A major consideration is whether you want to either:
- Upgrade the existing installation in place.
- Migrate the content from the current system to a fresh, new installation.
When moving from a previous version to the current version there are two options:
For either option, remember to read the Release Notes appropriate to the target version. If you have skipped any versions then it is recommened to read the release notes for those as well.
For example, see AEM Release Notes.
Metrics are used to define quantitative measurements for the quality of your website - they are basically a definition of the performance goals that you want to achieve.
Many metrics can be defined, but often the ones you define cover your goals for performance and concurrency. In particular, factors which can be difficult to quantify, and are often prone to emotional assessment:
- "our website is much too slow today" - when does slow qualify?
- "everything grinds to a halt when my colleague logs in" - how many concurrent users can the system support?
- "when I search, the system grinds to a halt " - which sort of search requests are impacting the system?
- "it takes ages to download the file" - what are acceptable download times (under normal network conditions)?
- indicate the expected dimensions of the website you will offer
- indicate the minimum quality which you want to achieve
- define how these factors will actually be measured
- if set too high they may be completely unattainable
- if set too low fluctuations may not be highlighted
- to ensure that they can be repeatedly and consistently measured
- to provide a balance across the different factors being measured
- certain metrics will relate to a test environment, but some should reflect real-life scenarios as they must be measurable, and reproducible, on your production website
- prioritize the metrics according to their significance to the website
- limit the metrics to a set that can be realistically monitored
During development of the project they can be updated and tuned as appropriate. After the project has been successfully implemented, they can be used to help you control your installation and monitor/maintain the required levels of service for ongoing operation.
When used properly these metrics can provide a useful tool; when used irresponsibly they can be a time-wasting distraction. As always, you need to understand what you are measuring, how you are measuring it and why.
This section will deal with the basic principles and issues to be considered. Each installation is different, so the actual values to be measured will differ.
All metrics to be measured will, in some way, be affected by the design of your project. Conversely, many issues will be best solved by design changes.
Therefore, you should define your target metrics before deciding on your design. This allows you to optimize your design based on these factors. Once your project has been developed, it will be difficult to make any changes to the basic design principles.
When you create the structure for the website, follow the recommended structure for AEM websites. Make sure you understand the following issues and/or principles:
- How to structure website content.
- How templates and components work.
- How caching works.
- The impacts of personalized content.
- How the search function works.
- How you can use CSS and related technologies to create compact, non-redundant HTML code.
If you feel that your design does not follow the guidelines, or if you are unsure about some of the implications, clarify these issues before you start either the programming phase or filling in the content.
To define or assess the infrastructure it will help to define target values such as:
- visitors/day; both average and peak
- hits/day; both average and peak
- number of web-pages being made available
- volume of web-content
Depending on your situation, and the strategic significance of the website, this will help you to assess and choose your infrastructure:
- number of servers
- number of AEM instances (author and publish)
There are several performance factors which can be evaluated:
- response times for individual pages, taking into account:
- response times on an author environment
- response times on the publish environment
- response times for search requests
This section can be read in conjunction with Performance Optimization that expands the technical details of actually measuring the performance.
A key issue is the time your website takes to respond to visitor requests.
Although this value will vary for each request, an average target value can be defined. Once this value is proven to be both achievable and maintainable, it can be used to monitor the performance of the website and indicate the development of potential problems
Differing targets on author and publish environments
The response times you will be aiming for will be different on the author and publish environments, reflecting the target audience:
This environment is used by authors entering, and updating content, so it must:
- cater for a small number of users who generate a high number of requests when updating content pages and the individual elements on those pages
- be as fast as possible to maximize their productivity for getting your content onto your website
So how can you decide on achievable (average) response times? This is often a matter of experience:
- past experience on your website
- experience with AEM
- recognizing complex pages which have above average response times (these should be individually optimized if possible)
However, (under controlled circumstances) the following guidelines can be applied:
- 70% of the requests for pages should respond in less than 100ms.
- 25% of the requests for pages should respond in less than 100ms-300ms.
- 4% of the requests for pages should respond in less than 300ms-500ms.
- 1% of the requests for pages should respond in less than 500ms-1000ms.
- No pages should respond slower than 1 second.
