Maximum Throughput?

CHorn

Hi,

Just wondering what the best up/down I can request from my ISP given stats below. Line seems pretty sound (no errors to date) and 3 upped it from 18 to 25 a year ago with no problems.

In particular, It's the attainable throughput of 50 I'm looking at- why can't I get it, is it possible or what needs to happen to get my ISP to enable it?

Any help/pointers appreciated.

The SNR margin isn't good, you're pretty much stuck with what you have to be honest. Your upload looks quite low to me though, maybe there's a problem somewhere.

CHorn

Karsini wrote: »

The SNR margin isn't good, you're pretty much stuck with what you have to be honest. Your upload looks quite low to me though, maybe there's a problem somewhere.

Thanks for this.

My understanding is ISPs consider 6 SNR OK these days, 10+ is good, 9 close enough to good. On the upload, 2887 send is close to attainable 2936 so OK by me (the line has to carry VOIP too?).

But it's the difference between "Attainable Throughput" of 50774 versus Max and Current of 25600 and 25136 (so current is close to capped max) that I'm looking to understand/improve on.

ED E

6dB is OEs ADSL target. The VDSL target is 9/10dB, the DSLAM reduces as far as 9 to sync. Remember 9db is 2x as strong as 6dB so the difference is huge.

Your upstream is about 50% the normal profile and thats the modem doing the best it can. Try and confirm your distance from the cabinet. 28dB says you're close to 2KM so probably doing well for your location or theres a problem in the copper.

CHorn

Thanks. Yes, distance c2Km.

If the down max is ISP-capped at 25600 I'm getting awful close to that with 25136. The up I'm not that worried about as I don't do much 'up' work. But how can I get closer to the magic "Attainable Throughput" # of 50774?

Is "Attainable" in this context under perfect conditions, or has this been tried and tested versus my modem by the ISP in some form or fashion and found to work? And would turning "Seamless rate adaptation" on help, as in it would start at the 50774 # and work downwards to a stable connection?

ED E

Fritz is just being optimistic as hell, possibly a wobbly g.Vector implementation.

Most ISPs hide that figure as it creates queries like this thread. Really you're doing very well getting as much as you are.

ED E

ED E wrote: »

Fritz is just being optimistic as hell, possibly a wobbly g.Vector implementation.

It was never a good estimation. I remember a Fritz 7140 saying that a G.DMT line could do 12Mb whereas the spec doesn't go over 8.

ED E

Karsini wrote: »

It was never a good estimation. I remember a Fritz 7140 saying that a G.DMT line could do 12Mb whereas the spec doesn't go over 8.

Put a decent modem in there and they can be ok (at the time of sync). Can be.

CHorn

Thanks to both of you. The trouble great thing with boards.ie is the education in the answers you get just leads to more Qs ;-).

I just got the Fritz 7360 from Digiweb last week and am kicking the tyres, after pulling the plug on the Huawei HG658c I had with 3. It seems pretty handy- I was pleasantly surprised when it immediately recognized my bridged router (another 658c) used to extend WiFi range via a Powerlink connection without me having to reconfigure anything.

You mentioned "put a decent modem in there and they can be ok", so just wondering how you'd rate the 7360 (or versus the 658c). And I'm aware you can split out the modem and router functions into 2 machines, but are there any (major) advantages to doing this given what I've got?

The 7360 provides one direct Ethernet connection and WiFi coverage in the front of the house, the bridged 658c sorts WiFi in the rear of the house (albeit slowly).

Personally I'd use the 7360 over a HG658 (the latter doesn't have gigabit Ethernet for a start) but I may be biased because I used to work for a company who were AVM's Irish distributor, so I know them fairly well.

ED E

The 658 is old tat, Vodafone were shipping them to VDSL subs until recently which IMO was the height of cheapness. The 659 is a much better beast (F2000 is a rebadge of it).

The advantage of Huawei is they make the DSLAMs so its pretty much guaranteed to play nice. Other niche OEMs like AVM/Dray etc can offer good value but also a slightly less tested product.

Use the Huawei as a sanity check for line stats, but use the Fritz.

CHorn

Thanks, I'll try configuring my 3 HG658c tomorrow to check stats on new Digiweb connection.

CHorn

I didn't bother testing the HG658c for stats. It seems the Fritz!Box is better all round, plus I can understand why Digiweb or anyone else wouldn't want to let customers loose with maximum possible throughput and try to rein them in as they encounter various issues.

So now I'm on to IPv6 to continue trying to get best performance out of the DSL connection and what a universe we seem to have there!

