Although public cloud is popular and looks like it's here to stay, it isn't the best option for every organization....
An on-premises private cloud is, in certain cases, more beneficial than a public cloud. If you're trying to figure out which type of cloud is right for you, this article is a good place to start: Here, we'll take a look at which applications public cloud is best suited for and how to provide a competitive on-premises private cloud offering.
Jack of all trades, master of none
When they look at the bottom-line costs on Amazon Web Services (AWS), most business-oriented people notice the low cost of compute without putting much thought into why it's cheap. While Amazon deploys VMs with ease, it doesn't include high-availability functionality most data centers require -- at least, not for free.
Compare on-premises private cloud to public cloud
Be sure to compare resources on an equal playing field when you develop a budget. Consider how you will back up ultra-cheap Amazon instances and whether your disaster recovery (DR) strategy for cloud instances is equivalent to the protection you maintain for on-premises servers. Think about how you intend to back up site and cloud servers. What is the recovery point objective? The recovery time objective? These considerations aren't exciting, but they are important. If your livelihood depends on a machine, ensure it has a working DR plan, regardless of where it's located.
The cloud isn't always cheaper because you need to factor in storage and compute costs. Again, never forget to compare resources. I often see potential cloud customers run the numbers for storage and find that cloud technology is significantly less expensive than AWS; this is especially true when high-performance storage factors into the equation. Of course, that means nothing if the storage resources don't meet your organization's needs. What works for one organization might not work for another.
Automate provisioning and reporting functionalities
One of the selling points for the cloud is that it spins up VMs in just a few clicks; however, this doesn't consider account management. Tier one vendors usually have automation platforms that provide self-service workflows and approval mechanisms. A word to the wise: Learn these tools and use them to their full capability.
Not only will this help your company provision VMs faster, but it will also bill the correct department for the cost of these VMs -- after all, unused VMs still cost money. You wouldn't want someone to subvert IT and pull out the credit card, as there is no protection against them misunderstanding costs. One way to save money is to show the user how much it costs to provision workloads so that they can see exact costs for subsequent months.
Reporting is critical as well. Anyone who has ever worked for a large company knows that the administration wants to see management and usage reports. These reports need to be accurate and ready to go from day one, so build cost estimation models and prove they work.
Use hypervisor features
One way to reclaim unused RAM and storage (thin provisioning) is to use hypervisor features. Optimize VM builds, ensure that all unused services are powered off and optimize the build to the environment. Every saved cycle of compute is a cycle you can use elsewhere.
Standardization is critical
When building the service offering, standardize… and then standardize some more. Build a service catalog and maintain it fastidiously; this keeps requesters honest and constrains them to what you want to provide. All services and systems should come from the service catalog. Give users the option to change compute and disk -- within reason -- but bill them for it. When users add nonstandard complexity, costs skyrocket, even though the cost to the end user remains the same. If an option isn't available in the catalog, then the option needs to be reviewed.
When is public cloud the answer?
The cloud is excellent at horizontal scaling. If your customers' scaling needs are seasonal, consider using a combination of private and public clouds to scale into the public cloud when a spike occurs. As the old adage goes, "Own the base, rent the spike."
Although the public cloud is an excellent option for some cases, it isn't the best option for every case -- sometimes an on-premises private cloud is better. Evaluate your decision on a case-by-case basis. Why are we putting this application in the cloud? Does it fit the cloud profile? What do we gain from putting it in the cloud? Sometimes the honest answer is somewhere in the middle, but it's good to show the application owner the pros and cons, the cost and the options.
Find a balance between on premises and the cloud
Store massive amounts of data with an on-premises cloud
Cloud computing experiences the boomerang effect