The process of migrating to the cloud is certainly complex. But the work doesn’t end when the last app is moved. In fact, in many ways, the work’s just beginning. Now you’ve got to make sure you continually manage your cloud presence to optimize both its financial and operational performance. And that’s not easy: You typically will have large volumes of data always coming in, and doing so with high velocity. In a consumption-based model, if you don't control every lever, it can spin out of control very quickly.
How do you optimize?
Based on our experience in working with a wide range of clients on their cloud journeys, we’ve identified 14 top tips for getting the most from the cloud.
#1 – Stop flying blind
The most basic tip is to truly understand what's in your estate. What are you spending your money on? Who's spending it? And where is that money going? When they embrace the cloud, most companies just migrate over and then kind of forget about it. They have little idea what's happening moving forward. Of course, you can’t continually improve performance if you have no awareness of it. So how do you build awareness into your estate? That leads us to the next tip…
#2 – Democratize access to data
This seems a little counter-intuitive. Most CTO's don't want to give out all their technology spend data to their business groups. But what we’ve found is that, if you give individual application owners or business groups as much information as possible on what they have and how much it costs, they tend to feel accountable and responsible. They're motivated to optimize that spend. But it’s also a lot of data and it can overwhelm a business user. For instance, for our largest provider, we get a 200 million-row bill every month. How do you convey this wealth of information in a way that’s useful and meaningful?
#3 – Tag it
To cleanly cut this data along the lines of owners or business units, you need tags—metadata that tells you who owns something, what it's supposed to be doing, and how important it is. But you can’t get carried away with tags. We’ve seen one company use 9,000 of them, which kind of defeats the purpose. In our experience, you really should have only five to 10 tags that focus on adding value, business contacts, criticality, and environment, and then restrict within those fields to three to five entries so you can, at a high level, quickly group information. Accenture Cloud Platform, our cloud management platform, can be a big help in providing the governance framework or foundation to enforce cloud tagging schemas and make sure they're standardized values so you can't create a custom input.
#4 – Scaffolding saves lives
The key is to avoid redundant tag values. Many times, people use an account name as a tag, and that's redundant. Microsoft Azure has a pretty robust organizational framework called the scaffold, which allows you to set levels of hierarchy within your account structure that go from the master enrollment level, to the departmental level, to an account level, and then to individual subscriptions and resource groups. Then, in a sense, that scaffold becomes your primary cost allocation tag—meaning, things that are related by budget would tend to be in the same part of that scaffold tree. So, you can generate reports at the CFO level, but then only expose what individual users use at their level—and it's all done through scaffolding.
#5 – Get a smart cloud thermostat
But what if you have thousands of machines that are changing all the time, on a day-to-day basis? For instance, at a large financial services company, one person was responsible for manually spinning up and down 3,500 virtual machines every night based on demand for them. How do you even schedule and track that? How do you scale it? Once again, this is where a cloud management platform can be a big help. With such a platform, you can automate this with a set policy-based rules that look at tags and say, "Okay, this is a development machine. I want all my development machines to shut off every night at 6:00 p.m. and come back up every morning at 6:00 a.m." You've now just cut costs in half.
#6 – Create guardrails
When dealing with optimization, you should put yourself in a position to not get things wrong, or to make it really hard to get it wrong. Our cloud management platform, Accenture Cloud Platform, provides these guardrails in the form of a governance plane that can help you enforce such things as scheduling rules and automatic resource termination. So, for instance, if you have development machines that are only used during the day, the platform can make sure to automatically shut them down in the evening so they aren’t incurring costs while they’re not being used. Putting up these kinds of guardrails is really important to preventing spend from getting out of control.
#7 – Tailor to fit
One of the most common mistakes a company makes when moving to the cloud is taking up way more real estate than they need. Think of it in terms of shirt sizes: If the workload in the company’s data center required a 2XL shirt, that’s what the company ordered in the cloud—but it really only need a large. So the estate is bloated and underutilized, which is just wasting the company’s money. You should tailor your servers to fit what you actually need—benchmarking, monitoring performance, and incrementally scaling down until you find that right size. The same is true with storage volumes. You're charged based on the amount of storage you’re provisioned, not the amount you actually use. So don't think, "Oh, I'm going to provision a five-terabyte disk because I may use it someday later."
#8 – Don’t forget about the little things
With the big charges like storage and compute commanding most of the attention, it’s easy to miss the smaller charges, which actually comprise the majority of the line items on a cloud bill and add up quickly. Don't forget about these peripheral services because, as you continue to optimize your compute and storage footprints, they make up an increasing percentage of your spend. Fortunately, cloud providers have done a great job of providing more and more visibility on what these ancillary services really entail in terms of cost and usage. You should take advantage of that to truly squeeze out every penny.
