Prepare for the complexity of agile change management - Cloud Adoption Framework (2023)

This article describes a high-level approach to agile change management in migration efforts. We will discuss the role of transition, release, and retry delays in migration setup and execution.

Figure 1: Relationship between migrations, releases, and backlog of iterations.

Three different backlogs correspond to different planning levels in the Scrum framework:

• Pending immigration =epic
• pending release =user history
• iteration pending =assignment

These delays will keep your migration organized and allow you to deliver value faster.

Define the workload

Migration should focus on moving the workload. So your backlog becomes your workload. A workload is a unit of application functionality. Defining the workload can be a challenge. That's why we've created a simple assessment guide using the familiar acronym INVEST to help you determine your workload.

The workload should be:

• independent:Workloads must not have unreachable dependencies. Before migrating the workload, all dependencies must be available and included in the migration attempt. Multi-wave migration must take temporary mixed connections into account.
• Negotiable:During the discovery process, your understanding of your workload may change. You can discover dependencies to include in your workload during discovery. Examples include pre-release operations, the sequence of migrating components in a multi-wave migration, or a single-release migration.
• precious:The value of a workload is measured by the ability to provide users with access to production workloads.
• discretion:Dependencies, assets, migration time, performance and cloud costs should all be quantifiable and should be estimated before migration.
• lil:The goal is to pack the workload into one iteration. When a single iteration is impractical, you should schedule iterations to minimize the time required to move the workload to production.
• Test bar:There should always be a defined method for confirming the completion of workload migration.

The workload you specify will be your migration backlog.

pending immigration

The migration backlog allows you to track the best combinations of workloads to migrate. Before migration, cloud strategy teams and cloud adoption teams should review the currentdigital real estateAnd agree on a prioritized list of workloads to migrate. This list forms the migration backlog.

Initially, workloads retired due to migration are unlikely to meet the INVEST criteria described in the previous section. Instead, they will serve as logical groups of assets in the initial inventory that serve as member states for future work. These placeholders may not be technically correct, but they are essential to align with the business.

In each pending migration file, the change management team should try to obtain the following information about any workloads in the plan. This data should at least be available for all workloads prioritized for migration in the next two or three releases.

Pending immigration data points

• business shock.Understand the impact of missing expected schedules or reduced functionality during a window freeze.
• Relevant business priority.A list of workloads organized by business priority.
• owner.Document who is responsible for making business decisions about this workload.
• technology owner.Document who is responsible for technical decisions related to this workload.
• expected schedule.When will the immigration plan be completed?
• Workload freezes.The workload is not appropriate for the time frame of the change.
• Workload name.
• initial estimate.Any component that provides functionality to a workload, including virtual machines, IT equipment, data, applications, development pipelines, and more. This information may not be accurate.

pending release

a migrationreleaseDeploy one or more workloads to production. A release usually covers several iterations or technical tasks. However, this is a single iteration of business changes. A release occurs after one or more workloads are ready for a production upgrade.

The decision to package a release is made when the migrated workload represents sufficient business value to justify introducing changes to the business environment. liberation is throughbusiness change planbackoperational testIt is done. The cloud strategy team manages planning and monitors release execution to ensure the necessary business changes are released.

INpending releaseis the determination of the future state plan. Release pending is business change management (pending immigration) and technical change management (pending rerun). The release backlog consists of a workload list with pending migrations that correspond to a specific subset of business outcomes. Defining and submitting pending releases to the cloud adoption team can trigger deeper analysis and migration planning. Once the cloud adoption team has validated the technical details associated with a release, they can choose to commit that release and set a release schedule based on current knowledge.

Given the level of analysis required for a release, cloud strategy teams should maintain a running list for the next two to four releases. Teams should also try to confirm as much of the following information as possible before setting and committing to a release. A disciplined cloud strategy team that can support the next four releases can greatly improve the consistency and accuracy of release plan estimates.

Release accumulated data points

A collaboration between the cloud strategy team and the cloud adoption team adds the following data points for any workload to the traffic backlog:

