Discussion:
[mrtg] 95th percentile?
Erik Van Riper
1999-09-24 13:41:23 UTC
Permalink
I noticed a few questions concerning 95th percentile about a week ago, but
saw no response. If someone did respond privately, can they also do so to
me? I am very interested in this capability.

Thank you in advance!

--
Erik Van Riper
Manager, Network Operations
(619) 302 0698 (Cell)
(619) 623 8400 x112 (Desk)

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
David C Prall
1999-09-24 14:13:39 UTC
Permalink
Post by Erik Van Riper
I noticed a few questions concerning 95th percentile about a week ago, but
saw no response. If someone did respond privately, can they also do so to
me? I am very interested in this capability.
What exactly are you looking for. I just use the WithPeak[]: wmy option.
This shows the high along with the averages. Giving you four lines
rather then two.

At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the standard
default configmaker script. And at that point WithPeak was implemented,
so I used it.

Hope this helps,

David C Prall, MCNE MCSE DCP Technologies
***@dcptech.com http://www.dcptech.com
***@dcptech.com


--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Erik Van Riper
1999-09-24 14:22:13 UTC
Permalink
Well, I guess that is a good question. I am interested in figuring out what
my bill from my different ISP's will be, since I am charged at the 95th
percentile. Since I am in multiple ISP's, and I have different contracts,
some bandwidth is more expensive than others. I can direct traffic to other
locations if things start to get too expensive. The "WithPeak" seems to
give me a peak, but not exactly what I am looking for unless I am just being
stupid this morning and not fully understanding the implications of it......

--
Erik Van Riper
Manager, Network Operations
(619) 302 0698 (Cell)
(619) 623 8400 x112 (Desk)



-----Original Message-----
From: David C Prall [mailto:***@dcptech.com]
Sent: Friday, September 24, 1999 7:14 AM
To: ***@list.ee.ethz.ch
Subject: [mrtg] Re: 95th percentile?
Post by Erik Van Riper
I noticed a few questions concerning 95th percentile about a week ago,
but
Post by Erik Van Riper
saw no response. If someone did respond privately, can they also do
so to
Post by Erik Van Riper
me? I am very interested in this capability.
What exactly are you looking for. I just use the WithPeak[]: wmy option.
This shows the high along with the averages. Giving you four lines
rather then two.

At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the standard
default configmaker script. And at that point WithPeak was implemented,
so I used it.

Hope this helps,

David C Prall, MCNE MCSE DCP Technologies
***@dcptech.com http://www.dcptech.com
***@dcptech.com


--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
George D. Nincehelser
1999-09-24 14:49:45 UTC
Permalink
Erik is correct. I'm looking for the 95th percentile parser for that same
reason. I've been trying to download it from null0.qual.net, but I can
never get connected.

WithPeak doesn't quite give that information.

George
----- Original Message -----
From: Erik Van Riper <***@intervu.net>
To: <***@list.ee.ethz.ch>
Sent: Friday, September 24, 1999 9:22 AM
Subject: [mrtg] Re: 95th percentile?
Post by Erik Van Riper
Well, I guess that is a good question. I am interested in figuring out what
my bill from my different ISP's will be, since I am charged at the 95th
percentile. Since I am in multiple ISP's, and I have different contracts,
some bandwidth is more expensive than others. I can direct traffic to other
locations if things start to get too expensive. The "WithPeak" seems to
give me a peak, but not exactly what I am looking for unless I am just being
stupid this morning and not fully understanding the implications of it......
--
Erik Van Riper
Manager, Network Operations
(619) 302 0698 (Cell)
(619) 623 8400 x112 (Desk)
-----Original Message-----
Sent: Friday, September 24, 1999 7:14 AM
Subject: [mrtg] Re: 95th percentile?
Post by Erik Van Riper
I noticed a few questions concerning 95th percentile about a week ago,
but
Post by Erik Van Riper
saw no response. If someone did respond privately, can they also do
so to
Post by Erik Van Riper
me? I am very interested in this capability.
What exactly are you looking for. I just use the WithPeak[]: wmy option.
This shows the high along with the averages. Giving you four lines
rather then two.
At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the standard
default configmaker script. And at that point WithPeak was implemented,
so I used it.
Hope this helps,
David C Prall, MCNE MCSE DCP Technologies
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Pete Templin
1999-09-24 15:04:05 UTC
Permalink
Post by David C Prall
What exactly are you looking for. I just use the WithPeak[]: wmy option.
This shows the high along with the averages. Giving you four lines
rather then two.
At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the standard
default configmaker script. And at that point WithPeak was implemented,
so I used it.
Most folks looking for 95th percentile want to be able to calculate the
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days, and
then it only stores averaged averages. It's impossible to recreate the
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour period.
There's no way to determine how many five minute samples were that high.

