hhai
608 posts
Sebagai yang mengikuti perpolitikan Vietnam termasuk soal HSR nya Hamba hanya bisa HEHEHEHE saja.😌
Support for "older" computers is critical for the future of computing. Linux is dropping 486 support... and that is a massive mistake.
@nxt888 Ho Chi Minh City just feels like a US ghost to me. That it's capitalist capital named Ho Chi Minh City seems ironically odd, it couldn't be more far removed from the man himself. Like walking through a US movie set of one of it's own cities where all the actors are Vietnamese.
“Use Redis as a global lock. Done.” This line has been responsible for races that only reproduce once a month under peak load, right when no logs are at the right level… The magic keyword is Lease, not Lock. In a distributed world, you must assume: processes crash clocks drift network partitions happen A “lock” you think you hold may already be considered expired by someone else. A trap: Service acquires a Redis lock to become “the leader” doing some periodic job. GC pause or network blip happens. Lock expires; another instance acquires it. Now two leaders run the same “single” job at once. A seasoned engineer designs for clock skew and failure modes: Locks have short TTLs. The holder frequently renews them. Critical operations are designed to be idempotent or detect concurrent ownership (version checks, fencing tokens). So, the real design question isn’t: “How do I make sure only one instance runs this code?” It’s: “If two instances run this code at the same time anyway, what’s the blast radius?” If the answer is “disaster,” you need more than a naive distributed lock.
Prediction: 2026 is going to be the year Memory Tiering eats the world. The pieces are already there: - Memory Shortage (duh) - CXL 3.1 realistically purchasable (memory pooling) - Linux Transparent Page Placement becomes opt-out instead of opt-in - Kubernetes DRA (dynamic resource allocation) graduated to stable - systemd-oomd kill threshold getting smarter All of these are building on each other to allow extreme over-provisioning of ram on hyperscalers. You might think your cloud instance has 128gb of RAM…it could be only 32GB of real DRAM. It’s just really, really clever. I’ll be writing a series over the next few days on each of these, stay tuned!
As a backend engineer. Please learn: - SOLID design principles - Multithreading - Immutability - Streaming , messaging - Caching - Security - SSL, JWT, OAuth - factory, decorator, singleton, obeservable design patterns - TDD All very important topics .
