[gentoo-user] {OT} Cable latency & Skype

[gentoo-user] {OT} Cable latency & Skype

[Grant]

I just switched from DSL to cable and I'm noticing a significant delay
when using Skype, even when nothing else is happening on my network.
Has anyone else noticed this and had success "fixing" it?  I'm using a
Gentoo router so I can try just about anything.

- Grant
-- 
gentoo-user@gentoo.org mailing list

Re: [gentoo-user] {OT} Cable latency & Skype

[Mick]

[-- Attachment #1: Type: text/plain, Size: 2808 bytes --]

On Monday 12 November 2007, Grant wrote:
> I just switched from DSL to cable and I'm noticing a significant delay
> when using Skype, even when nothing else is happening on my network.
> Has anyone else noticed this and had success "fixing" it?  I'm using a
> Gentoo router so I can try just about anything.

If this is a MTU problem (it may very well not be, see below) I would go about 
it as follows.

Do you know what is the recommended MTU of your ISP?  What MTU your modem is 
set at?  If not and they won't tell you you could try to find out by 
yourself:

Set the MTU of *all* nodes between your PC & your ISP's modem to 1500 (PC, 
router, etc).  Then run something like this:

$ ping -c 3 -v -M do -s 1452 www.yahoo.com  (better if you use the IP address 
of yahoo.com, of even better if you use one of your ISP's gateways, DNS 
servers, etc. to avoid problems with network congestion upstream - in that 
sense ebay on a Sunday afternoon may not be a good choice ;).

Work your way up to greater packet values until you cannot ping anymore 
because packet fragmentation occurs.  Once you find the value at which it 
switches over then add 28 (for the packet headers) to the maximum value at 
which ping works. This should be the value at which your modem is set.

Set all other network components to the same MTU (inc. all PCs in your 
network) and you should be good - BUT only if the modem's MTU was correct in 
the first place.

If your modem has the wrong MTU for the ISPs network assuming they don't tell 
you what that is, then try altering it from 1500 downwards to 1492, 1480, 
1458, 1430, 1400, while you test for maximum line speed.  Set all your 
network components to the same MTU and check with a broadband speed tester 
(e.g. www.speetest.net) to see when you get maximum speed.  Alternatively, 
try downloading some iso image from a main mirror near you (warning -  not 
all mirrors are born equal in terms of bandwidth, so some experimentation is 
required).

Eventually, you'll find some max MTU value at which you get the best 
comparative performance.  For the tests to be meaningful they should be 
better performed during quiet times (see contention below).  BTW, from the 
little I know about cable the MTU is usually set at 1500, but others could 
advise better.

If all that doesn't fix it, then it may have nothing to do with MTU and it 
could be related to faulty components in your network (cables, switches, 
modem, etc.); ISP throttling (is poor performance occurring only some 
times/days or is it universal); busy Skype servers(?), or good old high level 
pipe contention, because the world and his wife are on the Internet these 
days and the infrastructure has not grown enough for it.

HTH.
-- 
Regards,
Mick

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] {OT} Cable latency & Skype

[Mark Shields]

[-- Attachment #1: Type: text/plain, Size: 1100 bytes --]

On Nov 12, 2007 6:59 PM, Grant <emailgrant@gmail.com> wrote:

> I just switched from DSL to cable and I'm noticing a significant delay
> when using Skype, even when nothing else is happening on my network.
> Has anyone else noticed this and had success "fixing" it?  I'm using a
> Gentoo router so I can try just about anything.
>
> - Grant
> --
> gentoo-user@gentoo.org mailing list
>
>
I work for as a cable modem technician.  The first thing to check for when
you're having cable internet problems is the modem.  Call up tech support
and ask them to check the signals on the modem (upstream power, downstream
rcv, downstream SNR, upstream SNR, headend receive) and make sure they're in
range.  Also ask them to ping and (if available) rf ping to check for
latency/packet loss.  Also ask them to check the circuits/backbone.  Also,
can you reproduce this latency in the form of a ping/traceroute?  This will
go a long way with ISPs in determining where the problem is (although
Comcast just blows off high latency on pings as the result of dropping them
due to lower priority).