Some people have written scripts that archive a day's worth of five-minute
samples each day, and then crank through that unaveraged data at the end
of the month. The few folks I've talked to about this usually wrote it on
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales well
(or can be reengineered to scale).

Pete

--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator

On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
2414 Babcock Rd. Suite 106 ***@urdirect.net
San Antonio, TX 78229 (210)692-9911

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
David C Prall
1999-09-24 17:17:19 UTC
Permalink
Post by Pete Templin
Most folks looking for 95th percentile want to be able to calculate the
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days, and
then it only stores averaged averages. It's impossible to recreate the
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour period.
There's no way to determine how many five minute samples were that high.
Some people have written scripts that archive a day's worth of
five-minute
Post by Pete Templin
samples each day, and then crank through that unaveraged data at the end
of the month. The few folks I've talked to about this usually wrote it on
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales well
(or can be reengineered to scale).
Thanks for the excellent example. So I went back and took a look at
above.net which supposedly had the 95th Percentile patched and found it
actually listed on the Monthly graph, along with a Red Line across at
this value. http://stats.sjc.above.net/traffic/

So what I am to understand is that they take this number and charge a
higher rate for all traffic above and beyond this value? So if I decide
to copy the entire Windows NT i386 directory over the wire to another
server at night, I am going to get charged an additional amount for this
half-hour burst of abnormal traffic? Since it wasn't critical it would
probably be cheaper to send the CD and have the local admin insert it
into the CD-ROM drive instead?

I don't really understand this, all that well. I've never had to
provision major connections beyond a T-1 to an ISP that was X amount a
month. No funny stuff in the contract that I saw.

David C Prall, MCNE MCSE DCP Technologies
***@dcptech.com http://www.dcptech.com
***@dcptech.com


--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Chairman of the Bored
1999-09-24 17:13:17 UTC
Permalink
greetings,

i'm building an mrtg script archive to make life easier for all of us :)

could those of you who wish to contribute please
e mail your scripts to ***@L7.org

i will be going through them and changing the

[router.L7.org] ***@207.12.41.254
to
[routername] ***@0.0.0.0

just to keep some enterprising individuals honest..

thanks,
-dd

\\\\\//
\\|// _\\|//_ | | _\\|//_ \\|//
(@ @) (' 0-0 ') (.) (.) (' @-@ ') (o-o)
+-=oOOo-(_)-oOOo=oo0=(_)=0oo=oOO=-(_)-=OOo=oo0=(_)=0oo=oOOo-(_)-oOOo=-+
Plazma Networking Services / Level Seven inc.
Connecting the World....
http://www.plazma.net http://www.L7.net http://www.L7.org /"\
Olympia's "one stop" InterNetworking Provider 1 (360) 357 - 7315 \ /
+-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-+ X
ASCII Ribbon campaign against HTML E-Mail >- - - - - - - - - - - - - -> / \

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
George D. Nincehelser
1999-09-24 15:53:07 UTC
Permalink
That's exactly what I've been looking for.

I assume that there is a way an external script can gather the data and
calculate the 95th percentile, then pass that info in some way to MRTG so
it can be graphed?

I'm unclear as to the best way to graph the 95th percentile data. I'm
thinking that a simple straight line across the graph would be sufficient
for billing purposes, but a trending line might look cool. I imagine the
line would best be reset at the end of each billing period.

Thoughts?
George
Post by Erik Van Riper
-----Original Message-----
Behalf Of Pete Templin
Sent: Friday, September 24, 1999 10:04 AM
To: David C Prall
Subject: [mrtg] Re: 95th percentile?
Post by David C Prall
What exactly are you looking for. I just use the WithPeak[]: wmy option.
This shows the high along with the averages. Giving you four lines
rather then two.
At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the standard
default configmaker script. And at that point WithPeak was implemented,
so I used it.
Most folks looking for 95th percentile want to be able to calculate the
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days, and
then it only stores averaged averages. It's impossible to recreate the
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour period.
There's no way to determine how many five minute samples were that high.
Some people have written scripts that archive a day's worth of five-minute
samples each day, and then crank through that unaveraged data at the end
of the month. The few folks I've talked to about this usually wrote it on
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales well
(or can be reengineered to scale).
Pete
--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator
On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
San Antonio, TX 78229 (210)692-9911
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Ken Jones
1999-09-24 15:59:33 UTC
Permalink
I am also intereted in 95 percentile graphing for billing purposes.

