Hacker News Reader

Comment by tokyobreakfast | Hacker News Reader

RAM shortage or competent programmer shortage?

Can't get a Linux box to idle (or even install) under 512M these days.

Can't find a web developer worth a shit who doesn't think he needs a Python backend application server to print "Hello, world" when you could do this with a static page served with something like OpenBSD with two-digit RAM requirements.

It's not the RAM that's changed; it's everyone around the RAM.

A coddled generation who were taught that AWS is the Internet and live in abstractions certainly hasn't helped.

View on HN

My NixOS SSH jump host server here idles at 234 MB of which 64 MB is systemd-journald (which I assume can be reduced with some settings of how much to keep in RAM).

>which 64 MB is systemd-journald

why

Windows NT was routinely run with 32 MB of RAM TOTAL and the event log is basically unchanged 30 years later.

Achtung, you will draw the ire of the systemd downvote zealots.

Edit: Haha, withing a handful of seconds I got a downvote. :-D

>which 64 MB is systemd-journald

Wait till they rewrite it in Rust!

There are SoCs with 64 or 128 MB integrated, and people run reasonably complex stuff on it.

I still have 64 MB VPS (OpenVZ) which I use in production since 2012. It runs DNS, VPN, some logging stuff.

That is simply not true, Linux boots just fine with 8 MiB without MMU these days, which is half of the system memory I had available in 1995

You definitely can use Linux with few simple servers with 128 MB RAM.

Install can be tricky indeed, but if you have installed system, it's easier.

Yeah I'll need conclusive proof of that.

This is not difficult, you just need to run `htop` and perform addition of the RES column (which is in KB unless a unit is shown). Example:

    USER         RES▽ Command
    root       70436  systemd-journald
    root       14268  amazon-ssm-agent
    root       13508  systemd
    root       12160  systemd --user
    root       10240  sshd: root@pts/0
    root        9088  sshd: root [priv]
    root        8944  systemd-udevd
    root        8704  systemd-logind
    root        8320  nix-daemon --daemon
    systemd-ti  8192  systemd-timesyncd
    systemd-oo  7808  systemd-oomd
    root        6492  -zsh
    nscd        6272  nsncd
    messagebus  5888  dbus-daemon --system --address=systemd: --nofork --nopidfile -
    root        5888  htop
    sshd        4904  sshd: root [net]
    root        4736  sshd: sshd -D -f /etc/ssh/sshd_config [listener] 1 of 10-100
    root        2960  (sd-pam)
    root        2816  agetty --login-program login ttyS0 --keep-baud
    root        2192  dhcpcd: [privileged proxy]
    dhcpcd      1680  dhcpcd: [manager] [ip4] [ip6]
    dhcpcd      1468  dhcpcd: [BPF ARP] ens5 172.31.8.86
    dhcpcd      1168  dhcpcd: [control proxy]
    dhcpcd      1040  dhcpcd: [network proxy]

Why are you using systemd in a minimalist system?

I'm not trying to run a minimal system. This is just out of the box NixOS.

>> You definitely can use Linux with few simple servers with 128 MB RAM. > > This is not difficult, you just need to run `htop` and perform addition of the RES column (which is in KB unless a unit is shown). Example:

I'm not quite sure what points this makes... That's supposed to fit on 128MB? And it doesn't include memory consumed by the kernel itself (which is not negligible at this scale), and linux needs spare for cache to work remotely decently.

    $ awk '{ tot+=$2 } END { print tot /1024 }' < list
    214.035

I'm sure you can run a linux with 128MB of ram, but certainly not with systemd and a default kernel... Perhaps DSL (damn small linux) or alpine.

Toms root boot (TOMSRTBT) is what you need! Used to fit on a floppy disk.

https://en.wikipedia.org/wiki/Tomsrtbt

The point was that most of the shown software is optional, and you can check how much selecting those parts you care about sum up to.

To provide "conclusive proof" that it's possible to run a 128 MB Linux system.

For example, if you remove (or configure smaller memory use for) journald, the Amazon daemon, oomd, timesyncd, you are already at around 128 MB userspace, with more that can be removed if desired.

And this is on a distro that is not at all designed to be minimal in memory usage.

How much cache is good depends entirely on what you want the system to do.

netBSD! ... o wait not linux... damn

Reply on news.ycombinator.com