Not alone there, Microsoft is especially expensive for what you actually get but the more common players such as AWS and Rackspace are highly cost ineffective in many situations. I really wish people would stop calling it 'the cloud' and call it what it is 'outsourced hardware'.
And a whole bunch of other capabilities. My company couldn't exist if we had to buy & configure our own servers, storage, switching, load balancers, databases, monitoring &c. Characterising the AWS & Azure clouds as "outsourced hardware" demonstrates a real lack of awareness of the services that are available. As a third-party offering, they are about as far from old-style mainframe bureau computing as you can get, in every dimension from the architectural to the commercial.
Cloud computing means the entire DC is programmable, and I use it as such.
Moreover the unprecedented level of automation means I can spend a lot more time on creating customer value rather than faffing around with admin. The shift I've seen in the last three decades* has been phenomenal. Teams aren't smaller but they are vastly more productive.
If you host your own servers or even PaaS wherever, and you don't have an API and automation framework then yes - you will see benefits from having those things provided to you. If you already have these things and you're not spending a lot of time to make sure they continue to exist then you have a lot more freedom than locking into the tools that your 'cloud' provider has given you.
What lock-in? I'm deploying standards-based applications to standards-compliant platforms. Cloud services are simply saving us heaps of time & money. There's no loss of freedom, far from it; the disposability of cloud infrastructure provides enormous opportunity for adaptation and change.
In 100% of my experience to date, "cloud lock-in" is a myth trotted out by server huggers and hardware salesmen. Some SaaS providers may be data prisons, sure, but that's a different conversation.
If the economics of establishing and operating off-cloud resources ever made sense for us, we'd go for it, but it looks increasingly unlikely.
One example of that I've seen have spent significant time on tools like cloudformation can only (as far as I know) be used on AWS. Another would be where when you need decent storage performance - we have several cases where we use 20-40K IOP/s quite easily and doing that kind of work on the current cloud offerings is very expensive and usually involves significantly increase complexity if say you need this in your database layer as you suddenly have the need to scale horizontally while maintaining consistency and durability which is difficult. We can provision 6TB of networked 1M 4K random read IOP/s storage in a highly redundant, load balanced form that's easy to upgrade and scale for less than $500/month and it has little to no management overhead. Now while this may not be what your average startup requires it opens up a world of opportunity to how and what you do with your data.
Edit: I should note that we do, where appropriate use 'cloud' services including Rackspace, AWS and Azure where appropriate. Azure has had significant performance issues and has a lot of provable downtime especially due to internal network routing and DNS problems that they fail to acknowledge and we've found their support to be a joke if you know what you need / are doing, even their own O365 service has weekly outages that can take several minutes to resolve. Rackspace's support has been good but they do have a lot of small outages again often network related. AWS' has been alright be very costly unless you're doing either very small deployments or at the other end of the scale massive, horizontally scalable deployments, however their storage performance is woeful. For our mission critical or high performance deployments using our internally hosted platform is significantly fast and almost always cheaper. Our uptime across the platform is fantastic and it generally 'just works' while we watch our cloud hosted services suffer from inconsistent performance and service 'blips'.
Huh? Networked I/O at 1M random iops for $500/month? I feel like at least one of these numbers is off. Can you detail your setup a bit more? How many machines or drives are you striped across and how much of your 10gige link are you assuming you can dedicate to this?
(I'm genuinely curious but I feel like there's a missing upfront cost that's not being included here)
True for cloudformation , when spinning up multicloud systems it pays to invest in a tool like terraform.Currently we are running a hybrid Google Cloud AWS deployment and it helps to keep the infrastructure consistent.
Ever used spot instances? If you use "cloudy" strategies and shop your work to the cheapest AZ's, and only run the jobs when the spot price is right, you can run some pretty nice instances on AWS for cheaper than any other provider. However, if you just want a rack of servers running 24/7, regardless of utilization, you might as well go back to the colo.