I don't think Digiweb (or mainstream vendor anyone else?) natively support it, but if I enable IPv6 support, aside from 'use native IPv6 connection' as options I've:

1. Use native IPv4 connection (recommended)- First a native IPv4 connection is established. If a 6RD server address was learned through DHCP, a 6RD tunnel is established. Otherwise the device will attempt to establish a native IPv6 connection (Dual Stack).
2. Use native IPv6 connection- For this operation mode your Internet service provider must support native IPv6 on your line.
3. Use IPv6 with a tunnel protocol- Use IPv6 with a tunnel protocol over a conventional IPv4 connection. To use this operating mode your Internet service provider does not have to support IPv6.

So selecting 3, we have 6to4, SixXS (now defunct), 6RD and 6in4 as protocols and reading up, let's try 6RD. And it works, at least in the sense I get an IPv6 address, an IPv6 prefix for tunnel use, and I've set the IPv6 DNS servers to Google's rather than get handed Digiweb's (if that happens anyway).

If I then run http://test-ipv6.com/ I get a 10/10 score and can connect to almost all the 'Other IPv6 Sites' (Tab).

But then I wonder whether I'm completely wasting my time, since 'ping -6 www.google.com -n 4' (or .ie) gets me an average time of 36ms, and that's on a good day, other times 1 or 2 of the 4 will time out, whereas ping www.google.com -n 4 comes in at 6-7ms consistently. And this with the 2 sites most trafficked on my network.

I'd note if you (try to) use IPv6, Google's "Chrome" has a "fast fallback" mechanism. On the first try to a site, it will prefer IPv6. If connections take longer than a third of a second, IPv4 is attempted in parallel; and the better of the two will be used for that site (From: http://test-ipv6.com/)

Further outlined @ https://codereview.chromium.org/7029049):

When a hostname has both IPv6 and IPv4 addresses, and the IPv6 address is listed first, we start a timer (300ms) (deliberately chosen to be different from the backup connect job). If the timer fires, that means the IPv6 connect() hasn't completed yet, and we start a second socket connect() where we give it the same AddressList, except we move all IPv6 addresses that are in front of the first IPv4 address to the end. That way, we will use the first IPv4 address. We will race these two connect()s and pass the first one to complete to ConnectJob::set_socket().

Adds 4 new TCP connection latency histograms to assess the new behavior:

1. IPv6 raceable (includes both when it races and doesn't, which are distinguished by whether or not the samples exceed the race timeout of 300ms)
2. IPv6 solo (no IPv4 addresses to race against)
3. IPv4 no race (IPv4 is the first address, so we're not racing)
4. IPv4 wins race (IPv4 raced and won, even though it started behind).

So if the clever lads at Google have been there/done that, who am I to argue, and maybe it's best to stay in an all-IPv4 universe until there's no longer a need to stuff IPv6 packets into IPv4 envelopes and deal with the vagaries of all the routing equipment on the way to wherever you're trying to move data from/to?

Is it just better to turn anything IPv6 off, so you don't even bother with the overhead of the fast fallback test since Google Chrome (my main web access tool) is almost always going to use it?

Any thoughts appreciated.

P.S. I could post a new thread and ask specifically if anyone has managed to get a Digiweb Fritz home router to work effectively with IPv6, but thought I'd follow up here first.

ED E

Turn it all off. V6 has been going for 20yrs and is still dead.

CHorn

Thanks. You- or others who stumble across this- might have interest in the following which would represent state of play in late 2016 as regards 'Happy Eyeballs' (i.e. end users- sometimes aka customers- who want quick search++ response times):

TLDR (well almost): https://orhanergun.net/2017/04/what-is-happy-eyeballs/

State of Play: https://ripe73.ripe.net/presentations/125-index.pdf

Background++:
https://tools.ietf.org/html/rfc6555#page-4
https://blog.apnic.net/2017/06/13/stops-ipv6-traffic-dual-stack-isp/

It would seem dual stack using this 'Happy Eyeballs' algorithm is way forward? I wish it was configurable in Chrome (say) so I could reduce it from the 300ms it is today with a lot of IPv4 fallbacks and increase it as rate of IPv6 adoption increases.

Anyone with any extra info, feel free to add.

CHorn

End of story (so far- late 2017): I turned everything IPv6 off via https://support.microsoft.com/en-sg/help/929852/how-to-disable-ipv6-or-its-components-in-windows, downloading and then running the following:

1) Disable IPv6 on nontunnel interfaces (except the loopback) and on IPv6 tunnel interface
2) Prefer IPv4 over IPv6 in prefix policies
3) Disable IPv6

Then unchecked anything IPv6 in the Ethernet Properties of my Network Adapter.

I think the only case for using IPv6 here (Digiweb, Dublin- Ireland, Fritz!Box 7360, 25Mbps FTTH package, Windows 10) is if I wanted to set up a Home Network, and that's only because it uses IPv6 internally.

Otherwise the overhead involved in trying to navigate among v4 and v6 packets and routers just seems too high. Maybe when there's more IPv6 uptake I'll revisit.