[Micronet] Logging network performance?

classic Classic list List threaded Threaded
10 messages Options
Reply | Threaded
Open this post in threaded view
|

[Micronet] Logging network performance?

Ian Crew
Hi all:

Asking this about my home--Comcast (Berkeley)--net connection, but since I telecommute a bunch, and I'm guessing that the answer might be of general interest, it doesn't seem TOO inappropriate to ask here:

I've been having trouble with my net connection at home where it works fine much of the time, but periodically seems to get really bad for a little while.  When it's "fine", I see--per speedtest.net--something in the range of 15-20ms ping times, and about 16Mbps down/4Mbps up.  When it goes screwy, I see ping times in the 250ms range, and speeds at about 1Mbps in both directions.

Of course, if I try to call Comcast, they're likely not to see anything wrong, as most of the time it's OK.

So, my question is: Does anyone know of a tool which would allow me to log the performance (bandwidth, ping times) of the network on a periodic basis over time so I can build up some data to argue with?

If it makes a difference, we're an all-Mac household (though I could fire up Windows in a VM if necessary).

Thanks!

Ian

Ian Crew
Platform and Services Manager, Research Hub
Information Services and Technology-Research and Content Technologies
University of California, Berkeley
2195 Hearst Ave, Second Floor
http://hub.berkeley.edu


 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

Jay Bryon
Hi Ian,

On the mac you can use the command line in terminal to ping with
intervals and log the results to a text file.

Open a terminal window and use ping -i [number of seconds between pings]
-c [number of pings to send] [target ip address] > pinglogfile.txt[or
whatever text file you want to log the results to].

example: Ping berkeley.edu (169.229.131.81) once a minute, for 8 hours
(480 minutes) then append the output into file test.txt (in the current
directory, generally your user home folder).  I like to use append(>>),
as it doesn't erase the file like a standard redirect (>) does.

ping -i 60 -c 480 169.229.131.81 >> text.txt

There are more options, like setting a larger packet size, or only
printing output when criteria are met etc, if you use

man ping

at the terminal prompt there's a vaguely readable list of options.

Cable modems are basically a channelized Ethernet, shared by
neighborhood.  So if there's local congestion from heavy use by one of
your neighbors on the same channel, that would explain what you're
seeing.  (But it's been a while since I've read up on it since we don't
use it here, so YMMV).

-J


On 7/29/11 10:47 AM, Ian Crew wrote:

> Hi all:
>
> Asking this about my home--Comcast (Berkeley)--net connection, but since I telecommute a bunch, and I'm guessing that the answer might be of general interest, it doesn't seem TOO inappropriate to ask here:
>
> I've been having trouble with my net connection at home where it works fine much of the time, but periodically seems to get really bad for a little while.  When it's "fine", I see--per speedtest.net--something in the range of 15-20ms ping times, and about 16Mbps down/4Mbps up.  When it goes screwy, I see ping times in the 250ms range, and speeds at about 1Mbps in both directions.
>
> Of course, if I try to call Comcast, they're likely not to see anything wrong, as most of the time it's OK.
>
> So, my question is: Does anyone know of a tool which would allow me to log the performance (bandwidth, ping times) of the network on a periodic basis over time so I can build up some data to argue with?
>
> If it makes a difference, we're an all-Mac household (though I could fire up Windows in a VM if necessary).
>
> Thanks!
>
> Ian
>
> Ian Crew
> Platform and Services Manager, Research Hub
> Information Services and Technology-Research and Content Technologies
> University of California, Berkeley
> 2195 Hearst Ave, Second Floor
> http://hub.berkeley.edu
>
>
>
> -------------------------------------------------------------------------
> The following was automatically added to this message by the list server:
>
> To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:
>
> http://micronet.berkeley.edu
>
> Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
>

--
Jay Bryon
Senior Network Engineer, U.C. Berkeley/IST/IS/Network Operations and Services
[hidden email]
2-5636

"Next generation Dragon... will use the same escape engines for on target propulsive landing... This enables landing on any solid surface in the solar system."
-SpaceX on Crew Development Program currently building hardware.

