Direct work logs from the software build process—candid debugging sessions, thesis revisions, and architectural notes.
First day of practicum. Had a meeting with other assistants afterwards. The lecturers have been calling for a progress report regarding the 5g research tomorrow morning. They actually meet in the lab this evening, but I just got home and got a headache. Hope everything goes well tomorrow.
Went out to get a cup of hot coffee with a friend till almost midnight, talking about old friends. Can't sleep afterward. Finished my thesis proposal. Waiting for approval from both lecturers, possibly do a consultation next Friday.
Finally got a clear view of my research position, some inspiration from papers on scheduling (and rescheduling) methods, and guidance on how to get metrics from the orchestrator.
Spent the morning consulting about my PKL report. Finally got a clear idea of what we are actually doing here. Still juggling with papers. It's so hard to find the right one that matches your expected method to solve the problem.
I thought there was still something worth saving between us.
Then you made it clear how little my goodwill was worth.
Rather than hold a grudge, I gave you every chance to prove me wrong.
I should have known better than to mistake your indifference for circumstance.
Even after everything, I kept my side of the door open.
Done asking for the respect I thought we owed each other.
My portfolio website got DDOS-ed a few days back and ran out of computing resources in the free tier of Vercel. My deployment got paused, and I had to deploy my portfolio website locally. Since I don't want to repeat the same mistakes, I need to be able to observe the traffic flowing through my website, and unfortunately, Cloudflare didn't give me the full picture of the logs. So I had to build my own monitoring system using Caddy as the proxy server and connected through Loki, Grafana, and Alloy. Right now it's working just fine to see traffic and logs. Might integrate a firewall sometimes in the near future for ease of use in blocking malicious traffic.
Updated the placement controller of the 5g network cluster to use different Kubernetes Services for each node. A single Service for all nodes should be sufficient for the placement per se, but I'm not sure about offloading. Since I'm too lazy to argue over the architectural design, I just go along with what she wants to say and hope it's for the best.
Implemented a centralized controller for deploying services across multiple nodes/clusters in a Kubernetes environment. Since the image used in each pod is a local image, I also created a small program to distribute the images among all nodes. The current system is working just fine for creating, scaling, and deleting existing pods. Haven't tested it using a real tunnel tho. Spent the whole day at home.
Tried to use the uesimtun directly as the default route to the internet, with a small caveat. I had to make an exception for the local subnet to keep using the wlp1s0 interface. Otherwise, the UE won't be able to connect to the gNB, and the whole thing will just loop itself. It works just fine for lightweight data transfer. The UE lost signal to the gNB when I tried to download a 6 GB ISO image. Possibly due to the simulation limitation in ueransim. Found an open issue since 2021.
Straighten up my research focus on the heuristic approach. Might consider greedy best fit based on computing resources with SLO constraints. Consulted my intern report with my lecturer. I have to revise the first chapter. Still confused about where to bring the focus of the report.
Had a meeting with a lecturer regarding the 5g research. I dismissed the AI idea as it will make the implementation more complex, especially when reinforcement learning (i.e., Q-learning) is involved. I have never had my hand touch that field before, and I'd better not underestimate it, not under a short time constraint. We have approximately 4 months left if we want to graduate before the year ends. I need to do the literature review again, focusing more on heuristic or ILP approaches for dynamic and online placement. Also need to refine the presentation, as it was still written under the supervised learning assumption.
Finish the CCNA Introduction to Network final exam. It was easy. Far easier than the practice lab.