What does a slow ping response mean?

I have a client whose website I was unable to connect to for about 3
weeks. When I try to load their URL at work, I'd get a "Server not
responding" message after several minutes. Yet I was able to
successfully connect from home.

My ISP tech support person said he pinged the site and got response
times averaging 200-300 ms. He said that was very slow and that the
site is probably being overwhelmed with traffic. In the evening when I
can get through on my home account, traffic is probably slower.

I tried pinging the site myself and got the same repsonse times. Then I
tried pinging several sites that I have no problem loading and got
response times ranging from .50 ms to 450.00 ms. That seems to suggest
that response time range of 200-300 ms doesn't significantly slow down
a page loading into a browser.

Whatever the cause of the problem, I discovered today that the client's
site now loads fine -- instantly on my broadband connection at work.
For grins I ping them and got response times avg 775.708 ms! What the
hell? What is the significance of the ping response time?

And yes, I'm a networking neophyte...

On 14 Jun 2006, in the Usenet newsgroup comp.os.linux.networking, in article
<1150299638.761374.171610@u72g2000cwu.googlegroups .com>, Joe Shockey wrote:

>I tried pinging the site myself and got the same repsonse times. Then I
>tried pinging several sites that I have no problem loading and got
>response times ranging from .50 ms to 450.00 ms. That seems to suggest
>that response time range of 200-300 ms doesn't significantly slow down
>a page loading into a browser.

Apples are not Oranges. TCP/IP is not ICMP. There is a thing called "policy
routing" which may prioritize which packets get forwarded first. ICMP is
very often set to a low priority (it's not carrying the jpeg of that
bunch of dancing gerbils you are trying to load) and the packets may even
be dropped. As for what is a long or short delay, that obviously depends
on where the packets are being routed. A single hop (out and back going and
coming back) to a satellite takes just over half a second ignoring any
delays within the hardware itself and other traffic that might be using
the link at the same time. For ground based links, look at the traceroute
outputs to see where the packet is being routed. A packet stuck into a
fiber in New York is going to take 39 msec for the round trip to Los Angeles
(again, assuming no hardware delays and no competing traffic). See also the
response to your other posting "How can traceroute fail, yet the site still
open in a web browser?" which is part of the same concept.

>For grins I ping them and got response times avg 775.708 ms!

How busy are they - how busy are the links in between?

>What is the significance of the ping response time?

Not to much. Still, the gamers want "low ping times" so that the ogre that
is guarding the castle gate doesn't zap them before the even realize he's
there.

Old guy

On Wed, 14 Jun 2006 08:40:38 -0700, Joe Shockey wrote:

> When I try to load their URL at work, I'd get a "Server not
> responding" message after several minutes.

> My ISP tech support person said he pinged the site and got response
> times averaging 200-300 ms.

> Then I
> tried pinging several sites that I have no problem loading and got
> response times ranging from .50 ms to 450.00 ms.

> For grins I ping them and got response times avg 775.708 ms!

Are you pinging the name, or the IP address?

Mike

Mike Playle <usenet@vfx.org.uk> wrote:
> Are you pinging the name, or the IP address?

IIRC it shouldn't matter. The name to IP lookup would happen once,
before the first ICMP Echo Request (ping) was sent and wouldn't be
part of the timing window. Slow DNS might mean it takes a while
before the first ICMP Echo Request is sent, but would not otherwise
affect ping times.

rick jones
--
No need to believe in either side, or any side. There is no cause.
There's only yourself. The belief is in your own precision. - Jobert
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...

On Fri, 16 Jun 2006 00:06:52 +0000, Rick Jones wrote:

> Mike Playle <usenet@vfx.org.uk> wrote:
>> Are you pinging the name, or the IP address?
>
> IIRC it shouldn't matter. The name to IP lookup would happen once,
> before the first ICMP Echo Request [...]

Yes, you're absolutely right... I jumped to a conclusion.
Whenever I see a problem that takes "several minutes" I
immediately think of DNS trouble!

Mike

ibuprofin@painkiller.example.tld (Moe Trin) wrote:
>>What is the significance of the ping response time?
>
>Not to much. Still, the gamers want "low ping times" so that the ogre that
>is guarding the castle gate doesn't zap them before the even realize he's
>there.

Actually, for someone who knows what they are looking at, it is
*very* informative. But it requires knowing more than just the
ping times to make heads and tails of it.

Knowing the network topology, it is very possible to glean a
great deal of information about network status by using ping.
And sometimes is it possible to learn what the network topology
is by using ping.

But trouble shooting a misbehaving network without knowing the
topology is not possible. (Which is to say, if you don't know
what ping _should_ show, you cannot make decission about the
signficance of what it does show.)

For example, during normal traffic there should be periods of
relatively light loading. Perhaps not often, but once in
awhile. Hence on occasion the ping time will be as small as it
can possibly be. If you let ping ring for some period of time,
you'll see what that smallest ever value is. And that value
will tell you what kind of a link exists between your host and
the target. Or, more correctly, it will tell you what the
slowest part of the link is. A 14.8Kbps modem has a slower time
than a 28.8Kbps, which is slower than a 45Kbps v.90 connection,
which is slower than a 56Kbps data circuit, which is slower than
a T1, which is slower than a T3...

And of course, for those of use who live at the end of a
satellite link, that is also very obvious.

--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd@apaflo.com

On Thu, 15 Jun 2006, in the Usenet newsgroup comp.os.linux.networking, in
article <8764j1vg6s.fld@apaflo.com>, Floyd L. Davidson wrote:

>Actually, for someone who knows what they are looking at, it is
>*very* informative. But it requires knowing more than just the
>ping times to make heads and tails of it.

Agreed - used to was when turning on a system that had been physically
moved to a new netdrop, one of our standard tests was 'ping -c 25, -s
16384 name_of_some_host' which most would consider an abuse case. If
so much as one reply (with each ping made up of 11 actual Ethernet
packets) was dropped, there was cause for further investigation. We
don't do that any more, instead grabbing a one Meg file and noting
the '/sbin/ifconfig' errors as well as response times. This test is
more representative of what the network is used for anyway.

Old guy

Thanks everyone for the helpful answers.

So what I'm getting from all this is that ping is pretty useless on the
internet in general. Since packet travel is dyanamic it's impossible to
map the topology -- it can change with every packet. And because it
uses ICMP, servers treat it differntly than a TCP/IP request, and
ignore it or give it lower priority. And since traceroute use ICMP
also, it suffers from the same problems?