Returning to Roots: The Trend of Reverse Application Migration from Cloud to On-Premises Infrastructure
The initial business case proposed for migration to public cloud services was appealing to the businesses as it promised cost reduction
Fast forward to the present, five years post-pandemic, businesses are now encountering some challenges associated with this cloud migration. While I don't claim expertise in Cloud technology, these are my general opinions based on the current business and technological environments.
1.???? High Cost
There is recognition that public cloud services have become significantly more costly, and the expected cost savings have not materialized as initially estimated. This observation remains valid. Below, is the spending forecast provided by Gartner. The image will take you to the press release.
2.????Lack of cloud optimization
The second problem arises from unsuccessful migrations, especially during the pandemic when organizations moved significant workloads to the cloud without cloud optimization / modernization. Since optimizations needed time and cash, they opted for a lift-and-shift approach
3.????Diminishing need
Over the last three to four years, there has been a worldwide increase in startup numbers and the race to attain unicorn status. Moreover, the pandemic pushed many businesses to the brink of insolvency. Consequently, not all existing businesses and startups have abundant financial resources. While cloud costs may benefit startups with substantial user bases (e.g., 10,000, 20,000, etc.), they can lead to excessively high bills for those with smaller user bases.
?4.????Hidden costs
The financial tools provided by hyper scalers may initially seem straightforward, but upon closer examination, hidden within the details are additional costs associated with each layer of service and support. For instance, crucial security measures come with a premium price tag. Similarly, the pricing structures for storage can be ambiguous; while the initial cost per terabyte appears attractive, companies may find themselves charged not just for data storage but also for data deletion. While uploads might be free, organizations incur charges for every object downloaded. As a result, monthly bills often exceed projections, sometimes doubling or even tripling the expected amount, leading to budget shortfalls.
Moreover, bandwidth limitations, extra fees for CPU or RAM usage, and the requirement for businesses using VMWare to pay again based on these same usage factors exacerbate the situation. It's no wonder that the actual expenses associated with utilizing public cloud services frequently surpass the initial expectations of any Chief Technology Officer (CTO).
Business Agility - Going back to the Drawing board
Considering the challenges encountered in today's competitive world, it's prudent to revisit the drawing board and reevaluate the initial business case with the new financial parameters - those are changed with time.
An increasing number of businesses are considering a reverse migration
领英推荐
1.???? Cheaper hardware
Over the past decade, the prices of hard drive storage, networking hardware, computer hardware, power supplies, and other technological equipment have decreased.
Now, it is the most logical use case to bring back workloads and data storage that primarily serve basic functions, such as long-term data storage without advanced data processing (e.g., no advanced artificial intelligence or business intelligence). These workloads can be relocated to owned hardware, resulting in a net gain in return on investment (ROI). Even with the additional costs associated with taking over and internalizing operations, the enterprise can save money (or a substantial amount) compared to equivalent public cloud hosting.
?2.???? Modern infrastructure
For any business concerned about the necessity of rebuilding a server room or hiring dedicated technical personnel, neither presents an obstacle. The latest generation of servers can operate at higher temperatures, eliminating the need to replicate the air-conditioned server rooms of the past. These servers can conveniently be housed within existing network rooms or office spaces. Alternatively, if space is limited, the entire system can be securely co-located within a dedicated and locked rack. Technical support
?3.???? Ideal business case for startups / small businesses
During their initial phases or when serving only a small user base, startups find it cost-effective to operate with on-premise setups.
?4.???? No hidden costs
As on-premise setups and data centers are owned by the businesses themselves, they offer inherent transparency, facilitating easy cost management
In recent years, as businesses enthusiastically adopt the cloud for existing and new applications and packages, not all are thoroughly assessing costs and benefits in their decision-making process. Rather, decisions are frequently swayed by the imitation of competitors, industry or market trends, and a "Me Too" mindset, among other factors.
There are some pioneers in this area, who have successfully done this reverse migration and are now acting as a lighthouse to guide other sailors to reach their destinations. Hey.com is one of them who have also published an FAQ on this topic. The Big Cloud Exit FAQ (hey.com)
In closing, I'd like to leave you with the words of Hendrith Vanlon Smith Jr, who famously said...
Every business is unique. Every business must dance to the beat of its own drum and sing it’s own song.
Account Delivery Head @ Wipro Technologies
11 个月Very well written ????
Technology | Architecture | Leadership | Project Delivery
12 个月A few weeks ago, I engaged in a conversation with my colleagues on the very same topic. It appears that the world is poised to embark on a different trajectory soon.
Head of Cyber Protection & Digital Asset Security | IT/OT/ICS Security | SABSA, AWS Security
1 年Returning to the on-prem can require a lot of work for those who have refactored their architectures to take advantages for cloud native services. The amount of effort might be even more than what they spent on refactoring when migrating to cloud in the first place. There are also a lot of workloads which were built from ground-up to be cloud native. As the need for cloud admins surged, it also changed what the new to-be admins studied. It could be that there aren't that many juniors/mid-level admins who have the knowledge of maintaining hardware and on-prem systems (just my assumption). It's special knowledge need to know which Bios versions work with what hardware, and how to diagnose tricky network issues. A long, long time ago, when virtualisation came, there were these same expectations from customers, that when you migrated your workloads lift-and-shift to a virtualisation platform, you got cost savings. And it was needed to explain to customers, that you now have also the virtualisation platform to manage. It's important to have good reasons to make such moves, like having the flexibility, pays-as-you-go, scalability - and if organisation is willing to pay extra for these features.
Interesting read. Here’s my take. Ultimately the workload are going to be a commodity. The real action will happen on Kubernetes. If workloads are clntainerized, you can run them on Prem or on cloud. That’s where solutions like Redhat OpenShift and SUSE Rancher become very very relevant. The beauty of CNCF projects is you can run containers on bare metal using projects like Kubevirt.
Security Architecture and Engineering | CISO | Business Enablement | Cybersecurity Operations | Security Evangelist | Application Security | High-Performance Leadership
1 年Interesting observations. Cloud Repatriation is a growing sentiment in the industry. However, there is a catch to this in the GenAI times we live in. Software components required for LLMs are bloating up in size which will require significantly more hardware resources including resources that we didn't think we would need - GPUs or LPUs. From the GenAI perspective (enterprises will most certainly adopt them), hosting LLMs on-prem can cut costs and offer predictable scaling. Yet, an "expand into cloud strategy" might be more efficient, using cloud resources when on-prem capacity maxes out. This hybrid approach requires a flexible system architecture but offers the best of both worlds. As cloud hardware evolves, hosting LLMs may shift, incorporating technologies like WebGPU and on-prem, expandable PaaS workloads.