Checkpoint interval optimisation for fault-tolerant cloud workloads
“At what checkpoint frequency does the overhead of saving state exceed the cost of re-running the lost work?”
Method: CloudSim simulation, failure-injection experiments
Supervision · student research8 open topicsFUTA Computer Science
Supervision
Final-year and postgraduate projects I am willing to supervise. Each topic below is a real problem with a defensible scope — not a title looking for a student.
01
Open project topics
Each written as a research question with a defensible scope — not a title looking for a student.
Cost-benefit model for deduplication in low-bandwidth environments
“Does chunk-level deduplication still pay for itself when upload bandwidth, not storage, is the binding constraint?”
Method: Trace-driven analysis on real file corpora
Measured availability of Nigerian cloud providers
“Do indigenous providers meet their advertised uptime, measured independently over a full semester?”
Method: Longitudinal probing, statistical reporting
Hybrid risk assessment applied to a Nigerian university network
“What does a hybrid qualitative-quantitative risk model actually score when applied to a real campus network?”
Method: Field assessment, model instantiation
Lightweight authentication for constrained IoT multicast groups
“Can asymmetric-signature multicast authentication run within the energy budget of a microcontroller node?”
Method: Hardware benchmarking
Pattern-cube mining beyond telecoms
“Does the Pattern Cube Algorithm generalise to transaction streams — mobile money, POS, transport ticketing?”
Method: Algorithm adaptation, comparative benchmarking
Churn prediction from call detail records
“How much predictive signal survives when CDRs are aggregated to the privacy level operators are willing to release?”
Method: Feature engineering, supervised learning, ablation
Record linkage across fragmented national identity databases
“How accurately can citizens be matched across agency databases with inconsistent name spellings and no shared key?”
Method: Probabilistic record linkage, evaluation on synthetic registries
02
What is expected of a supervisee
Six non-negotiable principles of working together.
01
Come with a question, not a title
A topic is a question that can be answered wrongly. If your proposal cannot fail, it is not a project.
02
Read before you code
Ten papers read properly beat a hundred skimmed. Bring an annotated bibliography to the second meeting.
03
Instrument everything
If a number appears in your report, the script that produced it must exist and must run again. Keep a metrics file under version control.
04
Negative results are results
If your proposed method loses to the baseline, report it, explain it, and defend it. That is a stronger project than a suspiciously perfect table.
05
Weekly written updates
One page: what you did, what broke, what you will do next. Meetings are for the parts writing cannot resolve.
06
Integrity is not negotiable
Fabricated data, ghost-written text and undisclosed tool use end the supervision relationship. There is no second conversation.
03
Frequently asked questions
Answered here so we do not spend the first meeting on logistics.
How do I get you as a supervisor?
Send a one-page note: the topic that interests you, why, and what you have already read. Allocation still follows departmental process, but a prepared note tells me who to argue for.
Can I propose my own topic?
Yes, and I prefer it. It must state a question, a method, and how you will know whether you succeeded.
How often do we meet?
Fortnightly by default, weekly in the final six weeks, plus written updates in between.
What format for the final report?
Departmental template, IEEE-style citations, every figure numbered and referenced in text, and a reproducibility appendix listing seeds, versions and runtimes.
Can I use AI tools?
For learning, debugging and editing — yes, and say so. For generating results, analysis or prose you then sign as your own — no.