Early reports on cloud computing indicate that more than 70% of organizations are considering moving some data and workloads back on premises. Why were those workloads put on the cloud in the first place?
In some cases, the answer is simple timing. Data management needs and requirements change over time, and while cloud computing might have been ideal for those workloads a few years ago, it might not be now. The one constant in the data management industry is change. But another problem is hype. Companies have been told all kinds of things about cloud computing. Some of those things are true, some aren’t. Some are true in certain circumstances and false in others.
Cloud computing is one of the greatest IT revolutions of our time, but many of the things that “everyone knows” about cloud are actually myths. Let’s explore the most common concepts surrounding cloud computing to determine which have been proven out by time, and which myths have been busted.
1: Public Cloud Is Not Secure
A common misconception about the public cloud is that putting data on the cloud is a big security risk. Technology giants like the hyperscalers, Amazon, Microsoft, and Google (News - Alert), have far better security on their public clouds than most enterprises. When properly configured, public clouds are exceptionally secure.
This myth is BUSTED.
One caution to this is that while the public cloud definitely offers more security features and has dedicated security personnel monitoring the infrastructure 24/7, the number one cause of security breaches on the cloud is misconfiguration on the customer side. This can happen, and has happened, on even the most secure public clouds.
Another thing to keep in mind is government regulations that may render the security level on the public cloud moot. Some highly regulated industries handling sensitive data are required by law to keep data contained within a specific geographic location, or simply on premises. This means that, regardless of how secure a public cloud may be, check regulations in your area before putting sensitive data there.
2: Cloud Will Relieve Us From Maintenance And Upkeep Duties
Many companies moved their data and workloads to the cloud to save overhead in maintenance time and labor costs. The assumption was that they no longer needed all those smart, skilled, aka, expensive IT people. Cloud software and infrastructure such as databases promised security patches, hardware upgrades, etc. would be the responsibility of the cloud or software provider, not the customer. This concept is only partially true.
TRUE: Many aspects of administration and maintenance are handled for you on cloud systems. This can greatly reduce downtime and maintenance costs.
FALSE: The cloud does not manage itself. Cloud providers operate on a shared responsibility model, which means that cloud customers are equally responsible for infrastructure configurations, network controls, access rights, and data security. Organizations still need skilled infrastructure managers, cloud database administrators and other resources to manage, maintain, and secure the cloud.
3: Cloud Will Improve Analytics Performance
Cloud has several advantages, but sadly, performance isn't one of them. Cloud virtual machines, Kubernetes, or any other style of containerized or virtual infrastructure takes about a 10% performance hit compared to a bare metal implementation. What’s more, cloud performance can also be affected by “noisy neighbors.” If multiple people are using the infrastructure or application at the same time in a multi-tenant environment, the network can become clogged with high traffic. Performance can bog down even when your own company is barely using the system.
This myth is BUSTED.
4: The Cloud is Ideal for Spiky Workloads
If you have workloads that are sometimes low, and sometimes orders of magnitude higher, maintaining peak workload infrastructure when it sits idle most of the time is an expensive energy sucking black hole. The cloud delivers elasticity and on-demand performance. If your workloads tend to be steady, then on-prem might make more sense, but if you need to scale up suddenly, cloud computing is ideal since there is virtually no limit on supply of compute power.
This is absolutely CONFIRMED.
Auto-scaling, the automatic increase in compute power as workload demands increase, is one of the greatest conveniences and benefits of the cloud. The caution here is that as the infrastructure automatically scales up, so does your bill. Auto scaling compute nodes without having guardrails in place can auto-inflate costs and auto-destroy budgets. Set a cap on how much infrastructure you can afford to pay for, and under what conditions auto-scaling makes sense for your business. Don’t trust SaaS (News - Alert) applications (that make bank the more infrastructure you use) to put any kind of sensible limits on autoscaling for you.
5: The Cloud Will Save You Money
One of the biggest misconceptions about the cloud is that it will enable cost savings for organizations in comparison to buying, operating, and maintaining on-premises infrastructure or using a co-location provider. Cloud providers can demonstrate short-term savings due to low up-front capital expenditure, and all the hype that keeps repeating – Pay only for what you use! It sounds like a good idea, but what it really means is pay over and over every time you use compute, rather than paying for it once and using it indefinitely. The CTO of Catch Media was quoted as saying that for the cost of operating on the cloud, “I could re-buy the hardware every three months.”