"The person who says it cannot be done should not interrupt the person doing it"
-Chinese proverb

[Unless stated explicitly otherwise, all opinions are my own and do not represent official policy of any part of IST, U.C. Berkeley or the U.C. Regents]


 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

vanngogh
Hi,

I've seen the same events with my home connection as well Ian, but
haven't encountered it in awhile.  Many months ago, I had a period over
a couple of days where the speed would literally jump up and down over
a short interval.

As Jay mentioned, I know it used to be a problem years ago with cable
modems and if your neighbors were using heavy bandwith, you'd see the
effects.

I haven't personally used it, but I hear good things about Wireshark:

http://www.wireshark.org/

Thanks,

Anthony Vann
Systems Administrator
IST - Campus Technology Services
UC Berkeley

> Hi Ian,
>
> On the mac you can use the command line in terminal to ping with
> intervals and log the results to a text file.
>
> Open a terminal window and use ping -i [number of seconds between pings]
> -c [number of pings to send] [target ip address] > pinglogfile.txt[or
> whatever text file you want to log the results to].
>
> example: Ping berkeley.edu (169.229.131.81) once a minute, for 8 hours
> (480 minutes) then append the output into file test.txt (in the current
> directory, generally your user home folder).  I like to use append(>>),
> as it doesn't erase the file like a standard redirect (>) does.
>
> ping -i 60 -c 480 169.229.131.81 >> text.txt
>
> There are more options, like setting a larger packet size, or only
> printing output when criteria are met etc, if you use
>
> man ping
>
> at the terminal prompt there's a vaguely readable list of options.
>
> Cable modems are basically a channelized Ethernet, shared by
> neighborhood.  So if there's local congestion from heavy use by one of
> your neighbors on the same channel, that would explain what you're
> seeing.  (But it's been a while since I've read up on it since we don't
> use it here, so YMMV).
>
> -J
>
>
> On 7/29/11 10:47 AM, Ian Crew wrote:
>> Hi all:
>>
>> Asking this about my home--Comcast (Berkeley)--net connection, but since
>> I telecommute a bunch, and I'm guessing that the answer might be of
>> general interest, it doesn't seem TOO inappropriate to ask here:
>>
>> I've been having trouble with my net connection at home where it works
>> fine much of the time, but periodically seems to get really bad for a
>> little while.  When it's "fine", I see--per speedtest.net--something in
>> the range of 15-20ms ping times, and about 16Mbps down/4Mbps up.  When
>> it goes screwy, I see ping times in the 250ms range, and speeds at about
>> 1Mbps in both directions.
>>
>> Of course, if I try to call Comcast, they're likely not to see anything
>> wrong, as most of the time it's OK.
>>
>> So, my question is: Does anyone know of a tool which would allow me to
>> log the performance (bandwidth, ping times) of the network on a periodic
>> basis over time so I can build up some data to argue with?
>>
>> If it makes a difference, we're an all-Mac household (though I could
>> fire up Windows in a VM if necessary).
>>
>> Thanks!
>>
>> Ian
>>
>> Ian Crew
>> Platform and Services Manager, Research Hub
>> Information Services and Technology-Research and Content Technologies
>> University of California, Berkeley
>> 2195 Hearst Ave, Second Floor
>> http://hub.berkeley.edu
>>
>>
>>
>> -------------------------------------------------------------------------
>> The following was automatically added to this message by the list
>> server:
>>
>> To learn more about Micronet, including how to subscribe to or
>> unsubscribe from its mailing list and how to find out about upcoming
>> meetings, please visit the Micronet Web site:
>>
>> http://micronet.berkeley.edu
>>
>> Messages you send to this mailing list are public and world-viewable,
>> and the list's archives can be browsed and searched on the Internet.
>> This means these messages can be viewed by (among others) your bosses,
>> prospective employers, and people who have known you in the past.
>>
>
> --
> Jay Bryon
> Senior Network Engineer, U.C. Berkeley/IST/IS/Network Operations and
> Services
> [hidden email]
> 2-5636
>
> "Next generation Dragon... will use the same escape engines for on target
> propulsive landing... This enables landing on any solid surface in the
> solar system."
> -SpaceX on Crew Development Program currently building hardware.
>
> "The person who says it cannot be done should not interrupt the person
> doing it"
> -Chinese proverb
>
> [Unless stated explicitly otherwise, all opinions are my own and do not
> represent official policy of any part of IST, U.C. Berkeley or the U.C.
> Regents]
>
>
>
> -------------------------------------------------------------------------
> The following was automatically added to this message by the list server:
>
> To learn more about Micronet, including how to subscribe to or unsubscribe
> from its mailing list and how to find out about upcoming meetings, please
> visit the Micronet Web site:
>
> http://micronet.berkeley.edu
>
> Messages you send to this mailing list are public and world-viewable, and
> the list's archives can be browsed and searched on the Internet.  This
> means these messages can be viewed by (among others) your bosses,
> prospective employers, and people who have known you in the past.
>