• Inventory improvement.Identify the items to migrate. You often validate components with logs or trace data at the host, network, or operating system level to ensure that each component's network and hardware dependencies are properly understood under typical loads.
• Use function.Understand end-user usage patterns. These patterns typically include analysis of end-user geographic distribution, network routing, seasonal usage peaks, daily/hourly usage peaks, and end-user composition (space vs. remote).
• performance expectations.Analyze available log data to capture performance, page views, network routing, and other performance data needed to reproduce the end-user experience.
• DEPENDENCE.Analyze network traffic and application usage patterns for other workload dependencies. You should factor these dependencies into sequencing and environment readiness. Do not include workloads in a release until one of the following conditions is met:
  • All relevant workloads have been migrated.
  • Network and security configurations have been implemented to allow workload access to all dependencies that meet existing performance expectations.
  • For multi-wave migration, mixed connections may cause temporary performance degradation. You should discuss this possibility with your company manager and develop a plan of action prior to relocation.
• Your preferred method of immigration.For the immigration backlog, hypothetical immigration attempt is the only consideration used in the analysis. For example, if the business outcome is an exit from an existing data center, all migrations are considered to re-host scenarios in the migration backlog. In the release backlog, cloud strategy teams and cloud adoption teams should evaluate the long-term value of investing in additional features, modernization, and continued development. A more modern approach could be better.
• business testing standards.After workload is added to the migration backlog, the test criteria should be agreed upon by both parties. In some cases, the test criteria may be limited to performance testing of a defined group of power users. However, automated performance testing is essential and should be included for statistical validation. Existing application instances usually lack automated testing capabilities. It is not uncommon for cloud architects to work with power users to create baseline load tests against existing solutions to create a baseline to use during migration.

Release pending rate

In a mature migration, releases occur on a regular basis. The velocity of cloud adoption teams typically normalizes to releases every two to four iterations (roughly every two months). Creating an artificial release rate can negatively impact the cloud adoption team's ability to achieve consistent performance.

To stabilize the business, the cloud strategy team should maintain regular dialogue with the business by establishing a monthly release process. They should also set expectations, it takes months to predict a normal traffic rate.

pending rerun

inrepetitionA task is a consistent, timed unit of work. Iterations are called sprints in Agile Scrum. Repetitions are two to four week periods of time. Timed repetitions enforce fixed intervals at which the task is performed and allow for more frequent adjustments to the plan based on new learning. During an iteration, there are typically jobs to evaluate, migrate, and optimize the workload defined in the migration backlog. You should track and manage these work items using the same project management tools that you use for migration and the release backlog. It will help maintain consistency at all levels of change management.

In an iteration, the cloud adoption team will work to achieve the promised technical effort leading to the migration of defined workloads. Once complete, these efforts can be tested by validating the production readiness of the cloud workload.

Large or complex repeat structures

For small migrations with separate migration groups, a single iteration can include all four phases of a single workload migration (Evaluate,migrate,optimization, andprotection and management). More commonly, each process is shared by multiple groups on different work objects in an iteration. Depending on the type of work, the size of the work and the role, repetitions can be of different types. We will discuss two types:

• Factory migration.Large-scale migrations sometimes require a factory-like approach to the execution model. In this model, you assign individual teams to perform specific migration processes (or subsets of processes). When complete, the result of a group iteration fills the backlog of the next group. The factory model is suitable for large-scale "lift and shift" migrations. It efficiently moves workloads of thousands of virtual machines through the assessment, architect, repair and migration phases. However, you need to have a streamlined change management and approval process to track transfers.
• Multibølge migration.Another approach that works well for large migrations is the wave model. In this model, the division of labor is not so clear. The team is working on performing the migration process for a single workload. However, the nature of each iteration changes. During an iteration, the team can complete evaluation and architecture tasks. On the other hand, it can do the job of immigration. In another example, the focus may be on product optimization and launch. This approach allows the core team to adapt to the workload and understand the entire process. However, skill diversity and context changes can slow down migration efforts by reducing the potential velocity of teams. Additionally, hurdles in the approval cycle can cause significant delays. With this model, it's important to keep opportunities in the backlog to keep the team moving during blocks. It is also important to cross-reference team members and ensure that skills align with the theme of each iteration.

Redo outstanding data points

The results of the iterations capture and record the changes made to the workload, closing the change control loop. Once done, note the following points. This documentation must be completed in relation to the technical work points when performing the iteration.

• Activation has been entered.Any component that is deployed in the cloud to host workloads.
• treatment.Any changes to assets in preparation for cloud migration.
• configuration.Selected configurations for all deployed components, including any references to configuration scripts.
• Apply the model.The method used to deploy the component to the cloud, including references to any deployment scripts or tools.
• architecture.Document the architecture implemented for the cloud.
• Implementation.Run output from automated tests or business tests to confirm performance during development.
• unique requirements or configurations.Any unique aspect of the implementation, configuration, or technical requirements required to run the workload.
• Operation approval.Obtain post-implementation approval from the application owner and IT operations responsible for workload management to confirm operational readiness.
• Architecture approval.Get approval from workload owners and cloud adoption teams to confirm any architectural changes required to host each component.

Next step

With a change management approach, it is time to address the final conditions,Immigration Pending Review