Re: Question about latency

David Schwartz <(E-Mail Removed)> wrote:

> Really you need only do a few traceroutes to see this. You can find
> routes with twenty routers that take 6ms and routes with 5 routers
> that take 300ms. The difference is the geographic distance between the
> routers and the number of fiber miles the data has to take.

Doesn't MPLS and the "invisibility" of its "routing" to traceroute
rather complicate that these days?

rick jones
--
portable adj, code that compiles under more than one compiler
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...

Rick Jones <(E-Mail Removed)> wrote:
> David Schwartz <(E-Mail Removed)> wrote:

> > Really you need only do a few traceroutes to see this. You can find
> > routes with twenty routers that take 6ms and routes with 5 routers
> > that take 300ms. The difference is the geographic distance between the
> > routers and the number of fiber miles the data has to take.

> Doesn't MPLS and the "invisibility" of its "routing" to traceroute
> rather complicate that these days?

For example, the distance per the driving directions for Cupertino, CA
to Ft. Collins CO is 1225 miles. I'll go out on a limb and figure
that any "directish" fibre run between those two points would span
roughly the same distance - most of it is along I80 which I suspect is
a route also not too different from that taken by trains - two
possible rights of way one might expect things to follow.

I can traceroute between systems in both places on a company internal
network. The RTT change for the hop between "last.cup" and "first.ftc"
is roughly 32 milliseconds when unloaded:

5 "last.cup" 4.049 ms 5.445 ms 3.941 ms
6 "first.ftc" 36.912 ms 37.441 ms 37.470 ms

1225 *2 miles in 0.032s is 76562 miles per second, or roughly 0.41c.
I think the lowest figure for fraction of c we've seen thusfar was
something like .65? I find it difficult to believe that the hop-free
distance is off by that much, but I could easily believe (especially
since I have very vague recollections of much shorter times 15 years
ago under a different intranet topology) it if there were actually
several MPLS hops "underneath" that one IP hop. How many and for what
"as the packet transmitted" distance I cannot say because I don't have
visibility to what is happening underneath there. So, I still don't
know how much of the difference between perceived 0.41c and 0.65c is
the added distance and how much is the added hops. I "know" from
other investigations that the bandwidth of the "link" between Cup and
Ft.C is at least 100 Mbit.

Similarly, There is a 3ms difference between Cupertino and Palo Alto:

5 "last.cup" 4.092 ms 3.994 ms 4.905 ms
6 "first.pal" 7.058 ms 7.617 ms 7.126 ms

and roughly 13 miles of driving distance. Round-trip that works-out
to 8666 miles/second, or 0.05c. I find it even more difficult (no,
not impossible, but quite difficult) to believe the packets are really
travelling that much father than 26 miles round-trip. Even if it was
260 miles round-trip it would be only 0.47c.

rick jones
--
firebug n, the idiot who tosses a lit cigarette out his car window
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...

Moe Trin <(E-Mail Removed)> wrote:
> >Similarly, There is a 3ms difference between Cupertino and Palo Alto:

> > 5 "last.cup" 4.092 ms 3.994 ms 4.905 ms
> > 6 "first.pal" 7.058 ms 7.617 ms 7.126 ms

> Hmmm, are Ft. Collins and Page Mill both 6 hops away?

As far as IP TTLs are concerned, yes.

> That might be a hint of hidden stuff under the sheets.

Agreed. Ah the good old days, when a hop was what it was and not some
hidden sub-internet

rick jones
--
firebug n, the idiot who tosses a lit cigarette out his car window
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...

Moe Trin <(E-Mail Removed)> wrote:
> On Tue, 14 Sep 2010, in the Usenet newsgroup comp.os.linux.networking, in
> article <i6onou$9cn$(E-Mail Removed)>, Rick Jones wrote:

> >Moe Trin <(E-Mail Removed)> wrote:

> >> Hmmm, are Ft. Collins and Page Mill both 6 hops away?

> >As far as IP TTLs are concerned, yes.

> You might look at other domestic company facilities.

There are many similar situations for other domestic destinations.

> The idea of a (leased) tunnel makes a lot of sense from the economic
> standpoint as well as technically. You (as the customer) don't need
> to know that the packets are actually being routed via LAX and JFK
> to rack up extra Frequent Flyer miles,

Unless my applications are latency sensitive

rick
--
I don't interest myself in "why". I think more often in terms of
"when", sometimes "where"; always "how much." - Joubert
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 2010-10-02, Industrial One <(E-Mail Removed)> wrote:
> FIOS has a higher ping than cable? Y'all can't be serious... I thought
> FIOS was way faster?
>
> Based on everything I heard on this thread, I get that the shortest
> possible ping halfway around the world is 88 ms since signals at most
> travel 75% of Einsteins/s, but commonly would be double that due to
> the way UDP works to make sure you are connected and twice again
> because it doesn't travel in a completely straight line due to the
> bounces it has to make.
>
> Does this mean that I'd have to be on crack to look forward to latency-
> free real time online gaming in the future?

No. Just run the game all on your own computer.
It also depends on how much data and data packets are exchanged.

Industrial One <(E-Mail Removed)> wrote:
> FIOS has a higher ping than cable? Y'all can't be serious... I
> thought FIOS was way faster?

