Jan Stary wrote:
> last night, I installed 4.1 on the new ALIX.1C:
I recently got a Soekris net5501, which is uncannily similar (I
guess they're both based on the same reference design), and moved
the same kind of infrastructure functions to that box, so I had to
look at similar decisions.
> Firstly, swap (i don't really mind reinstalling).
Indeed. Just run without swap.
> (how would I do that? A 'b' partition of zero size, as it has to exist?),
Actually, it does not have to exist.
> but to be able to save possible core dumps, I am thinking of 300M
Do you want to do kernel development and debugging on that box?
It depends on how you view the machine. I decided to forgo the
usual multiuser system approach and treat the box as an appliance.
The whole point is that it will just sit there, performs its job,
and I won't have to touch it. I didn't twiddle with settings unless
required for functionality. No need for a pretty shell prompt. I
didn't even bother to create a user account. What for? I'd have
to prefix nearly all commands with sudo anyway. Partitions? There's
only a single partition 'a'.
> Secondly, the network interfaces. The box comes with an on-board
Well, near the top of /sys/dev/ic/rtl81x9.c you can find Bill Paul's
famous rant on just how crappy the rl(4) hardware is. He concludes:
"It's impossible given this rotten design to really achieve decent
performance at 100Mbps, unless you happen to have a 400MHz PII or
some equally overmuscled CPU to drive it."
That was written quite a few years ago, and as wimpy as a Geode
LX800 may seem today, it qualifies as an "overmuscled CPU". Any
of your cards above will be fine. I doubt you're going to notice
any difference.
> I don't have any idea about what amount of e.g. fragment reassembly the
Fragment reassembly doesn't happen in the driver.
> Thirdly, the CF storage. Having read
I'm not.
> As these articles are from 2005 - do these things still apply to
I don't think these things still applied back then either. At
EuroBSDCon 2005, Poul-Henning Kamp, who has a lot of experience
with this, broached the topic in one of his talks and basically
said that it wasn't a concern in practice and that he wanted to try
out a flash drive as his laptop disk.
> What else should I do to make the CF card live longer (noatime
Buy a bigger flash so wear-leveling can spread the writes around.
But with CFs now starting at 1 GB, this isn't an issue either.
--
Christian "naddy" Weisgerber naddy@mips.inka.de
| Jens Axboe | Re: [BUG] New Kernel Bugs |
| KAMEZAWA Hiroyuki | Re: 2.6.24-rc3-mm1 |
| Ingo Molnar | Re: [Announce] [patch] Modular Scheduler Core and Completely Fair Scheduler [CFS] |
| Greg Kroah-Hartman | [PATCH 001/196] Chinese: Add the known_regression URI to the HOWTO |
git: | |
| Jarek Poplawski | [PATCH] pkt_sched: Destroy gen estimators under rtnl_lock(). |
| Gerrit Renker | [PATCH 27/37] dccp: Integration of dynamic feature activation - part 2 (server side) |
| Jarek Poplawski | Re: Data corruption issue with splice() on 2.6.27.10 |
| Patrick McHardy | Re: [GIT]: Networking |
