1. run a local resolving name-server - bind9 or dnsmasq or something. this makes a *huge* difference. local dns lookups with caching and almost no latency (<0.1ms or similar), rather than 15-50ms RTT each way for *every* DNS lookup (and, of course, lookup times get longer and longer the busier your adsl connection is). you can configure a local resolving dns server to be completely standalone or to forward requests that can't be answered from cache to your ISP's nameserver. if you use your adsl modem to manage your connection (rather than have linux control the modem in dumb bridged mode), then it may already run a dns cache, and it probably does dhcp too so is probably already telling the dhcp clients to use it. note, however, that there are advantages to running a nameserver on a decent linux machine (or even an ancient pentium class linux gateway box) compared to a tiny little modem-class CPU with bugger-all memory. 2. run a local web proxy (e.g. squid or apache/nginx/whatever in proxying mode) to cache images, scripts, etc from frequently visited sites. just as with DNS, having a local cache of recurring elements like images on a page from a local cache can massively improve the responsiveness of web sites. make sure all host(s) on your LAN are configured to use them. If you run your own DHCP server, setting the name-server IP for dhcp clients is easy. For web proxy, you can either manually configure each client/web browser to use your proxy, or use some iptables prerouting REDIRECT rules on your router/gateway box for transparent proxying (transproxying is only evil if your ISP does it to you, not if you do it to yourself :) 3. you may also want to look into QoS traffic shaping to, e.g., prioritise small and/or latency-sensitive packets (e.g. ssh packets, dns lookups, VOIP etc) over bulk downloads like web traffic. some apps need low latency rather than high throughput. wondershaper does an OK job if you don't care to figure out all the intricacies of linux traffic shaping. packaged for debian, and also available at http://lartc.org/wondershaper/ On Mon, Mar 05, 2012 at 06:01:57PM +1100, Erik Christiansen wrote:
About every second day now, my ADSL broadband connection turns to sludge. Just loading the ABC news homepage, or BBC, or anything else, takes up to a minute, or sometimes stalls indefinitely.
weather related? bad wiring + water in the telstra pits?
In firefox, I have network.dns.disableIPv6 = "user set boolean true", but that's more for slow "Looking up ..."
just disable ipv6 at the host level if you're not using it. google shows a number of pages with distro-specific methods, and there's also: http://tldp.org/HOWTO/html_single/Linux+IPv6-HOWTO/ see also /etc/gai.conf for configuring your lookup preferences.
The link looks more than OK:
Connection Status Connected Us Rate (Kbps) 128 Ds Rate (Kbps) 1536
the upstream rate seems low for 1.5Mbps down - is it supposed to be 128K or do you just have a bad line? IIRC, most 1.5Mbps ADSL1 services offer either 384K or 512K up. either way, it'll be maxed out if downloading at full throttle (as a rule of thumb, you'll be sending 10-15% of your downstream bandwidth back upstream as request and ack packets etc). any chance of getting adsl2 where you live? and even ADSL1 should be capable of up to 8MBps down and about 800K up. IIRC, i've been on adsl2 for a few years now (and waiting impatiently for NBN).
Repeated long latencies within Internode's network don't look like they'd help much, but I don't have anything to compare that with. The fact that it seems to have failed to get to the target in the end looks ghastly, and beyond my fixit toolkit.
outside of your network mostly-equals outside of your control.
$ traceroute www.linuxcnc.org traceroute to www.linuxcnc.org (69.163.248.64), 30 hops max, 60 byte packets 1 router (192.168.1.254) 1.110 ms 1.313 ms 1.532 ms 2 lns20.mel4.internode.on.net (150.101.212.44) 16.206 ms 22.500 ms 29.324 ms
that second hop is actually pretty good for a mostly-idle ADSL1 link.
3 te2-2.cor3.mel4.internode.on.net (150.101.208.65) 36.055 ms 42.548 ms 49.196 ms 4 gi6-0-0.bdr1.mel4.internode.on.net (150.101.210.41) 225.556 ms 231.354 ms 238.425 ms
that's not so good if it's "normal" for that hop :( craig -- craig sanders <cas@taz.net.au>