Has anyone found code to do this?

Ken Jones
Inter7
Post by George D. Nincehelser
That's exactly what I've been looking for.
I assume that there is a way an external script can gather the data and
calculate the 95th percentile, then pass that info in some way to MRTG so
it can be graphed?
I'm unclear as to the best way to graph the 95th percentile data. I'm
thinking that a simple straight line across the graph would be sufficient
for billing purposes, but a trending line might look cool. I imagine the
line would best be reset at the end of each billing period.
Thoughts?
George
Post by Erik Van Riper
-----Original Message-----
Behalf Of Pete Templin
Sent: Friday, September 24, 1999 10:04 AM
To: David C Prall
Subject: [mrtg] Re: 95th percentile?
Post by David C Prall
What exactly are you looking for. I just use the WithPeak[]: wmy option.
This shows the high along with the averages. Giving you four lines
rather then two.
At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the standard
default configmaker script. And at that point WithPeak was implemented,
so I used it.
Most folks looking for 95th percentile want to be able to calculate the
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days, and
then it only stores averaged averages. It's impossible to recreate the
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour period.
There's no way to determine how many five minute samples were that high.
Some people have written scripts that archive a day's worth of five-minute
samples each day, and then crank through that unaveraged data at the end
of the month. The few folks I've talked to about this usually wrote it on
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales well
(or can be reengineered to scale).
Pete
--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator
On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
San Antonio, TX 78229 (210)692-9911
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
George D. Nincehelser
1999-09-24 15:57:59 UTC
Permalink
That's exactly what I've been looking for.

I assume that there is a way an external script can gather the data and
calculate the 95th percentile, then pass that info in some way to MRTG so
it can be graphed?

I'm unclear as to the best way to graph the 95th percentile data. I'm
thinking that a simple straight line across the graph would be sufficient
for billing purposes, but a trending line might look cool. I imagine the
line would best be reset at the end of each billing period.

Thoughts?
George
Post by Pete Templin
Most folks looking for 95th percentile want to be able to calculate the
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days, and
then it only stores averaged averages. It's impossible to recreate the
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour period.
There's no way to determine how many five minute samples were that high.
Some people have written scripts that archive a day's worth of five-minute
samples each day, and then crank through that unaveraged data at the end
of the month. The few folks I've talked to about this usually wrote it on
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales well
(or can be reengineered to scale).
Pete
--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator
On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
San Antonio, TX 78229 (210)692-9911
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Erik Van Riper
1999-09-24 16:08:13 UTC
Permalink
There is a perl module that I am looking at (Statistics::Descriptive) that
does this type of computation, the issue is mostly how to interpret the data
correctly from the MRTG log files. I will try to look at this during the
weekend, however I am not a math person so this may take me longer than
others may be able to figure out.

If anyone has ever manipulated the log files generated by MRTG in the past
with perl that could provide some insight, I would be most appreciative.

--
Erik Van Riper
Manager, Network Operations
(619) 302 0698 (Cell)
(619) 623 8400 x112 (Desk)



-----Original Message-----
From: Ken Jones [mailto:***@inter7.com]
Sent: Friday, September 24, 1999 9:00 AM
To: George D. Nincehelser
Cc: Pete Templin; David C Prall; ***@list.ee.ethz.ch
Subject: [mrtg] Re: 95th percentile?



I am also intereted in 95 percentile graphing for billing purposes.

Has anyone found code to do this?

