As I mentioned in An IT services overview I try to keep track of the architecture and designs of the IT services and solutions in a way that I feel helps me keep in touch with all the various services and solutions out there. Similar to how system administrators try to find a balance while working on documentation (which is often considered a chore) and using a structure that is sufficiently simple and standard for the organization to benefit from, architects should try to keep track of architecturally relevant information as well.
So in this post, I'm going to explain a bit more on how I approach documenting service and solution insights for architectural relevance.
Why I tend to document some of it myself
Within the company I currently work for, not all architecture and designs are handled by a central repository, but that doesn't mean there is no architecture and design available. They are more commonly handled through separate documents, online project sites and the like. If we had a common way of documenting everything in the same tool using the same processes and the same taxonomy, it wouldn't make sense to document things myself... unless even then I would find that I am missing some information.
It all started when I tried to keep track of past decisions for a service or solution. Decisions on architecture boards, on risk forums, on department steering committees and what not. Historical insights I call it, and it often provides a good sense of why a solution or service came up, what the challenges were, which principles were used, etc.
Once I started tracking the historical decisions and topics, I quickly moved on to a common structure: an entry page with the most common information about the service or solution, and then subpages for the following categories:
- Administration (processes, authorizations, procedures)
- Auditing and logging
- Configuration management
- Cost management
- Cryptography and privacy
- Data management (data handling, definitions, governance, lineage, backup/restore, ...)
- Design and development (incl. naming convention)
- High availability and disaster recovery
- Operations (actor groups, source systems & interactions/external interfacing)
- Organization (management, organisational structure within the company, etc.)
- Performance management
- Quality assurance & reporting
- Roadmap and restrictions (incl. lifecycle)
- Risks and technical debt
- Runtime information
Now, I won't go in depth about all the different categories listed. Perhaps some areas warrant closer scrutiny in later posts, but for now the names of the categories should be sufficiently self-explanatory.
If there is an internal service (website or similar) that covers the details properly, then I will of course not duplicate that information. Instead, I will add a link to that resource and perhaps just document how to interpret the information on that resource.
The entry page
The entry page of a service or solution always looks the same. It starts off with a few attributes associated with the service:
- The taxonomy used within the company
- The main point of contact(s)
- The backlog where the responsible team tracks its progress and evolution
- A link to the main documentation resources
- The internal working name
The taxonomy is something I strongly appreciate in my current company. It is a uniform identifier associated with the product, service or solution, and is used for several of the operational processes the company has. This taxonomy comes up in things like chargeback, service level agreements, responsibility overviews, data classifications, enterprise architectural overviews, etc.
For instance, a managed macbook (asset) might have a taxonomy identifier of
mmac, or we have a service for exchanging internal company data identified as
cd70 (it doesn't need to have an identifier that "reads" properly). Of course,
people don't just run around shouting the identifiers, but when we go through
the information available at the company, this identifier is often the primary
key so to speak to find the information.
For the main points of contacts, I usually just document the person that is my go-to person to get some quick insights. The full list of all contacts (like product owner, product manager, system architect, business analyst, release manager, etc.) is managed in a central tool (and uses the taxonomy to quickly find the right person), so I just have the few names that I quickly need listed here.
The backlog is something I recently added to support any questions on "when will we have feature XYZ". In the past, I had to contact the people to get this information, but that often became cumbersome, especially when the team uses a proper tool for tracking the work on the service.
The main documentation is often the most important part. It is a link to the documentation that the team provides for end users, architects or other roles. Some teams still have their information on a NAS, others in a document library on SharePoint, others use a wiki, and there are teams that use a structured document management system. Working for a big company has its consequences...
Finally, the internal working name is the name that a service or solution receives the most. For infrastructure services, this is often the name of the product from the time the product entered the organization. While the vendor has switched the name of the product numerous times since, the old name sticks. For instance, while I will document IBM's cloud offering as "IBMCloud" (its current name) I will list its working name as "Bluemix" because that's how the company internally often refers to it.
After the basic attributes, I have
- a few paragraphs for description
- a diagram or architecture view to give a high level design
- the most important questions surrounding the service or solution
- some tips and titbits for myself
The high level design is often a view that I maintain myself, which uses the abstraction that I mentioned earlier in my blog. It is not covering everything, but to a sufficient level that I can quickly understand the context of the service or solution and how it is generally implemented.