Higher bandwidth does not guarantee lower latency.

rick jones
--
I don't interest myself in "why." I think more often in terms of
"when," sometimes "where;" always "how much." - Joubert
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 Sunday 24 October 2010 14:30 in comp.os.linux.networking, somebody
identifying as Industrial One wrote...

> On Oct 2, 12:11Â*pm, ibupro...@painkiller.example.tld.invalid (Moe
> Trin) wrote:
>
>> NOTE: Posting from groups.google.com (or some web-forums)
>> dramatically reduces the chance of your post being seen. Â*Find a real
>> news server.
>
> Such as?

Eternal September:

http://www.eternal-september.org

AIOE

http://www.aioe.org

There are others too. Some are free of charge - as the above two -
while others may require a small annual fee. Sometimes ISPs also offer
their own newsfeed, but this beginning to diminish now. My own ISP
still does offer the newsfeed and posting through their own SMTP, but
they have outsourced their news service to an American company now
(Highwinds Media) and they are regularly having difficulties with
regard to posting or retrieving. That's why I have subscribed to
Eternal September myself.

I personally recommend Eternal September; they're very reliable. Just
register at their website (with a valid e-mail address) and they'll
send you the login and password for setting up your newsreader. It's
totally free of charge. (So is aioe.org, but I believe they limit the
number of articles you can send per day.)

As for newsreader programs, GNU/Linux comes with a great choice of them,
ranging from dedicated newsreaders to combined newsreader-and-email
applications - e.g. Thunderbird. There are typical GUI newsreaders,
such as KNode - which is part of the KDE PIM suite - or Pan, or Mozilla
Thunderbird and SeaMonkey. There are also character mode
terminal-based newsreaders, such as slrn, which work with an editor of
your choice - typically GNU Emacs or vi/vim, but other options are
available.

Whatever newsreader you will eventually choose, you must then set it up
with the login information from the news service you wish to use, and
then you subscribe to the newsgroups you are interested in - this is
called "subscribing" but it does not imply that you must pay for it.
It simply means that you are telling your newsreader what groups it
must monitor. Upon the first time you are subscribing, it might be
useful to mark all the posts as read, because there will be a lot of
them - 500 or more per individual group is not exceptional. You can
also tell your newsreader not to download more than a given number of
articles per group, and to download new articles automatically every so
many minutes. (I think I have mine set up to a maximum of 500 new
articles per group, per download.)

Another bit of information in the event that you are not aware of these
things, Usenet is *not* a part of Google Groups. All Google Groups
really does is /archive/ Usenet posts, and offer a (bad) web-based
posting interface to Usenet. What they also do is conveniently omit
these facts, so that inexperienced people tend to think that Usenet is
just part of Google. Yet Usenet already existed long before the World
Wide Web came into being, and thus also before the advent of Google.

http://en.wikipedia.org/wiki/Usenet

Google Groups is most often used by trolls and spammers, and therefore
many news servers have already resorted to blocking posts made through
Google Groups. In addition to that, many regular Usenet denizens whose
newsfeed does not block Google Groups - Eternal September doesn't, but
it does filter out spam messages - have set up a filter to kill posts
coming from Google Groups, based upon the User-Agent or Message-ID
headers. (All Usenet posts have a series of headers, which you won't
get to see when using Google Groups, but once you start using a real
newsreader you will be able to see them as well, and set up your own
filters if need be.)

--
*Aragorn*
(registered GNU/Linux user #223157)

Industrial One <(E-Mail Removed)> writes:

> On Oct 2, 12:11Â*pm, ibupro...@painkiller.example.tld.invalid (Moe
> Trin) wrote:
>> On Fri, 1 Oct 2010, in the Usenet newsgroup comp.os.linux.networking, in article
>> <9d7adea4-9acf-4502-82e9-ab7b3cbea...@l20g2000yqm.googlegroups.com>,
>>
>> Industrial One wrote:
>>
>> NOTE: Posting from groups.google.com (or some web-forums) dramatically
>> reduces the chance of your post being seen. Â*Find a real news server.
>
> Such as?

I use eternal-september.org
--
As we enjoy great advantages from the inventions of others, we should
be glad of an opportunity to serve others by any invention of ours;
and this we should do freely and generously. (Benjamin Franklin)

(E-Mail Removed)d (Moe Trin) writes:

> NOTE: Posting from groups.google.com (or some web-forums) dramatically
> reduces the chance of your post being seen. Find a real news server.

Can you quantify 'dramatically'?

--
http://www.greenend.org.uk/rjk/

On Mon, 25 Oct 2010 11:31:02 +0100, Richard Kettlewell wrote:
> (E-Mail Removed)d (Moe Trin) writes:
>
>> NOTE: Posting from groups.google.com (or some web-forums) dramatically
>> reduces the chance of your post being seen. Find a real news server.
>
> Can you quantify 'dramatically'?

Nothing on usenet can be quantified -- except, perhaps, the effluent
issuing via spammers using google groups. :-)

I do not see google group postings -- I have them killfiled.

Tho', I do see gg followups on the assumption that they may provide
information or amusement. That follows from the triumph of hope over
experience, tho'.

Far too many overwise 'nice' folks feel compelled to reply to spammer
postings in the insane belief that the spammer actually reads usenet.

Jonesy
--
Marvin L Jones | jonz | W3DHJ | linux
38.24N 104.55W | @ config.com | Jonesy | OS/2
* Killfiling google & XXXXbanter.com: jonz.net/ng.htm