Ken Jones
Inter7
Post by George D. Nincehelser
That's exactly what I've been looking for.
I assume that there is a way an external script can gather the data and
calculate the 95th percentile, then pass that info in some way to MRTG so
it can be graphed?
I'm unclear as to the best way to graph the 95th percentile data. I'm
thinking that a simple straight line across the graph would be sufficient
for billing purposes, but a trending line might look cool. I imagine the
line would best be reset at the end of each billing period.
Thoughts?
George
Post by Erik Van Riper
-----Original Message-----
Behalf Of Pete Templin
Sent: Friday, September 24, 1999 10:04 AM
To: David C Prall
Subject: [mrtg] Re: 95th percentile?
Post by David C Prall
What exactly are you looking for. I just use the WithPeak[]: wmy
option.
Post by George D. Nincehelser
Post by Erik Van Riper
Post by David C Prall
This shows the high along with the averages. Giving you four lines
rather then two.
At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the
standard
Post by George D. Nincehelser
Post by Erik Van Riper
Post by David C Prall
default configmaker script. And at that point WithPeak was
implemented,
Post by George D. Nincehelser
Post by Erik Van Riper
Post by David C Prall
so I used it.
Most folks looking for 95th percentile want to be able to calculate the
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days,
and
Post by George D. Nincehelser
Post by Erik Van Riper
then it only stores averaged averages. It's impossible to recreate the
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour period.
There's no way to determine how many five minute samples were that high.
Some people have written scripts that archive a day's worth of
five-minute
Post by George D. Nincehelser
Post by Erik Van Riper
samples each day, and then crank through that unaveraged data at the end
of the month. The few folks I've talked to about this usually wrote it
on
Post by George D. Nincehelser
Post by Erik Van Riper
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales well
(or can be reengineered to scale).
Pete
--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator
On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
San Antonio, TX 78229 (210)692-9911
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Landa, Brian
1999-09-24 17:23:55 UTC
Permalink
I would like to see a full config of this....

Brian Landa
Trend Consulting Services Inc.
Centre Plaza North, Suite #201
Eastlake, OH 44095
T: 440.942.4040 X243
F: 440.942.4848
Pager: 440.303.5426
Email: ***@trendcs.com <mailto:***@trendcs.com>
Network Support Specialist
WE make the NET * work




-----Original Message-----
From: David C Prall [mailto:***@dcptech.com]
Sent: Friday, September 24, 1999 1:17 PM
To: ***@list.ee.ethz.ch
Subject: [mrtg] Re: 95th percentile?
Post by Pete Templin
Most folks looking for 95th percentile want to be able to calculate
the
Post by Pete Templin
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days,
and
Post by Pete Templin
then it only stores averaged averages. It's impossible to recreate
the
Post by Pete Templin
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour
period.
Post by Pete Templin
There's no way to determine how many five minute samples were that
high.
Post by Pete Templin
Some people have written scripts that archive a day's worth of
five-minute
Post by Pete Templin
samples each day, and then crank through that unaveraged data at the
end
it on
Post by Pete Templin
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales
well
Post by Pete Templin
(or can be reengineered to scale).
Thanks for the excellent example. So I went back and took a look at
above.net which supposedly had the 95th Percentile patched and found it
actually listed on the Monthly graph, along with a Red Line across at
this value. http://stats.sjc.above.net/traffic/

So what I am to understand is that they take this number and charge a
higher rate for all traffic above and beyond this value? So if I decide
to copy the entire Windows NT i386 directory over the wire to another
server at night, I am going to get charged an additional amount for this
half-hour burst of abnormal traffic? Since it wasn't critical it would
probably be cheaper to send the CD and have the local admin insert it
into the CD-ROM drive instead?

I don't really understand this, all that well. I've never had to
provision major connections beyond a T-1 to an ISP that was X amount a
month. No funny stuff in the contract that I saw.

David C Prall, MCNE MCSE DCP Technologies
***@dcptech.com http://www.dcptech.com
***@dcptech.com


--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Henry Steinhauer
1999-09-24 17:10:55 UTC
Permalink
ALL -

It is easy to use Excel to bring in a Log file and then create the
information you are looking for.

Some ISPs use the value in a number of ways.

It is possible to have different bandwidths for each direction of a
connection, thus the 'normal' way of defining this does not work for all
people.

Normal :

1) - taking the highest value of IN or OUT for each interval monitored and
then reporting the 95th percentile level for the "billing interval'

a - If you are using difrerent bandwidths, this does not work.
--- You would need to keep IN and OUT seperate and create it
seperate.

b - each ISP has the option to create a different billing interval for
their own purpose
-- ie - have bills going out each business day so the cash flow is
not just one lump sum.
-- some on the 5th, some of the 10th, some on the 15th, some on the
20th, some on the 25th. (all months have these dates).

2) - Another approach is to use the 95% level for Each Day of the MRTG
cycles.
-- Since you are gathering this information on a 5 minute interval,
this amounts to 12 per hour or 288 per day.
-- 95% level would then be at the 273.6 value - you could round up
to the 274th or take the 273rd.
For each day then, you could replace the 'peak' value on the follow on
charts with this value for the monthly and yearly charts.

This would give you a better feel for how your actual capacity and
performance numbers are going rather than just your billing cycles.
After all the key is how do you service your users.

This would be higher than the ISP value on most days, but in general,
you would have a good feel for where you would be on their price sheet and
what you would need to do to keep below certain values.



