Engineering & Architecture: Build Decisions This Week
2026‑11‑13
The past week has highlighted a mix of external pressures that are reshaping how we think about platform design, team composition, and the security posture of our services. In South Africa, a sharp decline in IT recruitment coupled with sudden water‑restriction alerts underscores the need for resilient, cost‑effective infrastructure. Across the Atlantic, the UK housing market’s slide into a 16‑year low is a reminder that financial headwinds will filter through to the budgetary constraints of the companies we serve. These developments force us to re‑examine platform choices, architecture patterns, tooling, scaling tactics, and security safeguards.
---
The 30‑day Level 2 water restrictions announced by Tshwane mean that power grids and cooling systems for data centers are now a higher risk than ever. A single‑region deployment could see a cascade of outages that would interrupt service delivery. The trade‑off is clear: distributed multi‑region or multi‑cloud deployments add operational complexity and cost, but they provide the necessary redundancy to survive localized utility disruptions. For South African teams, a hybrid model that keeps the core of the application in a region with stable supply while off‑loading read‑heavy workloads or data‑processing jobs to a secondary region or a partner cloud can strike a balance between cost and resilience.
In the UK and EU markets, we must also factor in the GDPR, AI Act, and POPIA compliance layers. Moving workloads closer to the data subject can reduce egress costs and simplify audit trails, but it can also increase data residency obligations. The decision to adopt a globally distributed architecture should be driven by a risk‑budget that weighs the cost of extra regions against the potential loss of uptime.
---
Public‑sector job losses reported by Moneyweb show a tightening of budgets across many state‑run entities. When budget windows shrink, simpler architectures that require less operational overhead become attractive. A micro‑service stack offers isolation, independent scaling, and the ability to adopt different technology stacks per service, but it introduces inter‑service latency, operational complexity, and a higher surface area for supply‑chain attacks.
The Pragmatic Engineer interview on software correctness and supply‑chain security reminds us that every dependency is a potential attack vector. Serverless functions, with their single‑function packaging, reduce the number of moving parts and can simplify dependency management. However, they can be harder to debug in production and may suffer from cold‑start latency that affects user experience.
The trade‑off, therefore, is a “micro‑services‑within‑a‑service” approach: group related responsibilities into a bounded context, expose a small set of well‑defined APIs, and keep the deployment footprint minimal. This allows us to reap the benefits of isolation while keeping operational overhead manageable.
---
MyBroadband’s Employment Insights report reveals an 18 % year‑on‑year decline in IT recruitment in South Africa, largely driven by AI’s ability to handle routine programming tasks. The immediate consequence is a higher demand for seasoned developers, but also a shrinking pipeline of junior talent. The cost of hiring experienced staff is steep, and the talent pool is becoming increasingly competitive.
To address this, we must double down on AI‑assisted tooling that augments developer productivity without replacing the learning curve. A few actionable tactics:
The trade‑off here is between the cost of maintaining a robust tooling stack and the risk of code rot and security vulnerabilities. Given the rising supply‑chain risk profile, the investment is justified.
---
South Africa’s water restrictions mean that data centers may face unexpected power and cooling constraints, while the UK market’s financial strain could limit the ability to scale elastically. One solution is to adopt edge‑centric architectures—for example, leveraging Cloudflare Workers or AWS Lambda@Edge—to bring compute closer to users. The benefits are lower latency and reduced data transfer costs; the downsides are limited compute resources and vendor lock‑in.
An alternative is to use managed database services (e.g., Amazon Aurora, Google Cloud SQL) that provide automated failover and scaling, thereby reducing the operational burden. The cost of a managed service is higher per‑GB than a self‑hosted PostgreSQL cluster, but it also reduces the risk of data loss and ensures compliance with POPIA’s data protection obligations.
---
The Pragmatic Engineer article underscores that software correctness is not a luxury but a necessity in an age where AI can introduce subtle bugs. The recommendation is to adopt deterministic build pipelines, lock dependencies to specific commit hashes, and use reproducible builds. Additionally, integrating SLSA (Supply‑Chain Levels for Software Artifacts) into the CI process can provide an audit trail that satisfies both UK GDPR and EU AI Act requirements.
---
---
The recommendations above assume the availability of managed services in the South African region and the ability to legally host data across borders without violating POPIA or GDPR. The proposed hybrid deployment model also assumes that the vendor ecosystems (AWS, Azure, GCP) provide comparable service levels in both regions. Please validate the feasibility of these multi‑region strategies with legal counsel and the infrastructure team. Additionally, the cost‑benefit analysis of edge computing versus managed databases should be quantified against actual bandwidth and latency metrics for our target user base.