I was reading an article on The Stack a few weeks ago (https://www.thestack.technology/warren-buffetts-geico-repatriates-work-from-the-cloud-continues-ambitious-infrastructure-overhaul/), which was talking about their cloud journey, and more importantly about a recent decision to move some workloads back from the cloud to on-premises, the main reason provided in this article for the move was cost and complexity related to storage, and compliance considerations. It should be noted that the planned migration to the cloud had not been fully completed, which I suspect also impacted this decision.
This is not the only example I have come across recently, and I think cloud vendors are also taking this direction of travel on board, for example, Microsoft’s recent expansion of the Azure HCI Stack product line to include Azure Local and the hybrid-cloud features of Azure ARC are both aimed at supporting clients who want, or need to retain an on-premises estate.
I am also not surprised by this development, which stems from an incorrect understanding of the benefits of cloud computing, especially in terms of IaaS and PaaS services. For better or worse, I have worked in the cloud space since roughly 2011, including for Microsoft, and there are a couple of misconceptions that seem prevalent, which I do not think are that contentious.
Please note the following is more a stream of consciousness than a properly structured, organised post, but I hope it hangs together .
Choosing the right cloud strategy, or not all workloads belong in the cloud.
When creating this article, I thought I would revise what the Gartner 5 R’s actually was (https://www.gartner.com/en/documents/3893681#:~:text=When%20moving%20application%20workloads%20to%20cloud%20platforms%2C%20organizations,the%20right%20one%20for%20their%20goals%20and%20priorities ), and I was surprised to find that the earliest reference I could find (not an extensive search, to be transparent), consisted of the following:
- Rehost
- Revise
- Rearchitect
- Rebuild
- Replace
I was quite surprised that an “R” I always included when discussing cloud migrations did not appear in this list, which was Retain.
I understand that retaining a workload on-premises when the whole purpose of cloud migration is to migrate applications to said cloud seems anathema to most people and organisations. However, it is perhaps the most crucial option when discussing cloud migrations. The reason why, at least to me, is very simple. Some applications, usually legacy/monolithic, are either too expensive, complex, or have some regulatory compliance requirement that makes moving to the cloud financially and technically inadvisable, especially when this is done in a pure lift and shift approach, and that latter phrase is key. I have seen this several times where an organisation (often aided by a consultancy) embarks on a cloud migration to close one or more data centres, this is usually influenced by some sort of impending deadline, usually a contract renewal/extension date for the data centre. As a result, there is an understandable desire to migrate workloads at pace.
This tends to lead to the organisation adopting a Lift and Shift approach, with a transformation planned once the migration has been completed. However, the priority is to get the workloads out of the data centre, which tends to leave organisations in a position where they have, in effect, just moved their technical debt from one platform to another, which also may not be the best platform for that technical debt. Once migrated, the desire and budget for the transformation sometimes never materialise, and you are left with a solution not designed for running on the cloud, which is constantly (due to the nature of the cloud) incurring (potentially inflated) costs.
If I circle back to the original question, I think it is imperative that when organisations consider a cloud migration programme (for whatever reason), they need to consider not only the typical R lane treatment but also at least some of the below.
- Understand the current estate
This is often more difficult than it seems, I have seen a lot of migrations where the discovery process has focused purely on the infrastructure sitting in the DC, and the output is based on some discovery tool or existing CMDB. What these outputs typically lack is an understanding of the associated costs, integration points, compliance requirements etc. It then provides a false sense of confidence around the disposition of that solution and the potential cost savings gained by migration.
In this instance, the estate includes everything related to the servers, storage, and networking components, as well as surrounding elements like support contracts, license terms, resources, skills, etc. All of these need to be included if an organisation is to understand its future costs. - Understand the future estate
It is a simplistic statement but very important. This is theoretically simple for a single cloud target. But I am not sure there is such a thing anymore. Even if you stick with one vendor, say Microsoft, is Azure and M365 one cloud? I am not sure, and even so, you have to be able to understand the integration points and what the implications of this integration would be, i.e. storage, license, and development costs, let alone any compliance or regulatory controls that may be required.
