I. What is cloud repatriation?
Public cloud can undoubtedly improve the agility, flexibility, and synergy of enterprise services. However, it may also present challenges such as higher cloud migration costs, deteriorating performance, and higher security risks. For example, improper cloud migration policies can lead to continuous performance deterioration, compelling many enterprises to move some or even all of their applications off of public clouds. This reverse migration of services from public clouds back to enterprises' own data centers, and the replacement of public clouds with hybrid or private clouds, is known as cloud repatriation. According to IDC's August 2021 cloud repatriation report:
- 80% of respondents repatriated their workloads from the public cloud in the past year.
- Repatriation rates increase when public clouds cost more than alternative computing solutions.
- More than 70% of enterprises now use multiple cloud solutions.
II. Implications from a case study
Seagate CIO Ravi Naik's cloud migration experience highlights the problem of growing costs. After migrating a large amount of its services to the AWS public cloud, elastic computing enabled Seagate to integrate its four data centers, which were located around the world, into one data center and modernize its IT infrastructure. However, Seagate soon found that costs were rising due to the higher volumes of data being generated and bandwidth requirements in one of its big data systems running on AWS. This reached the point where it was cheaper for Seagate to simply use its own data centers.
Therefore, Naik led the Seagate team to reassess the company's cloud strategy, and used cloud-native architectures and tools to improve its big data analytics solution. On-premise testing showed that the new system could run a more efficient ecosystem beyond the public cloud. The new operating environment was a multi and hybrid cloud environment that included AWS, Azure, and Seagate's own data center. The team therefore moved the data lake application system off the public cloud and deployed it in a private cloud.
TCO was a major factor in the company's decision to repatriate its big data system. Reducing TCO and optimizing processes cut expenditure by 25%. Seagate generates more than 30 terabytes of data per day and while only a portion of this is moved into its data lake, transferring data between seven plants on three continents costs millions of dollars a year. Following the repatriation, Seagate was able to transfer large-scale datasets across the globe at a fraction of the cost of transferring data over the WAN and data acquisition time was shortened from weeks to just 72 hours.
III. Analysis of the causes of cloud repatriation
Through years of examining carriers' cloud transformation practices and observing numerous cases of cloud repatriation, the main reasons for cloud repatriation are summarized as follows: are as follows:
1. Services are subject to security and compliance restrictions
Public cloud security must be improved. Common issues include incorrect cloud platform configurations (71%), sensitive data breaches (59%), and unsecure APIs (54%).
Frequent cloud security incidents caused by data and security breaches mean that security and regulatory issues often force governments and enterprises to opt to repatriate their applications to private clouds or their own data centers. On-premises solutions typically feature robust overall management and well-defined security boundaries, leading to more effective data control. For this reason, many enterprises use hybrid cloud (44%) and multi-cloud solutions (43%) for data backup and redundancy. This will become a trend over the next few years.
Technically, public cloud and private data centers are equally secure. However, any issue that arises with public cloud may cause irreparable damage. Given the nature of remote access, refined access control in public clouds, and the additional security and compliance requirements of modern enterprises, there is no question that some critical applications are simply not suitable for deployment in a public cloud.
2. The cost of cloud migration remains high
In terms of computing resources, public clouds are highly scalable and agile, and can provide inexpensive capacity expansion to meet surging computing capacity needs. However, public clouds are not cost-effective for applications that are deployed year-round at scale. Ultra-large-scale instance deployments consume huge amounts of computing resources, resulting in higher cloud expenses. This is particularly true for legacy systems that have simply undergone lift-and-shift migration, rather than refactoring or rearchitecting. In such cases, massive computing resources cannot be promptly released even when applications are idle. In addition, computing resources will not be scaled down as needed after being scaled up, resulting in wasted resources. Users see costs rising, but neither cloud solution providers (CSPs) nor users can do anything about it, making cloud repatriation unavoidable.
In terms of storage resources, the cloud-related cost of an application can involve many hidden and changing resources. As services expand, the demand for data storage and cloud-to-local data transmission increases. Initially, storage and transmission costs are not significant during the planning of public cloud deployment. However, over time, as costs start to exceed the budget, the reverse migration of applications or data becomes inevitable.
In terms of services, vendor lock-in can be a reason for customers to opt for cloud repatriation. An over-reliance on the tools and systems of a single CSP can leave customers with no choice but to switch to alternatives when the CSP starts charging more for the same services.