-- 
- Mark Shields

[-- Attachment #2: Type: text/html, Size: 1494 bytes --]

Re: [gentoo-user] {OT} Cable latency & Skype

[Mick]

[-- Attachment #1: Type: text/plain, Size: 3305 bytes --]

On Tuesday 13 November 2007, Mark Shields wrote:
> On Nov 12, 2007 6:59 PM, Grant <emailgrant@gmail.com> wrote:
> > I just switched from DSL to cable and I'm noticing a significant delay
> > when using Skype, even when nothing else is happening on my network.
> > Has anyone else noticed this and had success "fixing" it?  I'm using a
> > Gentoo router so I can try just about anything.
> >
> > - Grant
> > --
> > gentoo-user@gentoo.org mailing list
>
> I work for as a cable modem technician.  The first thing to check for when
> you're having cable internet problems is the modem.  Call up tech support
> and ask them to check the signals on the modem (upstream power, downstream
> rcv, downstream SNR, upstream SNR, headend receive) and make sure they're
> in range.  Also ask them to ping and (if available) rf ping to check for
> latency/packet loss.  Also ask them to check the circuits/backbone.  Also,
> can you reproduce this latency in the form of a ping/traceroute?  This will
> go a long way with ISPs in determining where the problem is (although
> Comcast just blows off high latency on pings as the result of dropping them
> due to lower priority).

Interesting to hear this.  The OP will no doubt have a different traceroute to 
show the ISP, but does the comment on dropping pings explain the % loss shown 
below in certain hops, or is it just a matter of overloaded switches?
==========================================================================
HOST: lappy                       Loss%   Snt   Last   Avg  Best  Wrst StDev
  5. 217.41.177.66                 0.0%    15   17.9  18.0  15.7  22.8   1.7
  6. 217.41.177.134                6.7%    15   21.0  17.5  15.7  21.0   1.5
  7. 217.41.177.54                 0.0%    15   17.0  16.6  15.1  20.7   1.4
  8. 217.47.166.106                0.0%    15   16.0  16.9  15.3  18.9   1.1
  9. core1-pos5-2.faraday.ukcore.  0.0%    15   17.0  45.3  15.2 192.3  52.7
 10. core1-pos0-15-0-10.ilford.uk  0.0%    15   18.9  18.3  17.1  19.5   0.7
 11. 194.74.77.222                 0.0%    15   18.1  17.1  15.5  19.1   1.0
 12. t2c1-ge14-0-0.uk-ilf.eu.bt.n  6.7%    15   17.9  17.3  15.7  19.1   0.9
 13. t2c1-p4-0-0.us-nyc.eu.bt.net  0.0%    15  107.3 108.1 106.1 109.7   1.1
 14. 12.116.102.17                 0.0%    15  108.3 107.9 105.5 110.0   1.3
 15. tbr1.n54ny.ip.att.net         0.0%    15  133.2 133.8 131.2 135.4   1.4
 16. cr2.n54ny.ip.att.net          0.0%    15  135.2 133.5 131.6 135.7   1.3
 17. cr2.wswdc.ip.att.net          0.0%    15  132.2 132.9 131.3 134.7   1.1
 18. cr1.attga.ip.att.net          0.0%    15  134.2 133.6 132.1 135.7   1.2
 19. tbr2.attga.ip.att.net         0.0%    15  135.2 134.0 132.0 136.2   1.3
 20. gar4.attga.ip.att.net         0.0%    15  132.2 134.1 130.0 159.4   7.1
 21. 12.124.64.62                 20.0%    15  140.2 138.6 137.0 140.4   1.1
 22. te-9-1-ur01.south.tn.knox.co  6.7%    15  141.2 140.4 138.1 141.5   1.0
 23. te-8-3-ur02.west.tn.knox.com  0.0%    15  141.2 140.3 139.1 141.2   0.6
 24. ge-1-46-ur01.west.tn.knox.co  0.0%    15  138.2 138.6 137.8 140.6   0.9
==========================================================================

Note some of these are being dropped in the UK, rather than by Comcast.
-- 
Regards,
Mick

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [gentoo-user] {OT} Cable latency & Skype

[Mark Shields]

[-- Attachment #1: Type: text/plain, Size: 5092 bytes --]

On Nov 13, 2007 9:15 AM, Mick <michaelkintzios@gmail.com> wrote:

