In a perfect world, using infrastructure or technology services would be seamless, without impact, without risks. It would auto-update, tailor to the user needs, detect when new features are necessary, adapt, etc. But while this is undoubtedly what vendors are saying their product delivers, the truth is way, waaaay different.
Managing infrastructure services implies that the company or organization needs to organize itself to deal with all aspects of supporting a service. What are these aspects? Well, let's go through those that are top-of-mind for me...
When you have a service running, then you need to ensure operational support is in place in case there are issues. Be it to resolve a malfunction, a security issue, or a performance degradation - you need (human) resources to ensure that the service remains running adequately.
In many organizations, this is handled in a three- or sometimes even four-level support structure:
- First line is generally a call center that procedurally validates if an issue is service-bound, or if they can assist the user to correctly or better use the service. First line often does not require any knowledge of the customer base nor target infrastructure, and is strongly procedure-oriented. They do not have operational duties on the services themselves, and are an important part to weed out unstructured or invalid service requests. They then escalate the issues to second line.
- Second line is an organization that has knowledge on the customer base and the services themselves. They are also often the last line that has a wide view of all services within the company, as subsequent support levels are more specialized. Second line has the ability to take actions on the services themselves (like restarting services) if this is agreed upon in the past with the main stakeholders, and when this is executed in a controlled manner. If second line isn't able to resolve an issue, it moves to third line.
- Third line support is generally the team that is operationally involved in the service itself. If the problem lays with a customer portal for instance, then third line support is often the team that maintains the customer portal. They know the service and its usage in detail, and are in many organizations the last line of support.
Some organizations even continue this layering, with a fourth line of support being a technical expert in a particular component used by the service. If the customer portal has problems, and it is within the database, and the DBAs of the customer portal team do not have the knowledge to resolve it, they might escalate further to a dedicated DBA team (which is not assigned to any particular business service but specifically organized to be experts in database and database administration).
Not all companies have a technology-oriented support team though, and many companies will consider this as part of 3rd line support as well, if not just to be more aligned with market terminology on support structures. Still, organizing and optimizing this third line of support is often something that infrastructure service support is heavily involved in.
Regardless of the structure approached by the organization, these teams will need the knowledge and supporting tools and procedures to do their job. You need people that can develop support procedures, create simplified automation (for second line to execute), and continuously update that information. And if a vendor is involved, then the support line will need to have knowledge on how to approach the vendor: what are the procedures and processes for raising incidents, what is the priority queue like? Does the vendor have certain SLAs that the team should know about?
The several layers of support will need continuous training, even if it is just refreshing past information. It is also wise to involve these support lines in information sharing, like when you know there is a growth in database reliance in the business services, or when you know many databases are being migrated from one technology to another. Second line for instance might be able to use that information, together with their cross-organizational knowledge, to better triage issues.
Finally, this support structure is often tied to service level agreements that the company makes either internally, or with its customers. Hence, the support must be guaranteed within those SLAs. That implies that the support must be driven through multiple people, as the organization has to provide support during off-hours, during holidays, during absence... heck, even during global pandemics. I use the "3 FTE per technology" principle: you need three people with enough knowledge to support the technology.
As mentioned in the support part, teams exist that are responsible for the service itself, including patching and updates. This is part of the maintenance requirement on services, and infrastructure services are not different. Perhaps even more so than business services, infrastructure services have a wider impact if they are hit with a bug or with performance degradation, as many business services rely on the infrastructure to be up and running, highly available, well-performing, and secure.
Maintenance tasks for services include, alongside the participation in the operational support:
- Executing security- and stability patch validation and roll-out (updates)
- Proactively assessing the state of the service to see if improvements or mitigations are needed, before these result in actual issues
- Ensuring sufficient capacity is available for now and in the immediate future
- Resolving performance issues, be it by increasing resources, moving services to different locations or hosts, fine-tuning configuration parameters, offloading workload to different services, etc.
- Executing planned maintenance activities, which can include upgrades to higher versions. When the service cannot be transparently upgraded, this will involve a thorough alignment with all stakeholders as part of the change management processes.
- In case of larger, wide-spanning incidents (or even disasters), the teams play an active role in the orchestrated recovery together with all other teams.
Designing and architecting the integrations
To ensure that the services are well supported and can be maintained in an efficient manner, there is often a strong focus on the proper design of the infrastructure services and their role in the architecture. This design does not just include pointing out which components exist where, but also how the service integrates within the larger landscape:
- How are administration tasks executed? How do administrators authenticate themselves?
- Where are the monitoring metrics found and stored? Do you use trend analysis, and in which system?
- How are the assets related to the service tracked? What is the lifecycle for the individual assets, and how does that impact the maintainability of the service?
- What is the best way to use the service, and what will not be allowed?
- How are backups taken, and what does a restore procedure look like? Do you support both in-place restores, side restores, etc.
- What underlying resource requirements exist? Do you require certain tiered storage, or network performance requirements? If you allow your service to be instantiated by others (rather than maintain a platform yourself), what are the target resources that you allow? Are there minimal resource requirements?