The most important questions are mostly a refresher for questions that pop up during discussions. For instance, for an API gateway, common questions might be "What are the security controls that it enforces" or "Does the API GW perform schema validation on JSON structures?".
The history of a service
Below the entry page, the various categories come up. As I mentioned, it all started with the historical insights on the service. By having a chronological overview of all decisions and related material per service, I can quickly find the most important information in many cases.
Want to know the latest architecture for a service? Let's look in the history when the last architectural review was, and at which decision body or board it came up. Once found, I just need to go to the meeting minutes or case details to find it.
Want to know why the decision was taken to allow a non-standard integration? Take a look at the history where this decision was taken, and consult its meeting minutes.
Need to ask for a continuance for something but you're just new in the team and don't know why or how it was approved in the past? No worries, I'll share with you the history of the service, where you can find the information, and coach you a bit through our organization.
Having the history for services and solutions available has been a massive efficiency gain for myself and my immediate stakeholders. Of course, I would have loved if the organization tracked this themselves, but as long as they don't (especially since organization changes more often than technology) I will put time and effort to track it myself (at least for the services and solutions that I am interested in).
The historical information I track myself is not a copy/paste of the meeting minutes of those entries. I try to use a single paragraph explaination, like so:
ARB 2020/12/05 "Switch of PostgreSQL authentication provider to PAM+sssd" Approval to switch the authentication provider of the PostgreSQL database assets from the internal authentication to a PAM-supported method, and use the RHEL sssd solution as a facilitator. Link with Active Directory.
ARB is the name of the decision body (here it could be the Architecture
Review Board) and tells me where I can find more details if needed. I don't
really bother with trying to add actual links, because that takes time and often
the links become invalid after we switch from one solution to another.
Since then, I also started adding information related to the service that isn't just decision body oriented:
Incident 2021/06/08 "Fastly major outage" A major outage occurred on Fastly, a widely used cloud edge provider, between 09:47 UTC and 12:35 UTC. This impacted service ABC and DEF.
Incidents can be internal or external, and if they are internal I'll document the incident and root cause analysis numbers associated with the incident as well.
It also doesn't need to be about problems. It can be announcements from the vendor as well, as long as the announcement is or can be impactful for my work.
How complete is this overview
My overview is far, far, far from complete. It is also not my intention to make it a complete overview, but instead use it as a quick reference when needed. Services that are commonly discussed (for instance because they have wide implications on other domain's architectures) are documented more in depth than services that are barely influential to my meetings and projects. And that doesn't mean that the services themselves are not important.
Furthermore, the only areas that I truly want to have up-to-date, is the entry page and the history. For all the other information I always hope to be able to link to existing documentation that is kept up-to-date by the responsible teams.
But in case the information isn't available, using the same structure for noting down what insights that I gather helps out tremendously.
I also don't want my overview to become a critical asset. It is protected from prying eyes (as there is occasionally confidential information inside) and I am coaching the organization to take up a lot of the information gathering and documentation in a structured way. For instance, for managing EOL information, we are publishing this in a standard way for all internal stakeholders to see (and report on). The roadmap and strategy for the services within the domain are now being standardized within the backlog tool as well, so that everybody can clearly document when they expect to work on something, when certain investments are needed, etc.
In the past, architects often had to scramble all that information together (hence one of my categories on Roadmap) whereas we can now use the backlog tools of the teams themselves to report on it.
Which tool to use?
Personally, I use a wiki-alike service for this, so that I can search through the pages, move and merge information, use tagging and labels and what not. I also think that, unless the company already has a central tool for this, a well maintained wiki with good practices and agreements on how to use it would do wonders.
I've been playing around in my spare time with several wiki technologies. Dokuwiki is still one of my favorites due to its simplicity, whereas MediaWiki is one of my go-to's for when the organization really wants to pursue a scalable and flexible organization-wide wiki. However, considering that I try to structure the information in a hierarchical way, I am planning to play around with BookStack a bit more.
But while having a good tool is important, it isn't the critical part of documenting information. Good documentation, in my opinion, comes from a good structure and a coherent way of working. If you do it yourself, then of course it is coherent, but it takes time and effort to maintain it. If you collaborate on it, you have to make sure everybody follows the same practices and agreements - so don't make them too complex.