There are few subject in enterprise IT that are quite as mysterious – and cause the degree of consternation – as the cost of cloud computing. Cloud costs from major providers like AWS, Azure and Google Cloud aren’t fixed, and can vary wildly (and unpredictably) based on usage and the number of cloud-based tools used. Yet companies have little choice but to wade into this budget quagmire: certainly the cloud is now foundational in enterprise IT. Leveraging the cloud is a non-negotiable part of keeping up with the competition.
In this webcast, we discussed:
- Are the primary benefits of the cloud cost savings? (Short answer: No.)
- Are there any unique cost advantages or trade-offs to multi-cloud, hybrid cloud, edge, and/or microservices?
- What are the cost tradeoffs between using cloud computing vs your own infrastructure? What about “cloud repatriation?”
- What are some concrete steps companies can take to manage their cloud costs?
- What about the “cloud enables digital” strategy?
To provide insight into controlling cloud costs, I spoke with these leading experts:
Bernard Golden, Executive Technical Advisor, VMware
Joe Weinman, consultant, and author of Cloudonomics
James Maguire, Managing Editor, Datamation – moderator
Download the podcast:
Watch the webcast:
Concrete Steps to Control Cloud Costs
Golden: Well, the old cliche is, you can only manage what you measure, and so the first thing is measure what your costs are. And there’s a variety of tools or services that you can use that will basically examine your bills, your accounts, and go, you’re spending this much money on X, you’re spending this much money on Y, you’re spending this much on Z.
And then you can start to think about, okay, Where are the areas I’m spending money on? And what can I do to manage those efficiently? ‘Cause the thing is, you don’t necessarily to spend the least. You want to spend the least to get the most value.
So it’s not necessarily how little can I spend? It’s how little can I spend to get what I want from this cloud thing, which has all these capabilities and characteristics like elasticity and so forth.
A first step is always just look for the stupid stuff. Like you’ve created a lot of volumes, you don’t have them attached to machines, they’re not serving data, but you’re still paying for them. Well, shut them off. Another place that people commonly look at is: you have lots of instances, virtual machines, running that are all running at 2% utilization. Well, that’s a clear case that you’ve bought too big a basket for the amount of goods you’re putting into it. You should be shrinking the size of the instance.
And then you can ultimately go to more sophisticated things, which is: how do I design my application so that it responds well to transient changes in traffic where I need a lot of capacity? But when those transient pieces all fall down and the load goes down, I can shed some of those computing resources and still have the application up and running, but just with less resource devoted to it?
And so those are the kinds of things that need to you look at. So the first thing is measure it. The second thing is to look for the obvious things that you can take that are sort of like don’t require a lot of insight. Then the third is start to examine your bill and your application architecture and so forth to see if you can design for a more efficient use of the resources.
The hybrid multi-cloud fog
Weinman: I think the war is over at heterogeneous architectures, and it’s what I call the “hybrid multi-cloud fog,” which is hybrids of private and public, either for workflow enablement at the SaaS layer or for infrastructure arbitrage and hybrids of cloud and edge. And the fog in between based on where things optimally belong. And so there’s no simple one-size-fits-all solution, I don’t believe, for many clients.
Anything and everything is possible [in cloud]. Those that stay pure cloud native, those that stay in data centers. Zynga’s an example of a company that started in the cloud, then went to its own data centers because they determined that the mix of performance games as well as unit cost differentials made that make sense.
And then, as they started to, unfortunately, decline, they said, “Uh-oh, now we have all this unutilized owned capacity, we should go back to the cloud.” So all options are possible.