Concurrent programming is one of the most demanding disciplines in software engineering — and one of the least understood at the hardware level. This thesis set out to close that gap, benchmarking concurrent counter and HashMap implementations across radically different hardware configurations to understand how low-level decisions propagate upward into high-level application behaviour.
The research reveals that assumptions programmers make about parallelism — that more threads always means more throughput, that lock-free data structures are universally faster — break down under real hardware constraints. Cache line invalidation cycles, NUMA topology, and contended memory buses all extract a hidden tax that no profiler exposes directly.
Using the Java Microbenchmark Harness (JMH), each implementation was subjected to controlled scaling tests: thread counts doubled from 1 to 32 across both shared-memory and NUMA configurations. The results are unambiguous — lock-free approaches scale significantly better under high contention, but introduce their own bottlenecks when thread counts exceed physical core counts.
The practical implication for systems designers: algorithm selection must account for the deployment environment. A design optimal on a developer workstation may degrade 4× under production load on a different machine class.
The IT-Connect internship compressed years of academic learning into months of production reality. Working directly with clients — designing web solutions, configuring NAS and network systems, setting up remote access infrastructure — the stakes were concrete and the feedback loop was immediate.
Web solutions were developed through continuous user involvement: design iterations validated by customer feedback before any code was committed. The result was a development cadence that married user-centred practice with technical delivery rigour.
Infrastructure work spanned NAS configuration for shared file access, network printer setup, PC client deployment, and internet infrastructure management — giving direct exposure to the operational layer that underpins every digital service.
Designed, developed, and published digital web solutions under real client requirements and feedback loops.
NAS systems, network printers, remote access, PC client deployment, and internet infrastructure configuration.
The introduction of service robots into educational institutions is rarely a technology problem — it is an adoption and value-creation problem. This bachelor's project studied the conditions under which service robots transition from novelty demonstration to genuine operational value, applying economic theory within a design thinking framework.
Workshop facilitation with institutional stakeholders generated concrete design decisions grounded in research analysis — demonstrating that the gap between “impressive demo” and “daily utility” is bridged by organisational readiness, not technical capability.
Technology implementation in social psychiatry settings is uniquely dependent on employee buy-in. This research examined how organisational anchoring, change management, and role development determine whether welfare technology delivers its intended benefit — or creates friction that undermines care quality.