> On Tuesday 13 November 2007, Mark Shields wrote:
> > On Nov 12, 2007 6:59 PM, Grant <emailgrant@gmail.com> wrote:
> > > I just switched from DSL to cable and I'm noticing a significant delay
> > > when using Skype, even when nothing else is happening on my network.
> > > Has anyone else noticed this and had success "fixing" it?  I'm using a
> > > Gentoo router so I can try just about anything.
> > >
> > > - Grant
> > > --
> > > gentoo-user@gentoo.org mailing list
> >
> > I work for as a cable modem technician.  The first thing to check for
> when
> > you're having cable internet problems is the modem.  Call up tech
> support
> > and ask them to check the signals on the modem (upstream power,
> downstream
> > rcv, downstream SNR, upstream SNR, headend receive) and make sure
> they're
> > in range.  Also ask them to ping and (if available) rf ping to check for
> > latency/packet loss.  Also ask them to check the circuits/backbone.
>  Also,
> > can you reproduce this latency in the form of a ping/traceroute?  This
> will
> > go a long way with ISPs in determining where the problem is (although
> > Comcast just blows off high latency on pings as the result of dropping
> them
> > due to lower priority).
>
> Interesting to hear this.  The OP will no doubt have a different
> traceroute to
> show the ISP, but does the comment on dropping pings explain the % loss
> shown
> below in certain hops, or is it just a matter of overloaded switches?
> ==========================================================================
> HOST: lappy                       Loss%   Snt   Last   Avg  Best  Wrst
> StDev
>  5. 217.41.177.66                 0.0%    15   17.9  18.0  15.7  22.8
> 1.7
>  6. 217.41.177.134                6.7%    15   21.0  17.5  15.7  21.0
> 1.5
>  7. 217.41.177.54                 0.0%    15   17.0  16.6  15.1  20.7
> 1.4
>  8. 217.47.166.106                0.0%    15   16.0  16.9  15.3  18.9
> 1.1
>  9. core1-pos5-2.faraday.ukcore.  0.0%    15   17.0  45.3  15.2 192.3
> 52.7
>  10. core1-pos0-15-0-10.ilford.uk  0.0%    15   18.9  18.3  17.1  19.5
> 0.7
>  11. 194.74.77.222                 0.0%    15   18.1  17.1  15.5  19.1
> 1.0
>  12. t2c1-ge14-0-0.uk-ilf.eu.bt.n  6.7%    15   17.9  17.3  15.7  19.1
> 0.9
>  13. t2c1-p4-0-0.us-nyc.eu.bt.net  0.0%    15  107.3 108.1 106.1 109.7
> 1.1
>  14. 12.116.102.17                 0.0%    15  108.3 107.9 105.5 110.0
> 1.3
>  15. tbr1.n54ny.ip.att.net         0.0%    15  133.2 133.8 131.2 135.4
> 1.4
>  16. cr2.n54ny.ip.att.net          0.0%    15  135.2 133.5 131.6 135.7
> 1.3
>  17. cr2.wswdc.ip.att.net          0.0%    15  132.2 132.9 131.3 134.7
> 1.1
>  18. cr1.attga.ip.att.net          0.0%    15  134.2 133.6 132.1 135.7
> 1.2
>  19. tbr2.attga.ip.att.net         0.0%    15  135.2 134.0 132.0 136.2
> 1.3
>  20. gar4.attga.ip.att.net         0.0%    15  132.2 134.1 130.0 159.4
> 7.1
>  21. 12.124.64.62                 20.0%    15  140.2 138.6 137.0 140.4
> 1.1
>  22. te-9-1-ur01.south.tn.knox.co  6.7%    15  141.2 140.4 138.1 141.5
> 1.0
>  23. te-8-3-ur02.west.tn.knox.com  0.0%    15  141.2 140.3 139.1 141.2
> 0.6
>  24. ge-1-46-ur01.west.tn.knox.co  0.0%    15  138.2 138.6 137.8 140.6
> 0.9
> ==========================================================================
>
> Note some of these are being dropped in the UK, rather than by Comcast.
> --
> Regards,
> Mick
>

I would like to mention that while I am not a cable modem field tech, I do
work in an escalated dept (Tier II).  That said, most of the time when you
see packet loss/high latency at one hop, you'll see it at the sequential
hops after that if it's a true packet loss/latency issue and not just the
ICMP packets being given lower priority/dropped.  The packet loss could also
be that hop/ISP dropping the packet because it detected what it might
consider "too many pings" (flood protection, I assume).  I've seen Comcast
drop on a 3rd hop before.    In the case of ICMP packets having lower
priority, it's best to just ping the host you're trying to get to then go
from there - like an average of 100 sequential pings, for example.
Generally speaking, if a basic ping such as this returns latency/packet
loss, there's a problem somewhere along the line, and you can continue with
further testing such as traceroutes, speed tests, and individually pinging
possible problematic hops.  Concerning Comcast, I called them once and
complained about latency; they rebutted with the fact ICMP packets have a
lower priority on their network.

That doesn't make any sense to me, though.  If they're having to drop ICMP
packets, what does that say about the capacity of the network?  Regardless,
the best way to test for packet loss is to run a speed test.  If your speeds
are decently consistent and what you pay for (or close to it), then packet
loss isn't an issue (I recommend speedtest.net).

One last thing:  this thread is way off-topic.  I suggest we take this to
another forum or just e-mail off this mailing list if we wish to continue.

-- 
- Mark Shields

[-- Attachment #2: Type: text/html, Size: 8091 bytes --]