- How do users interact with the service? Do they access it directly, or do you require intermediate gateways (like reverse proxies)?
Now, design and architecture go beyond integrations. I focus strongly on integrations here as it is a part of design and architecture that has strong dependencies and relations with other teams and technologies. To work out the integration side of a service, you can't do this autonomously without assistance from the other service teams. Of course, if those teams have everything well documented and architected, and easy to consume, then the team responsible for the design and architecture can do a large part of the work independently, but having a final validation or confirmation is always beneficial.
One of the values of a proper integration design is a higher quality of the service delivery. If integrations are missed, you'll find that a service might not be ready to be activated in production, and then suddenly there is a stronger focus on timely delivery than on quality. Situations where firewall rules need to be quickly pushed and opened up because a project failed to assess their integrations, resulting in security risks, is sadly enough all too common.
Larger organizations will often have architects and designers within the teams or directorates to support this endeavor.
Secure setup and deployment
Given that some integrations might result in heightened risks, and infrastructure services are often widely used or consumed, the organization will need to make sure that the services are designed to be secure.
Security of a service is more than just ensuring it is up-to-date. You will also need to make sure it is configured correctly (as misconfigurations are a frequent occurrence of security incidents), that the authentication and authorizations are properly designed (and where needed or possible, using multi-factor authentication), that the deployment considers the placement and interactions (firewalls, zoning, etc.), that the service provides functional (or perhaps even physical) segregation, that the data governance is appropriate and aligned with regulatory and company requirements, that the service is continuously validated by the security tooling available (patch indications, vulnerability management, ...), etc.
As services also evolve when they are alive, secure setup and deployment is not a one-off (but the initial thoughts and designs are not to be underestimated): the teams will need to assess the impact of new insights (like security notifications, vulnerabilities, global changes by the organization, new threats in the wider IT world) which implies that the team has a continuous security and risk focus.
In many cases, there is even a coding and development component with a service. Not many services can be installed and deployed without requiring any integration scripting or even in-depth coding (such as custom plugins). And when there is coding, there is the need for secure coding: following a Secure Development LifeCycle (SDLC) approach to get assurance about the secure state of the developed code.
When the deployment uses infrastructure-as-code methods, follows a GitOps approach, or similar, then there should also be sufficient attention to the secure setup of these pipelines and the platforms on which they run. The code (or configurations) hosted should also follow appropriate security guidelines
Robust and reliable services
Designing for a trustworthy, secure environment is one thing. The other major focus area for infrastructure services is the availability, robustness, and resilience of the service. While not all services require to be up and running 24/7, nowadays it is hard to imagine many services to still have significant downtimes.
So, you often need to architecture and design for a resilient service, which can take a beating if it has to. Organize load- and stress testing. Organize disruptive testing if needed, and learn from the results to build a better service. Design for a service that you can do technical maintenance on without disrupting the customers themselves as much as possible - but don't overshoot.
If the service is set up in multiple locations, make sure that there is independence between these locations (often across different regions) so that failures in one region do not affect the other regions. Consider a setup as used in many public cloud environments: high availability across availability zones in the same region, and regional independence while allowing for cross-region usage scenarios for the customers.
Assess what can go wrong, and either try to update the architectures and designs to be more resilient against these failures, or establish procedures and processes to quickly recover. A common focus area here is to recover from disasters (using so-called Disaster Recovery Procedures), and there are plenty of disasters to assess: data center failures, large Internet outages, ransomware or other cyberattacks, worldwide epidemic outbreaks, etc.
Quality assurance at all stages
The organization has to be able to provide fixes for defects and problems that were raised, or to preemptively fix issues before they reach the customers. That doesn't mean that the organization needs to be able to develop the fixes itself: especially with off-the-shelf products the development is done by the independent software vendor (ISV). However, the organization does have the responsibility to track and put their weight on this so that the issues are indeed properly resolved (and in case of a third-party product, preferably through a fix that is applied to the main product, and not a one-off for that particular company). Of course, if the service is developed in-house, then the development of the fixes has to be guaranteed by the organization as well.
To be able to provide secure and reliable services, it is vital to have good change management processes and tools in place so that you can approach the various stages of quality assurance before reaching production. In The pleasures of having DTAP I mention the benefits of having four environments for the various stages of a development lifecycle (development, testing, acceptance, and production) and that is perfectly applicable to infrastructure services as well, even when the environments for infrastructure services might be isolated from those of the more business-oriented development stages: you want to make sure that the business-oriented development has production-grade services for its processes, and not the intermediate and possibly less reliable in-development infrastructure services.
Throughout these environments, testing can (should) be introduced to provide guidance to the engineers and developers to improve on the service. Testing can take several forms: unit testing within the frameworks, integration testing of code in larger environments, integration testing of new products in those environments, sanity testing with simple use cases to ensure nothing blows up (figuratively... hopefully), regression tests to ensure no fixed defects creep back in, acceptance tests for specific features or use cases, load testing to ensure the product stays up under the expected loads (and the individual components or services have the right performance profile), stress testing to ensure the product is resilient against higher loads or bursts, destructive tests to see the product behaves as expected when unauthorized usage is performed, security testing and penetration testing to provide assurance on the secure state of the product deployment, etc.
