"Just" 270 MB for...an idle server

[stefano@mastodon.bsd.cafe]

RE: https://mastodon.social/@nixCraft/115566280074527897

"Just" 270 MB for...an idle server?
Debian is still a great distribution but let's measure the ram consumption of a freshly installed *BSD or Illumos based server. The numbers are totally different.

#RunBSD #illumos #FreeBSD #OpenBSD #NetBSD #DragonflyBSD

[washbear@mastodon.sdf.org]

@stefano you can't compare the memory usage of different unix variants easily because they all use and account for caching differently anyway.

netbsd will be using 32mb and then top will report that "2gb" is in use after /etc/daily runs

[stefano@mastodon.bsd.cafe]

@washbear of course. But I've recently tested Raspbian vs NetBSD on an old Raspberry Pi A+. Raspbian was almost unusable, swapping all the time. NetBSD is totally usable - I'm not talking about desktop, of course.

[gumnos@mastodon.bsd.cafe]

@stefano curious what your other fresh installs rank in at.

A fairly fresh FreeBSD install on my MassiveGrid VM according to top(1):
7044K Active
109M Inact
1572M Wired
2232M Free
ARC: 1238M Total, 498M MFU, 654M MRU, 512K Anon, 9920K Header, 70M Other

And a fairly fresh OpenBSD install on a RackNerd VM:
Real: 29M/1210M act/tot
Free: 743M
Cache: 380M

I'm never quite certain how to measure RAM consumption though.

[cienmilojos@infosec.exchange]

@stefano none of this matters to measure with nothing running. An idle server with nothing running is a waste of compute. Servers best unit of measure is typically uptime and power usage. Willing to bet both Debian and BSD manage the task far more efficiently than a Windows server.

[stefano@mastodon.bsd.cafe]

@cienmilojos I know. But if the idle, useless system is using 270 MB for doing nothing, every single service will raise that number. IMHO, an idle server (freshly installed) should use some MB of ram, not hundreds. Let's not talk about GBs...

[peteorrall@mastodon.bsd.cafe]

@stefano I think if we are going to be analyzing RAM usage across OSes, then usage needs to be recorded every X minutes over a specific workload, with idle being baseline or control group. Just showing 270MB RAM at idle and then another single snapshot at 10GB of some kind of workload still doesn't tell much. It's difficult to infer anything else.