3. Cloud migration paths are unclear and service performance deteriorating
Migrating applications to the cloud requires comprehensive planning and extensive testing. Many services that are migrated to cloud are not designed for cloud and not refactored for cloud before migration. This is particularly true for enterprise applications developed in-house, where the cloud architecture does not match deployment environment requirements. As a result, latency becomes an issue and repatriating applications back to on-premises deployment is the simplest solution for enterprises.
In addition, public cloud performance (for example, data transmission rate to and from the cloud) varies greatly. This can make resources inaccessible to some applications that require them when WAN bandwidth is inefficient or the CSP's overall service volume is high.
4. Service availability does not meet SLA requirements
A public cloud system outage may last for hours or more, affecting numerous customers. When this occurs, cloud customers can only rely on CSPs' allocation of backup resources and remedial efforts under SLAs.
5. Carriers lack technological innovation capabilities
Driven by new technologies, enterprises must modernize applications that have been migrated to the cloud, including adopting microservice-based and container-based architectures. Cloud-native, for example, consists of applications and the supporting cloud environment. Based on cloud-native infrastructure and the platform tools provided by CSPs, enterprise applications, both ISV-provided and self-developed, must offer a microservice architecture with features such as modularization, orchestration, and container-based deployment. Generally, during IT modernization, the biggest challenge faced by CIOs is making legacy applications cloud-native. How can applications be optimized to become cost-effective? When application modernization encounters challenges, such as resource integration, TCO, and data security, it is reasonable to migrate applications back to on-premises deployments.
6. Carriers lack cloud service management skills
Another challenge to public-cloud-based enterprise applications is changes in cloud-related management, processes, and employee skills. Many regard the cloud as just another form of data center, and thus take a conventional approach to the public cloud, expecting the required management, processes, and skills to be similar to those of corporate data centers. However, this approach means missing out the opportunities that the cloud truly offers. It also creates additional challenges in areas such as architecture, processes, network security, and stable operating environment. Faced with such challenges, enterprises often consider moving their applications back to corporate data centers.
IV. Six measures to avoid cloud repatriation
The core reason for cloud repatriation is problems in the cloud migration path. Efforts in the following areas will help carriers reduce cloud repatriation and improve service levels.
1. Proper planning
The primary cause of many companies' cloud repatriation is a lack of proper planning, with many questions unanswered: Which applications are suitable for cloud migration and which are not? What preparations should be made before migrating to the cloud? How can we maximize business benefits by leveraging public cloud technology?
CSPs might not understand a customer's services, but they can help by playing a leading role throughout the lifecycle, including strategy development, migration planning, migration implementation, O&M, and technology upgrades. This approach can prevent customers from aimlessly migrating to the cloud.
2. Data sovereignty
A core reason for cloud repatriation is data security compliance. Deep insights into telecom network data are indispensable for developing solutions that properly support data security compliance and balance data sovereignty and service needs.
3. Training on cloud technology and O&M
A lack of necessary cloud technologies and cloud O&M expertise means that it is impossible for many enterprises moving to cloud to maximize the benefits. When problems start to arise, these enterprises often make the unwise decision of repatriating from the cloud. Therefore, when developing cloud strategies, carriers should come up with both cloud technology training and cloud O&M training plans to prevent losses caused by unnecessary cloud repatriation.
4. Hybrid cloud model
Enterprises can combine public- and private-cloud and on-premises resources to gain the agility required to gain competitive advantages. Hybrid cloud also enables enterprises to ensure that critical data stays on-premises for enhanced performance efficiency, while also meeting national/regional regulatory, data localization, and privacy regulations. This makes the hybrid cloud model the best choice. During communication with enterprises and solution design, carriers should help enterprises design a sound hybrid cloud architecture based on the nature of their services and the related requirements. Carriers should also provide cloud migration planning and design services. This will help enterprises develop effective cloud migration methods and application selection methods.
5. Cloud-native preferred
New applications should employ cloud-native architecture, so that microservice architecture can be orchestrated, distributed, and containerized. Such applications will bring value to customers on the cloud and eliminate the need to repatriate their services.
Carriers must help customers analyze legacy applications and provide appropriate suggestions and feasible upgrade solutions such as re-hosting, platform updates, rearchitecting, and replacement. Carriers should also make use of TCO and ROI analyses to show customers the value of cloud migration.
6. Digital priority strategy
In line with industrial digitalization, carriers are using new technologies, such as AI, 5G, big data, and cloud, to accelerate the digital transformation of industries. As top carriers shift from prioritizing cloud to prioritizing digitalization, they should focus on value-driven transformation based on customers' business scenarios. This will ensure that applications migrated to cloud can significantly benefit enterprises.