The vast inflation in operating costs is the number one reason cited for cloud repatriation, aka, moving analytics and other workloads off the public cloud and back on premises. This huge inflation in costs is particularly common in companies that have large amounts of data and complex workloads. According to a 2022 IDC report, 48% of companies are repatriating some workloads and controlling costs is the number one driver.
This myth is very emphatically BUSTED.
Other common drivers of cloud repatriation are regulatory compliance, and the need to improve performance, or gain greater visibility and control over workloads.
The big gotcha that many companies discover when trying to move their workloads back on-prem, to another public cloud provider, another region, or even another cloud application, are egress fees, a levy charged simply for moving your data off their cloud. My brain hears Hotel California lyrics every time someone mentions egress fees: “You can check out any time you like, but you can never leave.” Egress fees are egregious lock-in tactics. You’re never assessed ingress fees. If you have any negotiation leverage at all, do your best to demand that you not be assessed these fees.
How Do You Mitigate the Risks of Going to the Cloud? Think Hybrid, Think Long Term.
To move to the cloud or stay on-premises is both a business and technology decision.
1. Hybrid is a practical long-term strategy. If your company has some steady workloads, and other spikey workloads, it makes far more sense to keep the steady workloads on prem than to move everything to the cloud. Studies show that 82% of enterprises are switching to hybrid clouds so they can leverage the elasticity and scalability of public clouds while still meeting performance and security requirements using an on-premises infrastructure. One thing that will make this work is if your end user technology functions the same both on prem and on cloud, so the difference is transparent to users.
2. Ask about egress fees up front. Expect these asymmetrical expenses and budget them in. Change is the only constant in the data management and analytics industry, so expect that at some point your priorities will change and you will want to move your data. Egress fees, if you can’t avoid them, need to be in the budget and part of the plan.
3. Don’t be caught by surprise by cost increases. The advantage of cloud is shifting costs from up-front capitol expenditures to ongoing operating expenditures, not reducing costs overall. And one of the advantages of cloud – making analytics services easier to use – will also result in higher costs as adoption increases. This is a good thing, but plan for it. Don’t be shocked when the bill hits.
4. Put guardrails on auto-scaling. Make sure the cloud infrastructure you choose has guardrails that you can set for auto-scaling. You want the software to auto-scale as demand increases, but you don’t want it to auto-scale out of control. No CFO wants a surprise bill for five times what was budgeted. Set limits and decide under what conditions auto-scaling makes sense for your workload.
5. Know the regulations. If your industry or workload involves certain types of sensitive data or exists in certain regions, the cloud simply isn’t an option for you due to regulatory compliance restraints. Investigate those regulations thoroughly before making a jump to the cloud.
6. Be wary of applications that only optimize with server spread. Many SaaS and IaaS applications provide ease of use at the cost of flexibility and control. Most SaaS applications sell bundled infrastructure and software and get a markup on every bit of cloud compute they use, providing a counter-incentive to stay efficient. Good software optimizes performance without simply throwing more nodes at the problem. This is still true on the cloud, but even more important now that you’re paying for each instance as you use it. When deciding on cloud technologies, compare cost/performance, not just performance alone. Cost-effective concurrency and scalability will matter even more on the cloud than they did on prem.
Whatever route you choose, carefully consider the advantages and disadvantages of all options -- be it the public cloud, multi-cloud, on-premises, or hybrid cloud -- because the decisions made today will have an impact on the future of your organization.
About the Author
Paige Roberts is Open Source (News - Alert) Relations Manager for OpenText Analytics and AI, with focus on Vertica, a scalable analytical database with a broad set of analytics capabilities including end-to-end in-database machine learning. With over 25 years of experience in data management and analytics, Paige has worked as an engineer, trainer, support technician, technical writer, marketer, product manager, and consultant. She contributed to “97 Things Every Data Engineer Should Know,” and co-authored “Accelerate Machine Learning with a Unified Analytics Architecture” both from O’Reilly publishing. Vertica enables data-driven organizations with large and demanding analytical workloads to derive strategic predictive business insights on any data, at any scale, anywhere. For more information about OpenText [NASDAQ: OTEX, TSX: OTEX] visit OpenText.com.
Email: [email protected]