Any Perl hacks out there ?

I would think that a program that would read the log file daily and then
replace the 'peak' values with the 95th value for the monthly and yearly
sections of the log file would do the trick. My Perl coding is still not
up to speed, but I have outlined the approach.

I would be willing to work with someone if they want to do the Perl hack as
a sample program.

There is already a sample program to read a set of CFG files so you could
know which targets need to be gathered, or you could require that the list
of log files be put in a file so one could give a list of log files rather
than the entire directory of log files. (I do not need this for all the log
files). BUT if they really want it for all of them, then just put the
while list in the file to be read.

Suggestions and comments?

Please direct them to Henry Steinhauer ***@hewitt.com



--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
George D. Nincehelser
1999-09-24 17:59:31 UTC
Permalink
The way it has been explained to me is:

1) Record the rate every 10 minutes around the clock for the billing period
2) Sort the recorded rates where highest is at the top, lowest at the bottom
3) Knock off the top 5 percent of the list
4) The resulting top reading is the rate upon which you are billed

So, if you had an abormally high transfer for a short period, it will most
likely be lopped off the top of the list and won't matter.

In a contract I'm not working on, we pay a base rate, if our 95th percentile
calculation exceeds this rate, then we pay more for that billing period
based on their pricing scale (their billing scale is in 0.5 Mbps
increments).

If anyone has any corrections to this, please let me know...I'm in the
process of negotiating this contract and I'm not yet sure I fully understand
how the formula works and its implications. (I've yet to have anyone give
me the measurement process in writing, which really bugs me)

One thing I'm wondering about is their sampling rate. How would the 95th
percentile change with different sampling rates and different traffic
patterns? To take things to the extreme, suppose I sychronized myself to
their sampling schedule, loaded up the connection when they "weren't
looking", then choked back my transfer rate below my base rate during they
times they would be sampling.

Devious, I know. I'd never actually *D0* anything like this, of
course...I'm just curious if it would work and if anyone as tried it. ;)

George
Post by David C Prall
Thanks for the excellent example. So I went back and took a look at
above.net which supposedly had the 95th Percentile patched and found it
actually listed on the Monthly graph, along with a Red Line across at
this value. http://stats.sjc.above.net/traffic/
So what I am to understand is that they take this number and charge a
higher rate for all traffic above and beyond this value? So if I decide
to copy the entire Windows NT i386 directory over the wire to another
server at night, I am going to get charged an additional amount for this
half-hour burst of abnormal traffic? Since it wasn't critical it would
probably be cheaper to send the CD and have the local admin insert it
into the CD-ROM drive instead?
I don't really understand this, all that well. I've never had to
provision major connections beyond a T-1 to an ISP that was X amount a
month. No funny stuff in the contract that I saw.
David C Prall, MCNE MCSE DCP Technologies
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Tim Kennedy
1999-09-24 20:47:58 UTC
Permalink
there's a script around somewhere called "calculate_traffic.pl".
It takes a begining date and an ending date, and tells you how much
traffic passed in that time period, and you can do the math with your CIR
or allowed rate, and see what percentile you are in for the month.

I lost the script though. Somebody here ought to have it.

-Tim
Post by George D. Nincehelser
That's exactly what I've been looking for.
I assume that there is a way an external script can gather the data and
calculate the 95th percentile, then pass that info in some way to MRTG so
it can be graphed?
I'm unclear as to the best way to graph the 95th percentile data. I'm
thinking that a simple straight line across the graph would be sufficient
for billing purposes, but a trending line might look cool. I imagine the
line would best be reset at the end of each billing period.
Thoughts?
George
Timothy Kennedy
sugarat.net, Network Management Resources
http://www.sugarat.net

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Chairman of the Bored
1999-09-28 17:39:15 UTC
Permalink
I have a bunch of snmpwalk data from a device of mine (managed hub) and i
can't make heads or tails of it, let alone turn it into useful information.

if someone could help me with this i would greatly appreciate it.

also, anyone have a script for the cisco 675?
thanks,
-dd

\\\\\//
\\|// _\\|//_ | | _\\|//_ \\|//
(@ @) (' 0-0 ') (.) (.) (' @-@ ') (o-o)
+-=oOOo-(_)-oOOo=oo0=(_)=0oo=oOO=-(_)-=OOo=oo0=(_)=0oo=oOOo-(_)-oOOo=-+
Plazma Networking Services / Level Seven inc.
Connecting the World....
http://www.plazma.net http://www.L7.net http://www.L7.org /"\
Olympia's "one stop" InterNetworking Provider 1 (360) 357 - 7315 \ /
+-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-+ X
ASCII Ribbon campaign against HTML E-Mail >- - - - - - - - - - - - - -> / \

