questions about freebsd 8

IF you're on xDSL try to cut down your upload speed. Uploading close to the maximum of your connection seems to seriously impact download speeds.
 
It certainly does. Either stick to about 80% of your upload speed, or use ALTQ to prioritise outgoing acks. This is often underestimated! A download of ~6 Mbit can easily cause ~300 Kbit of outgoing acks. If they're obstructed by other traffic or lack of bandwidth, your download will plummet, and you will cause extra traffic because of all the retransmits.
 
phoenix said:
If you have a ZFS system, why aren't you using ZFS for /usr/src, /usr/obj, /usr/ports, and so forth? There's really no reason to keep those on the CF disk.

In fact you can also move /usr/local, /home, /var, /tmp off the CF and onto ZFS as well.

That way, the only time the CF disk gets written to is when you manually edit something in /boot or /etc, or when doing an installkernel/installworld.

A very good reason: so that I can import and export the pools when they go into degraded state and my SATA controllers start flipping around drive order at random when I pop a new drive in. I think that's a very good reason to keep the rootfs separate from the zpool. Maybe if I had another CF I could do zfsroot with an attached drive to form a mirror, but meh, may as well use gmirror for that.
 
I just installed 8.0 earlier today from the 200906 ISO snapshot. So far all good, but I'm still busy building everything. This is on a Dell XPS m1330 laptop.

The reason I made the jump was for USBv2 (and fixed umass/cam), mav@'s power saving improvements, and if_iwn(4). New power saving is incredible - as good as Windows XP it seems! Haven't tested USB much yet, and if_iwn is mostly working. FreeBSD 7 on a modern laptop is bit frustrating so I'm really happy with 8 so far.
 
DrJ said:
Does suspend and resume work?
Sort of. System comes back up, Xorg still works, but network cards are dead and I can't suspend a second time.

Unloading the network modules prior to a suspend might help, but I'll need to setup a custom kernel to test. Stay tuned...
 
FYI, I thought I'd want to use suspend/resume in 8.0, but when I saw how quickly it boots I'm not sure I'll bother. Resuming from suspend takes a bit of time for everything to settle anyway, and for a bit extra I can have a fresh boot.

For me, 8.0 kernel loads in 14 seconds, and after that userland loads in another 3 seconds, and Xorg another 7 seconds. I think I can shave a couple seconds off the kernel load time with a custom compile.
 
aragon said:
For me, 8.0 kernel loads in 14 seconds, and after that userland loads in another 3 seconds, and Xorg another 7 seconds.

That's impressive. I've not heard that about 8. For my desktops, this of course does not matter since they are never turned off. For a laptop, though, it does make a difference.
 
SirDice said:
IF you're on xDSL try to cut down your upload speed. Uploading close to the maximum of your connection seems to seriously impact download speeds.

no, i have cable. I've had a perfect setup for months. I hopnesty think comcast is throttling me. I've got a freebsd router (well, really it's pfsense, which is what got me using freebsd in the first place) and i've had it set to traffic shape for months. The problem is this:

Even if i start just 1 torrent, not even downloading fast, my ping goes up to about 1500 ms instead of around 40ms when i don't run a torrent. It's like a night and day difference. Something is SERIOUSLY wrong. I orignally thought this was due to me switching the server to freebsd because that's when it started but when i think back, it also coincided with an ip address change. My outgoing ip changed around the same time. I think comcast has started some sort of aggressive anti torrent traffic shaping in my area. I've had them out to "fix" my problem and they are due to come out again tomorrow so we'll see. It's pretty ridiculous. If i go to http://www.speedtest.net and test without a single running torrent my results are


now let me start a single torrent and test again


notice the difference in ping? that's the problem i'm having.

The top speed isn't that big of a deal but the ping is killing me.

And right now it's actually not nearly as bad as it's been....tonight it's actually a lot better...so maybe it's something they are working on.

oh, and btw, this torrent hasn't even started uploading yet...so the difference in upload makes no sense really.
 
aragon said:
Unloading the network modules prior to a suspend might help, but I'll need to setup a custom kernel to test. Stay tuned...
Sadly, my hopes were incorrect. Suspending my system breaks both my network devices - if_bge and if_iwn. Some USB devices fail to come back up too. At least it's half working, though. :)
 
I need to (not)write a short story about that.

aragon said:
Sadly, my hopes were incorrect. Suspending my system breaks both my network devices - if_bge and if_iwn. Some USB devices fail to come back up too. At least it's half working, though. :)
I'm just typing "out of my posterior" so to speak, but I had heard that suspend/resume & SMP do not play well together. My experience was that it is not worth the hassle, what with powerd(8) working quite well (powerd_flags=" -a hiadaptive -b adaptive") and boot times being trivial (& built-in session managers for all the applications that matter as far as that goes).

Maybe, some day far in the future when we're ruled by angry robots who whip us for their passionless, but still sadistic pleasure, it will become important. Hopefully I won't live that long.
 
Historically in FreeBSD, yes, SMP suspending has been impossible. It used to leave your system completely unusable. That changed a while ago and it does come back up now. I'm guessing more work just needs to be done to get all drivers suspend friendly. If Windows can do it... :)

My boot times are good, so this isn't a major hassle for me. Like you said, there are other things that can be done to save power, and for the rest nothing beats turning off completely.
 
aragon said:
For me, 8.0 kernel loads in 14 seconds, and after that userland loads in another 3 seconds, and Xorg another 7 seconds. I think I can shave a couple seconds off the kernel load time with a custom compile.
Looks like I shaved off a second with latest code from HEAD and a custom kernel - 23 seconds from the boot loader to the Xorg/XDM login screen. This is on a laptop with Core 2 Duo 2.5 GHz, 2 GB RAM, and 7200 RPM drive. Happiness. :)
 
Resume & SMP only works on amd64 at the moment.
 
Back
Top