 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

Graham Patterson
In reply to this post by Jay Bryon

I'd probably add a traceroute into the mix. With the number of hops and
varied services involved, a simple ping wouldn't give you much ammunition.

My service goes to Pinole, Oakland, Sacramento, San Jose, Cenic, and
then Berkeley - at least as far as the names of the routing stages are
concerned. Typically the first 10ms is getting from home to Pinole. The
remaining 7ms covers the remaining 15 hops. My critical stuff is wired,
so local WiFi conditions are not an issue.

On a few occasions I have had the DNS service fall back on Denver, which
slows up web access a little!

I try to avoid prime Netflix viewing times if I have to do heavy remote
access 8-(

For comparison, a connection to a British university is around 168ms,
and around 150ms of that is the trans-atlantic hop.

Graham

On 7/29/2011 12:33 PM, Jay Bryon wrote:

> Hi Ian,
>
> On the mac you can use the command line in terminal to ping with
> intervals and log the results to a text file.
>
> Open a terminal window and use ping -i [number of seconds between pings]
> -c [number of pings to send] [target ip address]>  pinglogfile.txt[or
> whatever text file you want to log the results to].
>
> example: Ping berkeley.edu (169.229.131.81) once a minute, for 8 hours
> (480 minutes) then append the output into file test.txt (in the current
> directory, generally your user home folder).  I like to use append(>>),
> as it doesn't erase the file like a standard redirect (>) does.
>
> ping -i 60 -c 480 169.229.131.81>>  text.txt
>
> There are more options, like setting a larger packet size, or only
> printing output when criteria are met etc, if you use
>
> man ping
>
> at the terminal prompt there's a vaguely readable list of options.
>
> Cable modems are basically a channelized Ethernet, shared by
> neighborhood.  So if there's local congestion from heavy use by one of
> your neighbors on the same channel, that would explain what you're
> seeing.  (But it's been a while since I've read up on it since we don't
> use it here, so YMMV).
>
> -J
>
>
> On 7/29/11 10:47 AM, Ian Crew wrote:
>> Hi all:
>>
>> Asking this about my home--Comcast (Berkeley)--net connection, but since I telecommute a bunch, and I'm guessing that the answer might be of general interest, it doesn't seem TOO inappropriate to ask here:
>>
>> I've been having trouble with my net connection at home where it works fine much of the time, but periodically seems to get really bad for a little while.  When it's "fine", I see--per speedtest.net--something in the range of 15-20ms ping times, and about 16Mbps down/4Mbps up.  When it goes screwy, I see ping times in the 250ms range, and speeds at about 1Mbps in both directions.
>>
>> Of course, if I try to call Comcast, they're likely not to see anything wrong, as most of the time it's OK.
>>
>> So, my question is: Does anyone know of a tool which would allow me to log the performance (bandwidth, ping times) of the network on a periodic basis over time so I can build up some data to argue with?
>>
>> If it makes a difference, we're an all-Mac household (though I could fire up Windows in a VM if necessary).
>>
>> Thanks!
>>
>> Ian
>>
>> Ian Crew
>> Platform and Services Manager, Research Hub
>> Information Services and Technology-Research and Content Technologies
>> University of California, Berkeley
>> 2195 Hearst Ave, Second Floor
>> http://hub.berkeley.edu
>>
>>
>>
>> -------------------------------------------------------------------------
>> The following was automatically added to this message by the list server:
>>
>> To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:
>>
>> http://micronet.berkeley.edu
>>
>> Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
>>
>