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
David C Prall
1999-09-28 23:05:16 UTC
Permalink
Post by Chairman of the Bored
I have a bunch of snmpwalk data from a device of mine (managed hub) and i
can't make heads or tails of it, let alone turn it into useful
information.
Post by Chairman of the Bored
if someone could help me with this i would greatly appreciate it.
also, anyone have a script for the cisco 675?
thanks,
-dd
Do you have the MIB files. Do a search for GETIF and install it on a
Windows machine.

Have fun,

David C Prall, MCNE MCSE DCP Technologies
***@dcptech.com Alexandria, VA
***@dcptech.com http://www.dcptech.com


--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Pete Templin
1999-09-24 20:59:37 UTC
Permalink
Post by Tim Kennedy
there's a script around somewhere called "calculate_traffic.pl".
It takes a begining date and an ending date, and tells you how much
traffic passed in that time period, and you can do the math with your CIR
or allowed rate, and see what percentile you are in for the month.
But if you're taking that from an ordinary MRTG log file, you've got
inaccurate data to calculate from (or data of differing accuracy). If you
want the 446th highest five-minute sample (95th percentile of 12 samples
per hour times 24 hours per day times 31 days in the month), you need to
do something different, as in archive daily data to a file so it doesn't
get averaged.

Pete

--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator

On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
2414 Babcock Rd. Suite 106 ***@urdirect.net
San Antonio, TX 78229 (210)692-9911

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Brian Pavane
1999-09-26 21:23:39 UTC
Permalink
I don't think the 95th percentile information really needs to be graphed.
Meerly a program that could pull the readings every 5 minutes from a piece
of equipment (say... a cisco), and then via cron at the end of the month
spit out a report would be sufficient. This type of program doesn't need
to be tied to MRTG, however, being that MRTG is already pulling this data,
it would be best to find a way to reuse what MRTG pulls.
Post by George D. Nincehelser
That's exactly what I've been looking for.
I assume that there is a way an external script can gather the data and
calculate the 95th percentile, then pass that info in some way to MRTG so
it can be graphed?
I'm unclear as to the best way to graph the 95th percentile data. I'm
thinking that a simple straight line across the graph would be sufficient
for billing purposes, but a trending line might look cool. I imagine the
line would best be reset at the end of each billing period.
Thoughts?
George
Post by Erik Van Riper
-----Original Message-----
Behalf Of Pete Templin
Sent: Friday, September 24, 1999 10:04 AM
To: David C Prall
Subject: [mrtg] Re: 95th percentile?
Post by David C Prall
What exactly are you looking for. I just use the WithPeak[]: wmy option.
This shows the high along with the averages. Giving you four lines
rather then two.
At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the standard
default configmaker script. And at that point WithPeak was implemented,
so I used it.
Most folks looking for 95th percentile want to be able to calculate the
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days, and
then it only stores averaged averages. It's impossible to recreate the
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour period.
There's no way to determine how many five minute samples were that high.
Some people have written scripts that archive a day's worth of five-minute
samples each day, and then crank through that unaveraged data at the end
of the month. The few folks I've talked to about this usually wrote it on
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales well
(or can be reengineered to scale).
Pete
--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator
On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
San Antonio, TX 78229 (210)692-9911
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
George D. Nincehelser
1999-09-26 23:05:32 UTC
Permalink
You don't *have* to graph it, but doing so should sure make make it a lot
easier to explain to your boss or customer (or maybe yourself) what it
means. (e.g. "If this little line (traffic) goes higher than that little
line (95th percentile) a little to often, open up your wallet.)

Since the 95th percentile figure translates directly into $$$, everyone one
(including the non-technical types) are really concerned with what the value
is and where it is going. It needs to be expressed quickly and clearly, and
putting it on the graph seems to me a good way to do it. Having this info
would give everyone viewing them a clear view of our costs and ideas on how
to manage them.

I'd really like to see a historic 95th percentile (like last month's billing
period) and a "current" 95th percentile based on usage in this billing
period.

Giving this some more thought, I'm a little unclear on the current 5-minute
MRTG graphs...Do the 5-minute graphs show peaks, or are they showing an
average? I've always assumed it was an averaged value, but I turned on
"WithPeaks" to see what would happen. I then get peaks on all the other
graphs, but *not* on the 5-minute graph. What am I not understanding here?