The above numbers assume the following conditions:
- measured on publish (no authoring environment and/or CFC overhead)
- measured on the server (no network overhead)
- not cached (no AEM-output cache, no Dispatcher cache)
- only for complex items with many dependencies (HTML, JS, PDF, ...)
- no other load on the system
Monitoring response times with the AEM request.log
A good starting point for performance analysis is the request log. Amongst other information, you can use this to see the response times of individual requests. See Performance Optimization for more details.
Search requests can have a significant impact on your website, in terms of both the:
- Response time of the actual search
- A fast search function is a quality goal for your website
- Impact on general performance
- As a search function must scan (potentially large) sections of the content, or a specially extracted index, this can impact the performance of the entire system if not optimized
Setting targets for search requests is, again, a matter of experience depending on:
- experience of AEM
- an assessment of how often search will be used in comparison to other goals
- your persistence manager
- your search index
- the complexity of your search function; a basic search function which only allows 1 search term to be input will be quicker than an advanced search allowing the user to build up complex search statements using AND/OR/NOT.
These should be planned and integrated from the very start of your project. Mechanisms available for monitoring include:
Programmed mechanisms for measuring search response times
To customize the information you collect about search requests, and their performance, it is recommended to include information collection in your project source code; see Performance Optimization for more details.
Your website will be made available to a number of users/visitors, on both the author and publish environments. The numbers are often more than you used when testing, but also fluctuating and difficult to predict. Your website will need to be designed for an average number of concurrent users/visitors without noticing a negative performance impact. Again the request.log can be used to make concurrency tests; see Performance Optimization for more details.
Targets for the number of concurrent users, are dependent on the environment type:
- This is more difficult to predict, so you must select a target value. Again this should be based on experience of your current website together with realistic expectations of your new website.
- Special events (e.g. when you publish new, very popular content) may exceed expectations - or even capabilities (as sometimes reported in the press when tickets for certain events are made available for sale).
- The system’s ability to deliver the volume.
- At each step, capacity and volume are measured differently, as shown in the table below. For best performance, make sure that the capacity matches the volume at each step, and that both capacity and volume are shared across all steps. For example, you may be able to compute the navigation on the client computer, or put it in the cache, instead of computing it on the server for every request.
What / Where Capacity Volume Client Computational power of the user’s computer. Complexity of the page layout. Network Network bandwidth. Size of the page (code, images and so on). Dispatcher cache Server memory of the Web server (main memory and hard drive). Web server (main memory and hard drive). Number and size of the cached pages. Output cache Server memory of the AEM server (main memory and hard drive). Number and size of the pages in the output cache, the number of dependencies per page. The dispatcher cache lowers this volume. Web Server Computational power of the Web server. Amount of requests. Caching lowers this volume. Template Computational power of the Web server. Complexity of the templates. Repository Performance of the repository. Number of pages loaded from the repository.
The preceding sections detail the main metrics to be defined.
Depending on your specific requirements it might be useful for you to define additional metrics, either in isolation, or taking the above classifications into account.
However, it is preferable to have a small set of accurate, core metrics that function easily and reliably, rather than try to measure and define every aspect of your website. By its sheer nature, your website will start to change and evolve as soon as it is handed over to your users.
The earlier you plan and design your acceptance tests, the easier the final deployment will be. They should be defined together with the customer and your Quality Assurance team.
Although you might not be able to define all details at the very start of the project, initial definitions should be discussed and agreed on. The acceptance tests will probably be based on fundamental requirements (functional and performance) of the system.
By involving interested parties as soon as possible you can encourage them to become stakeholders in the project, thus increasing their commitment to its success.
- On the customer side this includes the authors - who will have to work with the system on a day to day basis.
- Within your own project team this will also include the people responsible for quality assurance. The more they understand the customer's requirements the better they can plan the tests.
The processes to be defined will depend upon your individual project. Again try to keep these simple, with consideration for:
- Defining processes (and paths of communication) for interacting with any third-parties; e.g. design agencies and third-party software suppliers amongst others.
- Often the customer will have their own Project Management and Reporting procedures and tools.
There are many tools available for tracking information on bugs, tasks, and other aspects of your project - see Overview of Potential Tools for more details.
- The key point to note here is to keep only one copy of the information and share the information (and therefore access to the tool being used). This will ease maintenance and help prevent discrepancies.
