There was significant democratization in the technology sector with the advent of cloud technologies because it dramatically brought down the cost of starting a technology company. It’s fair to say that the cloud can be credited with the emergence, growth, and success of a vast majority of technology companies today. No longer did founders need to pour hundreds of thousands of dollars in expensive capital expenditure towards hardware provisioning. Instead, they could consume hardware resources such as compute power and data storage just like electricity – on a pay-as-you-go model.
As these technology companies scaled and added users at unprecedented velocity, the focus always remained on adding new features that could drive increased user adoption and stickiness. Backed by venture capital and tunnel-visioned by growth considerations, gross margin and technical architecture considerations took a backseat. While private capital financed growing infrastructure costs, technical teams deprioritized building for infrastructure efficiency – an eventuality that companies are recognizing where infrastructure costs continue to erode gross margins.
This article examines a relatively nascent question that CIOs and CTOs are asking themselves – is reverse migration to a purely on-premise or a hybrid on-premise and public cloud possible? If yes, what are the key considerations that need to be kept in mind, and how does one navigate such a complex undertaking.
Why move away from the public cloud?
Before we examine such a migration, let us take a step back and understand the key drivers of this movement. This is important because if you can relate to these symptoms, perhaps it is time for your leadership to prioritize this theme. We classify these reasons broadly under three distinct buckets – technical, operational, and commercial but together they converge to make a compelling case to “un-cloud”, even if it sounds counterintuitive.
Looking at technical challenges first, obvious bottlenecks emerge as applications mature and outgrow the initial growth stage. For example, mature applications increasingly depend on shared enterprise resources – from storage to directory services, user profiles, and data management. These dependencies lead to network-intensive workloads with these resources distributed across several data centers. With public clouds providers, this scenario introduces additional bandwidth costs and, worse, potential performance degradation in user experience.
A similar pattern reveals as applications grow to serve an increasing number of users, engineers need to optimize for performance. One such performance optimization technique involves replicating multiple copies of datasets each of which may be stored separately to reduce application latency. Public cloud services provide limited support for de-duplication and hierarchical storage management of such data, leaving it only to third-party de-dupe and storage tiering services that introduce additional administrative and cost overheads.
On the operational front, IT administrators responsible for managing cloud workloads have frequently expressed frustration with the absence of personalized support and the lack of freedom to perform routine configuration and testing without worrying about associated costs. The majority of these concerns stem from the ability to move data between the public cloud infrastructure and the on-premise one, especially activities that require large data transfers and those that involve integrating existing on-premise backup environments with data residing in the cloud. Administrators have also expressed concerns over understanding and configuring Roles Based Access Control – configurations that vary dramatically from one provider to another and adopt access management philosophies that may not be in line with those of the organization’s. Moreover, administrators and procurement executives get limited visibility on the granularity of consumption usage by various departments as billing is usually consolidated for the entire organization.
The commercial challenges of public cloud infrastructure are not as intuitive as the technical and operational ones but are probably the most important of the three. This is because moving to the cloud is an industry-standard for reducing the dreaded CAPEX that early-stage technology companies so loathe. However, while deploying on the cloud makes complete sense early in the organization’s life, the costs significantly outweigh the perceived scalability benefit as these companies mature. This is important because with scale, comes slower growth and economic efficiency emerges as a driving criterion for the company to transition to the next stage, usually via an exit for initial investors and founders. Despite committed use discounts offered by public cloud providers, infrastructure costs tend to be a significant percentage of revenue.
Planning for reverse migration
Despite these challenges, many companies find it extremely difficult to move off the cloud due to the sheer magnitude of work involved and the variety of technical and downtime impacts such a move would involve. However, the benefits are compelling and so is the need to assist organizations with a tested and validated approach that makes it possible while mitigating all foreseeable risks.
In terms of preparedness, both before and during the migration, CIOs and CTOs must focus on a holistic assessment of their existing cloud deployment followed by a detailed design and architecture of their proposed on-premise landscape. Among other activities, these would include a detailed hardware sizing exercise and planning for the required performance and security aspects of all applications. For a variety of reasons, we recommend organizations consider Converged or Hyper Converged IT infrastructure for their on-premise requirements.
Planning the migration involves detailing to the minutest granularity a cutover plan that includes the sequence of migration of various applications and workloads considering all interdependencies. Key among this is the availability of necessary network bandwidth during the actual migration. Since end-users are bound to be affected, it is necessary to plan for downtime and communicate planned outages in advance for them to minimize the business impact of the migration. Before initiating the actual migration, it is crucial to backup all cloud workloads using full-instance and blocks storage device snapshots, followed by a comprehensive application backup using either native cloud or application-specific tools.
Authentication migrations are usually one of the first steps in the migration process and the process to ensure a smooth authentication handoff depends on the nature of the directory services being used at the organization. Given the plethora of authentication mechanisms in use in various cloud environments, it is important to assess and determine the dependencies of authentication in applications and processes.
Several migrations have been compromised because the team did not ensure the availability of all software licenses. Therefore, the migration plan needs to ensure availability and documentation of all software licenses with a hotline to the procurement department as a fallback for any eventuality.
As your organization matures, you are bound to encounter the question of reverse migration from the cloud, sooner rather than later. Preparedness is key and given the sheer number of areas that could potentially be impacted, it is always recommended to start early. This is the initiative that could save significant percentage points on your gross revenue margins.Asit Sahoo August 25, 2021