#9 – Don’t be a data hoarder
Do you really need to keep every file you’ve ever created? It’s no big deal to do that on your computer if you must, but in the cloud, that costs real money and many people don't make the connection. And when thinking about storage, you need to pick the right type that balances functionality and cost. For instance, you don’t want to store archive data on an expensive SSD drive. There’s a much less expensive option. In fact, there are many flavors of cloud storage—everything from super-fast P30s, to managed discs that cost a ton of money, to cold archive storage that's very cost effective. And don’t forget: You may optimize the rest of your estate, but storage will become a greater percentage of your spend as you continue to migrate and mature within the cloud.
#10 – Kill the zombies
Zombies are fun in the movies, but not so much in the cloud. A zombie in the cloud is any resource that you’re spending money on but not deriving any business value from. Maybe it's a server that was spun up for some sort of test use case and then forgotten about and never decommissioned. A more common reason zombies get left on is no one knows what they’re for or who owns them. Whatever the reasons, zombies are running but you're not using them and they’re just costing you money—it's pure waste. If you suspect you have some zombies, shut them down for a week. If no one raises any alarm bells, those resources were probably not needed to begin with. At the very least, it’s the quickest and easiest way to identify the owners—they're usually the first ones to contact you.
#11 – Ditch your servers
About a year ago, we saw a marked increase in adoption of PaaS services in both our AWS and Azure estate. It means the number of virtual machines in use is going up but not at as quickly as things like RDS databases, red shift compute nodes, and SQL on Azure. In fact, all of these PaaS services are becoming more and more popular. And the reason is, the easiest (and least expensive) cloud server to manage is no server at all. Take, for instance, Lamda. It's a pure pay-per-use model—with no element of waste at all. We call this out to make a point: You've got to keep one eye on your current estate but the other one always on the future to know what's on the horizon—like serverless architecture—that can help you optimize your cloud presence.
#12 – Think before you commit
Reserved Instances (RI)—a reservation of resources and capacity for a fixed period, usually a year or two—are another area that can affect your ongoing cloud performance. Buying RI is a big investment that could save you a lot of money—or it could eliminate any ROI and even cost you if not done right. Companies can be attracted to RI by the potential value. But many times, companies dive in and commit to RIs with no track record of cloud usage. So how can they know what they need? They typically don’t. They expect RI will save them money but as it turns out, they didn't forecast their demand as well as they thought they may have. If they bought an RI on a 2XL machine, but their architecture only requires a medium or a large, they’re basically committing for a year for a much more expensive machine they don't need. They may have saved some money, but there's so much more they could've saved if they more accurately understood beforehand what they really needed. Now they’re stuck for the entire term of the contract—a sunk cost they’ll never recover.
The fact is, there’s a genuine sequencing for optimization. You need to create a solid roadmap for your cloud journey based on your business requirements, demand, decommission and migration schedules, and other key variables to make sure you’re realizing as much of the potential savings as possible.
#13 – Mind your metrics
Yes, it’s a cliché, but it still rings true: To be successful at anything, you have to mind your metrics. Your cloud presence is no exception. You need to establish a few key performance indicators that are most relevant to your organization. And when you do, you need to involve the people affected to discuss the things you’re going to track and why—whether it's some sort of per-CPU core cost, or per-gigabyte RAM, or more business-focused metric, whatever you define. Additionally, remember that that as you embrace the cloud and people get more comfortable with, you're going to use it more. That means your spend also will rise. But you can still be successful if spend goes up, as long as the rate isn't rising as high as it was at the outset. It's all about cost efficiency. You need a way to measure that efficiency gain, and then communicate that information throughout the organization.
#14 – Keep your finger on the pulse
While the preceding 13 tips are all useful, perhaps the most important is this: Keep your finger on the pulse of your cloud estate with Accenture Cloud Platform (ACP), our comprehensive and flexible service delivering the legacy-to-cloud control plane required by today’s global enterprise to successfully manage your journey to cloud. ACP is pre-integrated, on-demand and pay-as-you-go.
The cloud is dynamic and constantly changing, and because of this, it requires constant attention and continuous monitoring—for instance, to handle alerts that indicate spend has exceeded a certain threshold or something out of the ordinary has happened. In fact, you may consider appointing a dedicated resource—a “cloud financial admin”—to doing this full time. The fact is, the cloud never sleeps and, neither should optimization. It’s not a “one and done” exercise.
As the cloud becomes an increasingly core part of most companies’ business, optimizing your cloud estates financial and operational performance will only become more important. We hope this post has given you some food for thought on how to optimize your cloud presence. Feel free to reach out to our experts at Accenture. We can help you get the best out of your cloud.