cost savings
- you prefer operating expenses over CAPEX
- you want to shift the responsibility for uptime and security to a 3rd party
- your traffic patterns are unpredictable and spiky
DON'T use cloud services
- you can predict or control your traffic/storage/computing requirements
- you have network engineering and devops in-house
- you can build your own hardware
In summary, if you understand your capacity requirements, and have the expertise to provision it accordingly, your in-house solution can easily save you 60..95% per month over any cloud provider. If your company is unsure - feel free to reach out to me, and I will be happy to show you how.