--
Graham Patterson, System Administrator
Lawrence Hall of Science, UC Berkeley

 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

Michael Sinatra


On Fri, 29 Jul 2011, Graham Patterson wrote:

>
> I'd probably add a traceroute into the mix. With the number of hops and
> varied services involved, a simple ping wouldn't give you much ammunition.
>
> My service goes to Pinole, Oakland, Sacramento, San Jose, Cenic, and
> then Berkeley - at least as far as the names of the routing stages are
> concerned. Typically the first 10ms is getting from home to Pinole. The
> remaining 7ms covers the remaining 15 hops. My critical stuff is wired,
> so local WiFi conditions are not an issue.
>
> On a few occasions I have had the DNS service fall back on Denver, which
> slows up web access a little!
>
> I try to avoid prime Netflix viewing times if I have to do heavy remote
> access 8-(
>
> For comparison, a connection to a British university is around 168ms,
> and around 150ms of that is the trans-atlantic hop.

WARNING WARNING WARNING:

The round-trip-times that traceroute gives you are based on the receipt of
ICMP ttl-exceeded messages that each hop router deigns to send you.
Routers are designed to forward packets; they are generally not designed
to generate and return ICMP messages.  The router will happily punt such
activity to the slowest process-switched path it can find and you will
wait for the ICMP message to come back when the router gets around to
sending it to you.  Therefore the RTT time you see from "pinole" is the
combination of network path delay in both directions, plus a factor of how
busy the *control plane* (not necessarily the forwarding plane) of
the router itself is.

The point is, you should not infer network path delay characteristics from
traceroute RTT output, although traceroute is an otherwise very useful
tool.

As for the original problem being reported, as part of the logging
process, it would also be useful to understand what was happening at the
time that you noticed the slowness.  Specifically, was anyone else in the
household doing a bandwidth-intensive activity?  Downloading files, etc.?
TCP should be able to deal with congestion, but such abilities are
hampered by bufferbloat.  Simply put, so many devices (such as CPE
routers) try to make the Internet "faster" by buffering traffic, but this
wreaks havoc with TCP's congestion-control algorithms.  Jim Gettys's
presentation at NANOG 52 sums this up nicely, and you'll note that Jim's
symptoms are similar, but not identical to Ian's:

http://www.nanog.org/meetings/nanog52/abstracts.php?pt=MTc3MiZuYW5vZzUy&nm=nanog52

It may be worth trying Netalyzr from home and see what it says:

http://netalyzr.icsi.berkeley.edu


michael


 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

Ian Crew
In reply to this post by Graham Patterson
Thanks, all, for the suggestions below, as well as those off-list.

As Anthony Vann said, I'm seeing very transient poor performance--maybe 1 minute in 10?--but it's enough to foul up video chats or phone calls via our microcell.

One follow-up question:  I see how to monitor ping times (thanks Jay!) but is there any way to monitor bandwidth numbers, like what speedtest.net displays?  (Or is that not a good indicator of network health?)

To answer Michael Sinatra's question, no, this seems to happen regardless of what else is (or is not) going on on the network in the house, either internally or out to the internet....

Can you tell that I'm a bit of a networking novice here?  All this makes me all the more grateful for our quite-well-maintained setup here on campus!

