Prefixing everything with the word "digital" seems to make the concept sexier these days. The relatively new term "digital infrastructure" pops up recently in sales pitches from traditionally on-premise infrastructure service vendors. This article is an attempt to explain some of the more prominent elements of the indeed valuable concept of digital infrastructure.
As per my description above, the benefits of digital infrastructure together with some of the key techniques associated are:
- Costs Efficiency: For example, being able to run multi-cloud allows for selection of more economical cloud vendors. Cost effectiveness benefits from hybrid-cloud depends on scale. Roughly speaking, if the physical infrastructure scale is too small, then the lower TCO of running application workload on-premise might not apply because of the overhead of requiring system administrator/engineers to take care of the hardware. (Hint: For any serious operation, at least 3 system administrator/engineers will be required on rotation and as backup to support the operation). But if the scale of the infrastructure is sizeable with known, sustained and fixed workload, then on-premise infrastructure will be more cost-effective considering TCO of say 5 year down the road. CAS(Infrastructure As Code) or SDI (Software Defined Infrastructure), on the other hand, are cost effective from the get go, because there will be quite a number of deployments during SDLC and operation in the lifetime of a serious application.
- Risks Management: There are quite a number of risks if the infrastructure is locked in. For example, if there is a catastrophic failure of a particular cloud vendor, then workload migration will take significant efforts if everything is tailored for the one cloud vendor. Cloud native is cool and quick, but not really agile after the initial exploration phase. In these days of contention amongst nations, being cloud-agnostic or multi-cloud ready will help ensure business continuity. IAS/SDI will reduce risks because everything about a system/application is "documented" by the deployment code (assuming the software deployment of the system/application are scripted as part of devops as well). Even if the administrator god is sick out, or has won a lottery, the organization will still have a good handle on their systems.
- DevOps Agility: DevOps encompass more these days. We often hear about DevSecOps, DataOps, AIOps, etc. So, let's just be silly and use DevSecDataAIOps once and for all (for now). The agility of having the infrastructure side software defined together with software deployment automation can be understood from the fact that moving to different clouds to different regions can be done with just minor tweaking of the deployment script (assuming done well with abstraction using tools like Terraform, Ansible, etc). For unit testing, integration testing, and trying out of different clouds or on-premise can be done with just one click, speeding up R&D and solution evaluation. To those who might wonder why the infrastructure layer matter so much, please note that traditional VMware for the application server, database server, and NAS being sufficient to run most applications days are a bit of the past. With the advance of technologies, e.g. cache, CDN, NoSQL, streaming clusters, distributed storages, containerization, machine learning pipeline, and so forth, an application performance and even correct operation are coupled much more closely with the infrastructure, which if not configured, versioned, and maintained well will manifest in application issues. So, it is imperative to have the infrastructure configuration and deployment scripted as much as possible.
If it is not obvious from the above already, DevSecDataAIOps automation including IAS/SCI will be the foundation of digital infrastructure. One often neglected detail is test automation. Whether it is application regression, security penetration/code-scan, data hygiene, or even ML related testing, shall be included as part of the automation. The efforts to create the test will only be high if done as an aftermath. But just think of the human efforts savings for each deployment, and the costs savings from bugs prevented, should make it clear the necessity of test automation. Of course, if it is just a one-time/one-off one season only type of promotional application, then automation might not pay off.
A side note is that most organizations failing Agile because of the neglect of test automation. This is particularly apparent during frequent deployments, which require testing to avoid introducing more bugs than features. Agile can be fragile because the are nuts and bolts that cannot be missed.
There are a lot more important concepts and details on digital infrastructure. This article only picks on a few topics which the author deems more interesting to most.
Just to clarify, for most organizations which are still on-premise, going cloud first is a great transition step, because in the process of migrating to cloud, systems and infrastructures will be more standardized and observable. When priority allows, going multi-cloud and hybrid-cloud after being cloud-only will not be difficult if DevSecDataAIOps are automated. Just be mindful and control cloud-native dependencies.