云迁移的遗憾：是否应该回迁？_AI阅读总结 — 包阅AI

Enterprises are expressing regret about their cloud migrations. That’s one of the most common sentiments about the cloud we’ve heard in our cloud advisory practice over the past few years.

The legacy capital expense (CapEx) model was slower and more bureaucratic than a cloud operating expense (OpEx) model, and the cloud was viewed as a panacea that would increase business agility and shorten time to market. Enterprises pushed workloads to the cloud, regardless of appropriateness. The transition was more complex than anticipated and often took much longer. Once migrations were completed, the costs were significantly higher than on premises.

How the Cloud Failed to Live Up to Its Promise

For enterprises, the cloud feels like a black box that hides its true financial cost while failing to deliver business agility. Hindsight provides valuable lessons on how we got here:

Expensive Lift-and-Shift Strategy

One of the biggest problems we see is that businesses lifted and shifted applications, thinking they would slowly rewrite them, but then they didn’t. These are generally the most expensive types of workloads in the cloud. We’ve seen a company’s entire yearly cloud budget drained in a month because tightly coupled legacy applications were not rewritten to take advantage of cloud infrastructure.

Cloud Technology Is Disconnected From the Business

The shift to cloud moved a lot of the decisions from the business to the technology group, significantly decreasing transparency. If you look at a cloud bill, it can be millions of lines long with very specific technical information: This data was transferred here. This CPU was used this much. But that doesn’t mean anything to a business. The visibility and traceability of costs were lost, and cost/benefit decisions such as the level of system resiliency or application performance appropriate for the business were shifted to technologists rather than financial analysts.

Over-Rotating on Developer Flexibility

The shift to programmable cloud infrastructure saw a concurrent shift to developer autonomy. Many of the guardrails provided by the enterprise architecture and infrastructure teams were removed in the name of developer flexibility. Developers were free to spin up cloud resources, configure their own virtual machines and even decide which cloud provider to use, on a per-application basis. They were essentially free to use the fanciest, coolest, most modern technology. It is a trap that enthusiasts, and true believers in the cloud like me, can easily fall into.

Should You Repatriate Workloads Back to on Premises?

With increasing pressure to cut costs, many CTOs and CIOs are considering repatriating cloud workloads back on premises. As hard as it may seem, it’s important to think beyond just the cost. You must understand workload requirements to make sound decisions for each application. For example:

• If your application needs to scale quickly, say an e-commerce site that experiences sudden spikes in traffic, you will need an equally rapid expansion in capacity. That’s not possible on premises.
• AI workloads are best kept in the cloud. They require specialized hardware that cloud vendors can readily provide, and those vendors are rapidly building additional infrastructure in anticipation of future needs.
• Big data is one of the top workloads that has been moved to the cloud. Cloud vendors offer many tools and services that make analytics processing fast and efficient. It’s not necessarily less expensive in the cloud, but this is a good example of where the value gained in the cloud is worth the additional cost.

A lot of organizations have forgotten how much IT operations have changed since moving to the cloud. Cloud transformation meant revamping ITOps based on the chosen mix of Infrastructure-, Platform- or Software-as-a-Service (IaaS, PaaS or SaaS) services. Bringing applications back on premises strips away those service layers, and Ops teams may no longer be able or willing to accept the administrative and maintenance burden again.

One final consideration before moving workloads off the cloud is security. I think security is one of the many advantages of cloud infrastructure. When businesses first started moving to the cloud, security was one of the biggest concerns. It turns out that cloud providers are better at security than you are. They can’t fix security holes in your software or other operator error scenarios, but a cloud infrastructure provides greater isolation if a breach does occur. When a data center firewall is breached, the hacker is in your data center, not in a cloud instance that doesn’t connect to anything else.

The Value of the Cloud Is There

Enterprises were expecting cost savings from the cloud’s OpEx financial model, but the way I see it, server virtualization was the big cost saver for data center budgets. Once everything was virtualized, there weren’t many additional cost savings left to realize.

But the actual value of the cloud is still there. Cloud is more reliable, more secure and more scalable than an on-premises data center, and cloud native application development and programmable infrastructure result in exponential improvements in developer productivity and business agility.