George


----- Original Message -----
From: Brian Pavane <***@liii.com>
To: George D. Nincehelser <***@aaih.net>
Cc: Pete Templin <***@urdirect.net>; David C Prall <***@dcptech.com>;
<***@list.ee.ethz.ch>
Sent: Sunday, September 26, 1999 4:23 PM
Subject: [mrtg] Re: 95th percentile?
Post by Brian Pavane
I don't think the 95th percentile information really needs to be graphed.
Meerly a program that could pull the readings every 5 minutes from a piece
of equipment (say... a cisco), and then via cron at the end of the month
spit out a report would be sufficient. This type of program doesn't need
to be tied to MRTG, however, being that MRTG is already pulling this data,
it would be best to find a way to reuse what MRTG pulls.
Post by George D. Nincehelser
That's exactly what I've been looking for.
I assume that there is a way an external script can gather the data and
calculate the 95th percentile, then pass that info in some way to MRTG so
it can be graphed?
I'm unclear as to the best way to graph the 95th percentile data. I'm
thinking that a simple straight line across the graph would be sufficient
for billing purposes, but a trending line might look cool. I imagine the
line would best be reset at the end of each billing period.
Thoughts?
George
Post by Erik Van Riper
-----Original Message-----
Behalf Of Pete Templin
Sent: Friday, September 24, 1999 10:04 AM
To: David C Prall
Subject: [mrtg] Re: 95th percentile?
Post by David C Prall
What exactly are you looking for. I just use the WithPeak[]: wmy option.
This shows the high along with the averages. Giving you four lines
rather then two.
At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the standard
default configmaker script. And at that point WithPeak was implemented,
so I used it.
Most folks looking for 95th percentile want to be able to calculate the
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days, and
then it only stores averaged averages. It's impossible to recreate the
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour period.
There's no way to determine how many five minute samples were that high.
Some people have written scripts that archive a day's worth of five-minute
samples each day, and then crank through that unaveraged data at the end
of the month. The few folks I've talked to about this usually wrote it on
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales well
(or can be reengineered to scale).
Pete
--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator
On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
San Antonio, TX 78229 (210)692-9911
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Brian Pavane
1999-09-26 23:13:38 UTC
Permalink
George,

