Public cloud has helped accelerate digital transformation in the last decade. With the events of 2020 finally providing a nudge to many enterprises, more have realized the benefits. At the same time, organizations understand that those gains cannot be realized by all workloads immediately, or, perhaps, ever.
Source: BrandPost: We’re Seeing a Wave of Workload Repatriation. Here’s Why.
While significant cloud services growth is expected in 2021, evidence shows that some mission-critical workloads that moved to public clouds under a “cloud first” mandate are now coming back on premises.
Is this the start of a larger wave of repatriations and the rebalancing of workloads?
On a world-wide basis, my Hybrid Cloud Practice is seeing a wave of enterprises with mission-critical workloads in or targeted towards public cloud. As our customers exit the disruption of the last 18 months, with new outlooks on resilience, there is a material number that are either cancelling their full migration plan for ERP workloads, and in particular SAP®, or migrating the workload/data back to an on-premises model a year or so after moving to public cloud. The momentum with SAP workloads is particularly strong. Findings from my own SAP consulting and migration activities are also reflected from the Americas SAP User Group (ASUG) that notes enterprise intent to host SAP in cloud has moved from 16% in 2019 to 8% in 20211.
HPE (Hewlett Packard Enterprise) is not alone in seeing the repatriation momentum. IDC recently published “Cloud Growth, Migration & Repatriation Continue to Gain Momentum.” This market presentation highlights the trend that many early adopters of public cloud are repatriating back to on premises. Other organizations echo similar messages, demonstrating that repatriation is just one part of workload mobility or rebalancing as we get better at selecting each destination platform to run a specific workload. The act of bringing a workload back on premises or to a colocation is merely selecting the ideal location/platform to balance data security, sovereignty, privacy, latency, and upgrade requirement constraints.
Let’s examine what our consultants see as we engage with customers across many regions.
Data
Mission-critical apps are at the core of our businesses and connected to many other applications throughout the enterprise landscape. With heavy dependencies on these applications, many clients are finding the fees to move this data in and out of the public cloud much higher than originally budgeted. This presents challenges in accurate planning, both operationally and financially. Outbound data transfers over certain limits, some inter-network data transfers, and data transfers across zones all can cause unplanned fees in a public cloud landscape. Therefore, unpredictable costs to access data across a hybrid environment (public cloud, on-premises compute) is a growing concern.
For example, in working with a large U.S. retailer, we again confirmed many data sources and targets orbit SAP core applications. In several situations, it was difficult to scope the data coming from those sources until near production time. The egress charges are usually an unknown amount until they happen, and this organization found out too late in the implementation process. In this example, the egress charges were 4x original expectations.
Performance/Latency
A global manufacturer based in Europe requested our SAP Performance COE (Center of Excellence) to compare and benchmark multiple platforms. Our findings revealed root causes for latency issues in the application for end users. Deliberately, we ran identical performance tests that were executed on a single cloud platform at various times of the day. This produced a noticeable difference in results across a span of time. These tests revealed that the spread between CPU performance test results were higher than ideal. The conclusion: the end-user experience can be affected at different times of day in an unpredictable manner. This type of testing could help predict CPU bottlenecks that potentially impact user application performance experiences.
In other cases, we have found networking latency causes problems. Overcoming performance and latency challenges in this case is mitigated by deploying higher-bandwidth services from telco providers. This introduces added cost when moving a data rich ERP system to a public cloud.
That said, performance problems can happen on any platform in any location. Organizations just need to be aware of the limitations and plan for the costs to mitigate them.
Rolling Upgrades
Unplanned or forced system down-time is a factor to seriously consider when moving SAP applications to a cloud-native version such as SAP® S/4HANA Cloud. System down-time due to scheduled rolling upgrades is a way of life, as this is how code remains up to date with the latest fixes from SAP. But if you are not used to running your business according to someone else’s patch schedule, you might find this annoying or simply unacceptable. Some organizations can live with this. Most larger enterprises cannot.
In another example, a global manufacturer determined the forced quarterly rolling update process enforced by the hyperscaler/ISV license structure did not align with their business processes updates. Additionally, when upgrades are rolled out, downstream training and MOC (Management of Change) for users has a much larger impact on the business and must be more holistically planned. Business processes must be adjusted, and changed to accommodate an upgrade process that may not align to overall business priorities.
Workload Mobility
Cost and agility seem to be the root trade-off points where mission-critical ERP apps are being tested in public cloud platforms. Many of our customers are in the middle of investigating these choices and we often suggest considering the possibility of limiting the number of times moving an important workload across platforms.
Let’s face it, until there is automatic and complete mobility of mission-critical workloads, it takes time and money to migrate, test, and make the transition to production. Not to mention the increased risks of moving mission-critical workloads your company depends on daily to realize revenues. There will always be good reasons to place key workloads on public and private compute platforms with cloud features. Our goal for our customers is to help look into the future and make the best choices first.
Rebalancing Cloud Workload
Our customers appear to be rebalancing their cloud strategies to their right mix of clouds. Within our “right mix” approach we have a chart we refer to as our “ease and impact” chart of workload migration prioritization. Some workloads belong in cloud; others do not, and enterprises are discovering that and reassessing their IT investment strategy and working to identify trapped costs.
This application repatriation trend is a good reminder that advantages of cloud, such as ease of use amd fast deployments (essentially having the infrastructure outsourced), don’t necessarily translate to cost savings. And clearly there is value gained for public cloud-hosted workloads with certain characteristics, meeting companies with a certain profile of requirements for their data. I’ve commented on this in an earlier article.
If control over your upgrade schedule, application latency, and performance, and predictable costs are important, you might find a hybrid platform architecture is the solution for your enterprise ERP and other mission-critical workloads.
For further information on how to engage with a Digital Next Advisor contact digitaladvisor@hpe.com
1 ASUG Pulse of the SAP Customer 2020 and 2021: What’s on the Minds of SAP Customers (Feb 2020/2021)
____________________________________