In many cases, the QA part is driven by organizational requirements, and not all products require the same intensity on the tests and assurance levels. But, by introducing the right automation, it is far easier to include multiple test scenarios and test types in your pipeline.
Products then pass through the various stages according to the change management principles and processes that are in effect in the organization.
Strategy and roadmap
The team responsible for an infrastructure service will also need to consider the service in the long term: is the current technology (or set of products) still state-of-the-art, mature, and following market practices? Or does the team consider the technology to be relatively outdated and in need of an update? When would the right time be to address this update?
Perhaps the currently used technology is nearing its end-of-sale, end-of-support (EOS), or even end-of-life (EOL). In that case, the team has to be ready to address these lifecycle stages accordingly, be it through migrations, or refactoring of current usages. Perhaps the teams find that it is more sensible to get an extended support contract in place, or that they have the ability to take the support (including code development) completely internally. Whatever the choice is, it has to be made clear for the wider organization, and the support team has to be ready to take on its commitments.
There is a distinction between the service (or better yet, "capability") that is offered, versus the products and technologies that support and realize it. Capabilities will be needed by the organization for a long, long time, whereas the products that realize these capabilities can have much shorter timeframes.
The team will need to consider alternatives for the products, and see when those alternatives make sense to address and implement. Perhaps the organization might benefit from multiple implementations using different technologies because they serve different usage scenarios, and the costs of keeping multiple technologies in the air is less than the value it provides. Or perhaps an alternative needs to be quickly realizable in case of a sudden exit scenario of the current technology (or vendor). In that case, while the alternative doesn't need to be up and running, the team must be able to address the change in due time.
The need to have a proper roadmap on the capability and products that are being used also reflects in the relationship that that team has with the vendor. For strategically important products, an organization might even want to participate in that vendor's Customer Advisory Board (CAB) or equivalent programs. The team should have the time and resources to collaborate with that vendor to build a partnership, participate in the conferences and other events, as that provides input to the team on the progress and future of that product. Those insights are primordial to properly design and organize an internal roadmap.
Once the internal roadmap has taken shape, then the responsible teams are involved in supporting the organization through updates and upgrades, by communicating the need for these upfront (so that the internal customers can plan around it), and to track the progress so there are no lingering risks for the wide organization.
More elaborate (internal) customer support
The support for the infrastructure service often has to extend beyond the operational support that I started with in this article. Teams that provide certain capabilities within the organization don't do this as a vendor, but as part of that organization. So when an internal customer needs help in migrating away from a service, the team is still involved in this process.
Hence, more elaborate support such as migrations (both towards a new technology, service or capability, as well as migration away from it), finding suitable alternatives that are not properly handled by the capability (and thus might be best provided by a different team), ... helps in building out the trust that the wider organization has in IT.
Cost, licenses, and contractual obligations
Many services have certain contractual obligations associated with them, often known as the Terms and Conditions of the contract and product usage. Infrastructure teams will need to make sure that these T&Cs are known and that the organization adheres to them.
The cost of an infrastructure service usage also needs to be correctly devised and accounted for. Teams have to make sure the product usage remains within the allocated licenses (or, if there is no capping in place, that the usage is sufficiently constrained that the organization does not get any surprises), and is often involved in defining a chargeback towards the rest of the organization.
I tend to make a distinction between showback (show the organization how much a service costs to the company or usage group) and chargeback (a governance-driven or tax-driven requirement for charging usage to the organization), as the latter is more a company decision on how to approach this, whereas the showback is the actual, factual cost. Showback is needed to support conscious decisions on next steps or consumption patterns, whereas chargeback might be necessary for tax reasons in larger corporations where IT is considered as part of a different legal entity.
Addressing cost, licenses, and T&Cs is not to be underestimated. Many vendors make this very difficult, as that allows for many interpretations during license audits that can give a nice bonus to the vendor if he can show that his interpretation is more appropriate than how you thought that the contract or license was structured.
The cost of a product or technology is not just the cost of the purchase, and even that cost is still somewhat variable: many things are open for negotiation, and you also don't need to tackle this directly with the vendor, as there is a large market of middle parties that facilitate purchasing technologies. These can often provide better rates as they can bundle purchases of multiple customers and thus negotiate better deals with the main vendor.
The cost also depends on the support contract associated with it, as well as the costs of other depending technologies (such as capacity requirements) that come from its implementation. This is often neglected in SaaS purchases: even though you have correctly negotiated a good price for the SaaS service, you might be jeopardizing your internet connectivity and need to upgrade the bandwidth, the anti-DDoS service, your firewall capabilities, etc. because the consumption of that SaaS service is significant.
The responsibilities for managing and tracking infrastructure services are large and not to be underestimated. It is not a matter of deploying a new service and assuming everybody can deal with it, nor are all responsibilities equally visible to the end-user.