Clearly define what is to be covered by the project at various levels:
- the individual releases (if an iterative release process is used, and regardless of whether they are delivered to customers or your internal test team).
- the AEM project.
- the entire project; including any third-party software, their impact on testing, organizational issues and many others.
- For certain aspects it can also be useful to state what is not within the scope of the project. This can help prevent confusion and incorrect assumptions, though it should be limited to essential issues.
This information can be defined within a Project Handbook; the use of a Wiki can also help ensure that ongoing changes are handled efficiently. Wherever these are defined, the key factors are that:
- Information is defined and maintained
- Information is clearly communicated to all of the relevant people involved. Although standard Project Management practice, it cannot be repeated often enough that clear role definition and good communication can make, or break, a project.
- Only one version is kept of any information being tracked; for example, bug tracking, issue tracking, etc.
The following section offers an example overview of milestones and phases involved when first implementing an AEM project. The actual model you select to work with will depend on factors such as the project concept, budget and timelines amongst others.
Any time scales are theoretical.
This overview of standard AEM projects relate to new implementations (or versions). Often such projects also include a design relaunch.
The main phases are:
- Handover from the Sales process.
- Implementation of the customer application (Development).
- Installation and configuration of the infrastructure (and related processes) on customer site (Infrastructure).
- Creation (or migration) of the content (Content).
- Handover to operations (Maintenance/Support).
- Follow up releases.
It is recommended to treat the application, infrastructure and content as separate (sub-)projects.
The diagram below gives more detail about the three main phases, followed by some key points related to each:
Split the project-launch into Soft Launch(s) (reduced availability, multiple iterations) and Hard Launch (full availability - Live) to allow for tuning, optimization and user training under realistic conditions on the production environment.
See Example Tasks for examples of possible tasks which you may need to perform (or assess) during the life-cycle of your project. These tables are meant as quick-reference and to act as a springboard, they do not (cannot) offer a completely comprehensive list of all possible tasks. Your list of actual tasks will vary with each project.
- Define the base architecture first.
- Use several iterations (sprints) for development:
- First sprint equates to the first full development cycle.
- First sprint results in the first deployment to your test environment.
- Every sprint has a runable result.
- Each sprint gets a customer signoff (minimum of structured test with feedback).
- Plan for the eventuality of an update of the available AEM version during the project.
- Plan for tests and optimization during sprints.
- Plan for stabilization and optimization phases.
- Create a log of items to be planned for further releases.
- Plan for partner involvement and handover.
- Define the base architecture first:
- Define performance requirements.
- Define performance goals (ie clearly define expectations).
- Define hardware and infrastructure architecture; including sizing.
- Define deployment.
- Use several iterations; for the first sprint and initial configuration prepare:
- Development environment.
- Development process.
- Test environment.
- Deployment process (including configuration management).
- Plan for several load tests.
- Plan for tests and optimization during sprints.
- Plan for a stabilization and optimization phase.
- Deploy to the production environment as early as possible (let the operations team setup the system to gain experience).
- Use named users and defined roles as early as possible.
- Plan for training (for example, administrator training).
- Plan for handover to operations.
- Define the base architecture first:
- The base architecture:
- Drives the content hierarchy.
- Helps to define the content concept.
- Defines MSM usage and layout.
- Defines roles, groups, workflows and permissions.
- Consider whether offline page creation will be useful.
- Plan for the early creation of first pages and content (for use in tests and feedback).
- Plan for the migration of existing content.
- Plan for "in-sprint-migration" after refactoring.
- Plan "content burndown" (sitemap for go-live content).
- The base architecture:
Dependent on your resulting task list you can then make initial estimates of time and effort for (high-level) task definitions. These should include an indication of who (customer or partner) will do what and when.
The following list shows standard approximations and inter-relationships of effort involved, and therefore costs:
|Development||A rough estimation of 2 - 4 hours for each component node will cover all development requirements.Note: These can only be used for initial estimates. An experienced AEM developer must make the detailed analysis.|
|Developer Testing||15% of Development|
|Follow-up||10% of Development|
|Documentation||15% of Development|
|JavaDoc Documentation||10% of Development|
|Bug-fixing||15% of Development|
|Project Management||20% of project costs for ongoing Project Management and Governance|