Cheers,

Ian

On Jul 29, 2011, at 1:36 PM, Graham Patterson wrote:

>
> I'd probably add a traceroute into the mix. With the number of hops and
> varied services involved, a simple ping wouldn't give you much ammunition.
>
> My service goes to Pinole, Oakland, Sacramento, San Jose, Cenic, and
> then Berkeley - at least as far as the names of the routing stages are
> concerned. Typically the first 10ms is getting from home to Pinole. The
> remaining 7ms covers the remaining 15 hops. My critical stuff is wired,
> so local WiFi conditions are not an issue.
>
> On a few occasions I have had the DNS service fall back on Denver, which
> slows up web access a little!
>
> I try to avoid prime Netflix viewing times if I have to do heavy remote
> access 8-(
>
> For comparison, a connection to a British university is around 168ms,
> and around 150ms of that is the trans-atlantic hop.
>
> Graham
>
> On 7/29/2011 12:33 PM, Jay Bryon wrote:
>> Hi Ian,
>>
>> On the mac you can use the command line in terminal to ping with
>> intervals and log the results to a text file.
>>
>> Open a terminal window and use ping -i [number of seconds between pings]
>> -c [number of pings to send] [target ip address]>  pinglogfile.txt[or
>> whatever text file you want to log the results to].
>>
>> example: Ping berkeley.edu (169.229.131.81) once a minute, for 8 hours
>> (480 minutes) then append the output into file test.txt (in the current
>> directory, generally your user home folder).  I like to use append(>>),
>> as it doesn't erase the file like a standard redirect (>) does.
>>
>> ping -i 60 -c 480 169.229.131.81>>  text.txt
>>
>> There are more options, like setting a larger packet size, or only
>> printing output when criteria are met etc, if you use
>>
>> man ping
>>
>> at the terminal prompt there's a vaguely readable list of options.
>>
>> Cable modems are basically a channelized Ethernet, shared by
>> neighborhood.  So if there's local congestion from heavy use by one of
>> your neighbors on the same channel, that would explain what you're
>> seeing.  (But it's been a while since I've read up on it since we don't
>> use it here, so YMMV).
>>
>> -J
>>
>>
>> On 7/29/11 10:47 AM, Ian Crew wrote:
>>> Hi all:
>>>
>>> Asking this about my home--Comcast (Berkeley)--net connection, but since I telecommute a bunch, and I'm guessing that the answer might be of general interest, it doesn't seem TOO inappropriate to ask here:
>>>
>>> I've been having trouble with my net connection at home where it works fine much of the time, but periodically seems to get really bad for a little while.  When it's "fine", I see--per speedtest.net--something in the range of 15-20ms ping times, and about 16Mbps down/4Mbps up.  When it goes screwy, I see ping times in the 250ms range, and speeds at about 1Mbps in both directions.
>>>
>>> Of course, if I try to call Comcast, they're likely not to see anything wrong, as most of the time it's OK.
>>>
>>> So, my question is: Does anyone know of a tool which would allow me to log the performance (bandwidth, ping times) of the network on a periodic basis over time so I can build up some data to argue with?
>>>
>>> If it makes a difference, we're an all-Mac household (though I could fire up Windows in a VM if necessary).
>>>
>>> Thanks!
>>>
>>> Ian
>>>
>>> Ian Crew
>>> Platform and Services Manager, Research Hub
>>> Information Services and Technology-Research and Content Technologies
>>> University of California, Berkeley
>>> 2195 Hearst Ave, Second Floor
>>> http://hub.berkeley.edu
>>>
>>>
>>>
>>> -------------------------------------------------------------------------
>>> The following was automatically added to this message by the list server:
>>>
>>> To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:
>>>
>>> http://micronet.berkeley.edu
>>>
>>> Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
>>>
>>
>
> --
> Graham Patterson, System Administrator
> Lawrence Hall of Science, UC Berkeley
>
>
> -------------------------------------------------------------------------
> The following was automatically added to this message by the list server:
>
> To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:
>
> http://micronet.berkeley.edu
>
> Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.

Ian Crew
Platform and Services Manager, Research Hub
Information Services and Technology-Research and Content Technologies
University of California, Berkeley
2195 Hearst Ave, Second Floor
http://hub.berkeley.edu


 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

Mark Boolootian
In reply to this post by Michael Sinatra

> The point is, you should not infer network path delay characteristics from
> traceroute RTT output, although traceroute is an otherwise very useful
> tool.

