Why are some customers switching away from public clouds?

Cost of Public Cloud has spiraled down to an extent that it is getting difficult these days to make a real case for a private cloud or hosted infrastructure model. Yet, every other week I get a call from someone  who after running infrastructure on Amazon  (which is by far the cheapest among all public cloud providers) for months wanting to switch back to good old capex model, or a hosted infrastructure model.
I am not sure whether this is likely to be a trend. To my best knowledge more users are abandoning their infrastructure and switching to public cloud than ever before. But the idiom --grass is greener on other side is definitely true here too.
Based on the last five conversations which I have had with organizations wanting to switch back from a public cloud service back to last decade technology, there are some indicators which are emerging. Five is a very poor sample size, so treat this not as a research perspective, but just a mere observation from an armchair expert.
Predictability kills cloud advantage
Most public cloud adopters  choose the model because they are not sure about what their growth rates or what their usage spikes would be. And this is perfect thinking. They also do not want the meddle with setting up hardware, or even entertain the idea of buying hardware. On top they do not want to hire a sys-admin to set up the infra for them. This is perfect startup mentality.

But as they grow, there is some predictability to their usage model, and then there is some amount of visibility to where growth is taking them. This is where they see that the bills from a public cloud provider being higher than an alternate set up.

Public cloud providers are way too proprietary

Whether it is IaaS(Infrastructure as-a service) players like Amazon, Google or Microsoft, or PaaS (Platform as-a-service) players such as Heroku or Engine Yard, there is certain amount of lock-in which many users are finding it a bit suffocating. They all agree that a software API driven service works like magic, but as they grow their dependencies on a public provider there are several hidden costs that seems to be hurting.

Among the five companies who wanted to move away from Amazon/Google to a capex model just one have made the switch, because others have  locked their architecture to one or other of the public cloud providers.

Rented hardware models are cost-effective

The company I run Netzary manages a few hundred Linux servers for our customers. We have several of them hosted at third party data centers which provides bare metal servers at really cheap costs. We have 8 core servers as cheap as $ 54 a month with 32 GB RAM offered to us. The cost of running a fairly optimized server after paying 24x7 monitoring and management fees to us is 1/3rd of the bill a customer is likely to run compared to the costs of running the same on Amazon.

Some of you will demand  precise numbers. That is beyond the scope of this article. However to give you a rough idea, an Amazon M1 General Purpose Extra Large would cost $ 252.0 and $ 400.0 depending on bandwidth consumption, and other services you may consume. A similarly configured  server with a 10 TB of monthly bandwidth and with 24x7 sys-admin support would cost approximately the customer $ 110 a month when they sign on with us. This includes the money you pay the Data Center for the server including all costs.

We charge approximately Rs 25,000 ($ 400) a year for 24x7 monitoring and management of any Linux server whether that is a cloud instance, a root server or VPS. This includes any number of installations, configuration and hardening requests as long as it is limited to that single box. It even includes writing up to 3 custom scripts for monitoring software applications per server. We offer the same rate to manage an Amazon EC2 instance too.

At the same time, if you were to ask me whether you should invest in a public cloud or not, in 70 percent of usage cases I would advise you to go ahead. The reasons I will discuss in a later article. I will also tell you the 30 percent of cases where you should not bet on public cloud.