But suppose you extend this to a multi/hybrid cloud environment. In that case, this becomes incredibly important, complicated and raises several questions, which, off the top of my head, include storage and data exchange architectures, integration, HA/BC, licenses, management overhead, support costs, contract management, supply chain security/management and much more. - Have a clear idea of what can or should be migrated.
From a purely technical perspective (business decisions notwithstanding), it is OK to leave workloads on-premises if migrating them is not beneficial. Yes, this may lead to a more complex environment, e.g., a hybrid cloud, and may require that DC remain (hopefully with a smaller footprint and cost profile, etc.). However, this is a better scenario than having to endure a repatriation effort a few years later, as appears to be occurring. - Complete the transformation
In my experience, many migration programs are designed to transform the applications once they are migrated to the cloud. However, for various reasons, this does not always happen. Organisations should adopt cloud-aligned (Please note I am not saying cloud-native) architectures to benefit from the cloud most. - Constantly Monitor/Revise
One of the benefits of the cloud is that it is constantly evolving, and new features/services are constantly being released and integrated into services; for example, the adoption of Gen AI tools would not have been possible without the cloud. But this also adds challenges; what may be a good design for a cloud solution now may not be the case in a year’s time. Organisations have to stay on top of the changes in best practices and be constantly evolving the estate to take advantage of or maintain/reduce operating costs.
This last point may be the most difficult, and I have seen where organisations adopted the cloud several years ago, and implemented landing zones and application architectures that were aligned with best practices at the time, but now do not align, and this has and continues to cause “issue”.
There are several other recommendations that I could have added here, but I think this post is probably a bit too long already, but if anyone has got this far and thinks I have missed something, or disagrees with any of the points, please feel free to let me know in the comments.
In conclusion, I think it is almost inevitable, considering the rush to the cloud, that some organisations are considering repatriation to an on-premises environment. Unfortunately, I suspect this trend will continue for a time, as it probably takes 2-3 years before organisations realise the impacts of some of the decisions made when they initially embarked on their cloud migration. However, it is important to note that it is entirely possible to have a successful cloud migration, as long as several key principles are applied before, during and after the migration.
Elsewhere
Copilot Chat
I commented in the last pod about how I would like Microsoft to be consistent in its licensing model for the various Copilots; this was about the fact that Security Copilot is consumption-based, as opposed to license-based. I was, therefore, a little “surprised” to find that Microsoft has doubled down on the inconsistent licensing approach for Copilot, with their announcement of Copilot Chat, which is a consumption variant of M365 Copilot. Microsoft are positing this a an “on-ramp” to the full M365 Copilot experience, which I suppose makes sense, but also fits into my concerns that we are still missing the killer business application for Gen AI.
Introducing Microsoft 365 Copilot Chat
https://www.microsoft.com/en-us/microsoft-365/blog/2025/01/15/copilot-for-all-introducing-microsoft-365-copilot-chat/?msockid=2db0be1eba176e053806ab29bbf76ff3
IBM CSS Whitepaper on SOC Transformation using Microsoft Sentinel and Defender
Read the full whitepaper to learn more about each phase of the TDR Method and how IBM Consulting can help you achieve your security goals.
https://www.ibm.com/downloads/documents/us-en/10a99808baafd518
Watch the on-demand webinar where our experts discuss the benefits of platformisation on the Microsoft platform, including Microsoft Sentinel and Defender and how we infuse AI to accelerate threat detection and response.
https://ibm.webcasts.com/starthere.jsp?ei=1696478&tp_key=f457020e57
Understanding Zero Trust in the Microsoft Cloud: Part 3
Part 3 of my series on Zero Trust in the Microsoft cloud covers several basic concepts, such as MFA, Cloud and Hybrid Identities, RBAC and Identity Lifecycle by using Access Reviews.
2024 Retrospective and Looking forward to 2025 | Cloudy with a Chance of Insights | EP04
In this episode of ‘Cloudy with a Chance of Insights’, the hosts reflect on the past year in Microsoft Cloud, discussing key highlights and looking forward to future trends.
They emphasize the growing importance of security, shifts in security practices, and the shared responsibility model. The conversation also delves into challenges in cloud migration, the future of quantum computing, and the intersection of AI and quantum technologies.
https://themicrosoftcloudblog.com/2025/01/10/cloudy-with-a-chance-of-insights-ep04/