A very useful NANOG talk on traceroute:

  http://www.nanog.org/meetings/nanog47/abstracts.php?pt=MTQ0MiZuYW5vZzQ3&nm=nanog47

 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

Graham Patterson
In reply to this post by Michael Sinatra
True.

Though traceroute has been fairly consistent over the years. When I have
had issues the traceroute output has also changed. The actual data may
be of little value.

With all the various organizations the traffic passes through, I'm just
happy it gets there at all.

A periodic file upload/download might be a better real-world test. But
then you may find out what the local bandwidth cap is like. And it
presupposes you have a server that can do SSH or something else that
avoids service provider caching.

Graham

On 7/29/2011 1:59 PM, Michael Sinatra wrote:

>
>
> On Fri, 29 Jul 2011, Graham Patterson wrote:
>
>>
>> I'd probably add a traceroute into the mix. With the number of hops and
>> varied services involved, a simple ping wouldn't give you much
>> ammunition.
>>
>> My service goes to Pinole, Oakland, Sacramento, San Jose, Cenic, and
>> then Berkeley - at least as far as the names of the routing stages are
>> concerned. Typically the first 10ms is getting from home to Pinole. The
>> remaining 7ms covers the remaining 15 hops. My critical stuff is wired,
>> so local WiFi conditions are not an issue.
>>
>> On a few occasions I have had the DNS service fall back on Denver, which
>> slows up web access a little!
>>
>> I try to avoid prime Netflix viewing times if I have to do heavy remote
>> access 8-(
>>
>> For comparison, a connection to a British university is around 168ms,
>> and around 150ms of that is the trans-atlantic hop.
>
> WARNING WARNING WARNING:
>
> The round-trip-times that traceroute gives you are based on the receipt
> of ICMP ttl-exceeded messages that each hop router deigns to send you.
> Routers are designed to forward packets; they are generally not designed
> to generate and return ICMP messages. The router will happily punt such
> activity to the slowest process-switched path it can find and you will
> wait for the ICMP message to come back when the router gets around to
> sending it to you. Therefore the RTT time you see from "pinole" is the
> combination of network path delay in both directions, plus a factor of
> how busy the *control plane* (not necessarily the forwarding plane) of
> the router itself is.
>
> The point is, you should not infer network path delay characteristics
> from traceroute RTT output, although traceroute is an otherwise very
> useful tool.
>
> As for the original problem being reported, as part of the logging
> process, it would also be useful to understand what was happening at the
> time that you noticed the slowness. Specifically, was anyone else in the
> household doing a bandwidth-intensive activity? Downloading files, etc.?
> TCP should be able to deal with congestion, but such abilities are
> hampered by bufferbloat. Simply put, so many devices (such as CPE
> routers) try to make the Internet "faster" by buffering traffic, but
> this wreaks havoc with TCP's congestion-control algorithms. Jim Gettys's
> presentation at NANOG 52 sums this up nicely, and you'll note that Jim's
> symptoms are similar, but not identical to Ian's:
>
> http://www.nanog.org/meetings/nanog52/abstracts.php?pt=MTc3MiZuYW5vZzUy&nm=nanog52
>
>
> It may be worth trying Netalyzr from home and see what it says:
>
> http://netalyzr.icsi.berkeley.edu
>
>
> michael
>

--
Graham Patterson, System Administrator
Lawrence Hall of Science, UC Berkeley

 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

Bill Clark
In reply to this post by Ian Crew
Given Michael's explanation of why ICMP response times should be taken
with a grain of salt, I'm not positive that the methodology employed by
this tool is entirely trustworthy, but it might be worth a look:

http://bwping.sourceforge.net/

-Bill Clark

> Thanks, all, for the suggestions below, as well as those off-list.
>
> As Anthony Vann said, I'm seeing very transient poor performance--maybe 1
> minute in 10?--but it's enough to foul up video chats or phone calls via
> our microcell.
>
> One follow-up question:  I see how to monitor ping times (thanks Jay!) but
> is there any way to monitor bandwidth numbers, like what speedtest.net
> displays?  (Or is that not a good indicator of network health?)
>
> To answer Michael Sinatra's question, no, this seems to happen regardless
> of what else is (or is not) going on on the network in the house, either
> internally or out to the internet....
>
> Can you tell that I'm a bit of a networking novice here?  All this makes
> me all the more grateful for our quite-well-maintained setup here on
> campus!
>
> Cheers,
>
> Ian
>
> On Jul 29, 2011, at 1:36 PM, Graham Patterson wrote:
>
>>
>> I'd probably add a traceroute into the mix. With the number of hops and
>> varied services involved, a simple ping wouldn't give you much
>> ammunition.
>>
>> My service goes to Pinole, Oakland, Sacramento, San Jose, Cenic, and
>> then Berkeley - at least as far as the names of the routing stages are
>> concerned. Typically the first 10ms is getting from home to Pinole. The
>> remaining 7ms covers the remaining 15 hops. My critical stuff is wired,
>> so local WiFi conditions are not an issue.
>>
>> On a few occasions I have had the DNS service fall back on Denver, which
>> slows up web access a little!
>>
>> I try to avoid prime Netflix viewing times if I have to do heavy remote
>> access 8-(
>>
>> For comparison, a connection to a British university is around 168ms,
>> and around 150ms of that is the trans-atlantic hop.
>>
>> Graham
>>
>> On 7/29/2011 12:33 PM, Jay Bryon wrote:
>>> Hi Ian,
>>>
>>> On the mac you can use the command line in terminal to ping with
>>> intervals and log the results to a text file.
>>>
>>> Open a terminal window and use ping -i [number of seconds between
>>> pings]
>>> -c [number of pings to send] [target ip address]>  pinglogfile.txt[or
>>> whatever text file you want to log the results to].
>>>
>>> example: Ping berkeley.edu (169.229.131.81) once a minute, for 8 hours
>>> (480 minutes) then append the output into file test.txt (in the current
>>> directory, generally your user home folder).  I like to use append(>>),
>>> as it doesn't erase the file like a standard redirect (>) does.
>>>
>>> ping -i 60 -c 480 169.229.131.81>>  text.txt
>>>
>>> There are more options, like setting a larger packet size, or only
>>> printing output when criteria are met etc, if you use
>>>
>>> man ping
>>>
>>> at the terminal prompt there's a vaguely readable list of options.
>>>
>>> Cable modems are basically a channelized Ethernet, shared by
>>> neighborhood.  So if there's local congestion from heavy use by one of
>>> your neighbors on the same channel, that would explain what you're
>>> seeing.  (But it's been a while since I've read up on it since we don't
>>> use it here, so YMMV).
>>>
>>> -J
>>>
>>>
>>> On 7/29/11 10:47 AM, Ian Crew wrote:
>>>> Hi all:
>>>>
>>>> Asking this about my home--Comcast (Berkeley)--net connection, but
>>>> since I telecommute a bunch, and I'm guessing that the answer might be
>>>> of general interest, it doesn't seem TOO inappropriate to ask here:
>>>>
>>>> I've been having trouble with my net connection at home where it works
>>>> fine much of the time, but periodically seems to get really bad for a
>>>> little while.  When it's "fine", I see--per speedtest.net--something
>>>> in the range of 15-20ms ping times, and about 16Mbps down/4Mbps up.
>>>> When it goes screwy, I see ping times in the 250ms range, and speeds
>>>> at about 1Mbps in both directions.
>>>>
>>>> Of course, if I try to call Comcast, they're likely not to see
>>>> anything wrong, as most of the time it's OK.
>>>>
>>>> So, my question is: Does anyone know of a tool which would allow me to
>>>> log the performance (bandwidth, ping times) of the network on a
>>>> periodic basis over time so I can build up some data to argue with?
>>>>
>>>> If it makes a difference, we're an all-Mac household (though I could
>>>> fire up Windows in a VM if necessary).
>>>>
>>>> Thanks!
>>>>
>>>> Ian
>>>>
>>>> Ian Crew
>>>> Platform and Services Manager, Research Hub
>>>> Information Services and Technology-Research and Content Technologies
>>>> University of California, Berkeley
>>>> 2195 Hearst Ave, Second Floor
>>>> http://hub.berkeley.edu
>>>>
>>>>
>>>>
>>>> -------------------------------------------------------------------------
>>>> The following was automatically added to this message by the list
>>>> server:
>>>>
>>>> To learn more about Micronet, including how to subscribe to or
>>>> unsubscribe from its mailing list and how to find out about upcoming
>>>> meetings, please visit the Micronet Web site:
>>>>
>>>> http://micronet.berkeley.edu
>>>>
>>>> Messages you send to this mailing list are public and world-viewable,
>>>> and the list's archives can be browsed and searched on the Internet.
>>>> This means these messages can be viewed by (among others) your bosses,
>>>> prospective employers, and people who have known you in the past.
>>>>
>>>
>>
>> --
>> Graham Patterson, System Administrator
>> Lawrence Hall of Science, UC Berkeley
>>
>>
>> -------------------------------------------------------------------------
>> The following was automatically added to this message by the list
>> server:
>>
>> To learn more about Micronet, including how to subscribe to or
>> unsubscribe from its mailing list and how to find out about upcoming
>> meetings, please visit the Micronet Web site:
>>
>> http://micronet.berkeley.edu
>>
>> Messages you send to this mailing list are public and world-viewable,
>> and the list's archives can be browsed and searched on the Internet.
>> This means these messages can be viewed by (among others) your bosses,
>> prospective employers, and people who have known you in the past.
>
> Ian Crew
> Platform and Services Manager, Research Hub
> Information Services and Technology-Research and Content Technologies
> University of California, Berkeley
> 2195 Hearst Ave, Second Floor
> http://hub.berkeley.edu
>
>
>
> -------------------------------------------------------------------------
> The following was automatically added to this message by the list server:
>
> To learn more about Micronet, including how to subscribe to or unsubscribe
> from its mailing list and how to find out about upcoming meetings, please
> visit the Micronet Web site:
>
> http://micronet.berkeley.edu
>
> Messages you send to this mailing list are public and world-viewable, and
> the list's archives can be browsed and searched on the Internet.  This
> means these messages can be viewed by (among others) your bosses,
> prospective employers, and people who have known you in the past.
>



 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.
Reply | Threaded
Open this post in threaded view
|

Re: [Micronet] Logging network performance?

Jay Bryon
In reply to this post by vanngogh
On 7/29/11 12:59 PM, Anthony Vann wrote:
>
> I haven't personally used it, but I hear good things about Wireshark:
>
> http://www.wireshark.org/
>
>
Wireshark isn't the right tool for this IMHO, it's a packet inspection
tool, but you only have visibility into traffic that traverses the
network where you're sniffing with it.  You can waste a huge amount of
time trying to parse the information it presents, only to realize that
it's not relevant.

It is very powerful at what it does, but for bandwidth measurement and
diagnosis it's not really what I would go to.  The standby for that has
been iperf, but there are also web based bandwidth testers that will do
the job.

-J

--
Jay Bryon
Senior Network Engineer, U.C. Berkeley/IST/IS/Network Operations and Services
[hidden email]
2-5636

"Next generation Dragon... will use the same escape engines for on target propulsive landing... This enables landing on any solid surface in the solar system."
-SpaceX on Crew Development Program currently building hardware.

"The person who says it cannot be done should not interrupt the person doing it"
-Chinese proverb

[Unless stated explicitly otherwise, all opinions are my own and do not represent official policy of any part of IST, U.C. Berkeley or the U.C. Regents]


 
-------------------------------------------------------------------------
The following was automatically added to this message by the list server:

To learn more about Micronet, including how to subscribe to or unsubscribe from its mailing list and how to find out about upcoming meetings, please visit the Micronet Web site:

http://micronet.berkeley.edu

Messages you send to this mailing list are public and world-viewable, and the list's archives can be browsed and searched on the Internet.  This means these messages can be viewed by (among others) your bosses, prospective employers, and people who have known you in the past.