Well, with the graphing or without the graphing you also need some output
that you could easily take and enter into your billing records. Perhaps
even a webpage at the end of the month showing you (and your customer)
their total bandwidth for the month, avg, and 95th percentile. Point
being, a graph is nice, and I do see your reason for graphing it, but I
don't want to have my billing dept. looking at graphs to determine
billing. I would not mind having them going to a website though that
clearly stated what the bandwidth was.
You don't *have* to graph it, but doing so should sure make make it a lot
easier to explain to your boss or customer (or maybe yourself) what it
means. (e.g. "If this little line (traffic) goes higher than that little
line (95th percentile) a little to often, open up your wallet.)
Since the 95th percentile figure translates directly into $$$, everyone one
(including the non-technical types) are really concerned with what the value
is and where it is going. It needs to be expressed quickly and clearly, and
putting it on the graph seems to me a good way to do it. Having this info
would give everyone viewing them a clear view of our costs and ideas on how
to manage them.
I'd really like to see a historic 95th percentile (like last month's billing
period) and a "current" 95th percentile based on usage in this billing
period.
Giving this some more thought, I'm a little unclear on the current 5-minute
MRTG graphs...Do the 5-minute graphs show peaks, or are they showing an
average? I've always assumed it was an averaged value, but I turned on
"WithPeaks" to see what would happen. I then get peaks on all the other
graphs, but *not* on the 5-minute graph. What am I not understanding here?
George
----- Original Message -----
Sent: Sunday, September 26, 1999 4:23 PM
Subject: [mrtg] Re: 95th percentile?
Post by Brian Pavane
I don't think the 95th percentile information really needs to be graphed.
Meerly a program that could pull the readings every 5 minutes from a piece
of equipment (say... a cisco), and then via cron at the end of the month
spit out a report would be sufficient. This type of program doesn't need
to be tied to MRTG, however, being that MRTG is already pulling this data,
it would be best to find a way to reuse what MRTG pulls.
Post by George D. Nincehelser
That's exactly what I've been looking for.
I assume that there is a way an external script can gather the data and
calculate the 95th percentile, then pass that info in some way to MRTG
so
Post by Brian Pavane
Post by George D. Nincehelser
it can be graphed?
I'm unclear as to the best way to graph the 95th percentile data. I'm
thinking that a simple straight line across the graph would be
sufficient
Post by Brian Pavane
Post by George D. Nincehelser
for billing purposes, but a trending line might look cool. I imagine
the
Post by Brian Pavane
Post by George D. Nincehelser
line would best be reset at the end of each billing period.
Thoughts?
George
Post by Erik Van Riper
-----Original Message-----
Behalf Of Pete Templin
Sent: Friday, September 24, 1999 10:04 AM
To: David C Prall
Subject: [mrtg] Re: 95th percentile?
Post by David C Prall
What exactly are you looking for. I just use the WithPeak[]: wmy
option.
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
Post by David C Prall
This shows the high along with the averages. Giving you four lines
rather then two.
At some point there was a patch for doing this, but at somepoint I
believe it got converted into the WithPeak option. I have only been
using MRTG for 6 months where I have actually looked beyond the
standard
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
Post by David C Prall
default configmaker script. And at that point WithPeak was
implemented,
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
Post by David C Prall
so I used it.
Most folks looking for 95th percentile want to be able to calculate
the
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
95th percentile of all of the data over a given time period (usually a
month). MRTG only keeps its five minutes numbers for about two days,
and
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
then it only stores averaged averages. It's impossible to recreate
the
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
true 95th percentile five minute average using reaveraged numbers; the
peaks merely show the highest five minute peak across a two hour
period.
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
There's no way to determine how many five minute samples were that
high.
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
Some people have written scripts that archive a day's worth of
five-minute
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
samples each day, and then crank through that unaveraged data at the
end
it on
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
company time, and weren't permitted to release their script. It's
actually not that hard; the challenge is making sure that it scales
well
Post by Brian Pavane
Post by George D. Nincehelser
Post by Erik Van Riper
(or can be reengineered to scale).
Pete
--
Peter J. Templin, Jr., CCNA
Systems and Networks Administrator
On-Line Internet Services - URDirect.net
A division of Global On-Line Computers
San Antonio, TX 78229 (210)692-9911
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Stieers, Ken
1999-09-27 12:46:54 UTC
Permalink
Giving this some more thought, I'm a little unclear on the
current 5-minute
MRTG graphs...Do the 5-minute graphs show peaks, or are they
showing an
average?
Technically its a rate, not an average. Its not created by taking multiple
polls, its ONE poll.
I've always assumed it was an averaged value, but I
turned on
"WithPeaks" to see what would happen. I then get peaks on
all the other
graphs, but *not* on the 5-minute graph. What am I not
understanding here?
The weekly graphs are created by averageing 6 polls of the daily graphs. If
WithPeaks is turned on, it shows you the highest poll of those six for that
30 minute slice of time (6 polls * 5 min).


HTH,

Ken


--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Larry Sheldon
1999-09-27 13:30:58 UTC
Permalink
Post by Stieers, Ken
Giving this some more thought, I'm a little unclear on the
current 5-minute
MRTG graphs...Do the 5-minute graphs show peaks, or are they
showing an
average?
Technically its a rate, not an average. Its not created by taking multiple
polls, its ONE poll.
I am not a statistician, I don't even play one anywhere.

I think Ken is close, but my understanding is that it is actually the average
rate over the "5-minute" period--the arithmetic is I believe is something like
the number of bytes counted in the "5-minute" period divided by the number of
seconds in the period.

I put "5-minute" in quotes because I think MRTG records the actual time-stamp
for the sample, and given the vagaries of computers and such, the next
sample is not going to be exactly five minutes later.
--
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
. .
- L. F. (Larry) Sheldon, Jr. -
. Unix Systems and Network Administration .
- Creighton University Computer Center-Old Gym -
. 2500 California Plaza .
- Omaha, Nebraska, U.S.A. 68178 Two identifying characterstics -
. ***@creighton.edu of System Administrators: .
- 402 280-2254 (work) Infallibility, and the ability to -
. 402 681-4726 (cellular) learn from their mistakes. .
- 402 332-4622 (residence) -
. http://www.creighton.edu/~lsheldon Adapted from Stephen Pinker .
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Martin Ansdell-Smith
1999-09-25 11:31:00 UTC
Permalink
Post by Tim Kennedy
there's a script around somewhere called "calculate_traffic.pl".
It takes a begining date and an ending date, and tells you how much
traffic passed in that time period, and you can do the math with your CIR
or allowed rate, and see what percentile you are in for the month.
This highlights a point to consider when calculating percentiles: you need
to be using the same sampling period. If the provider is monitoring over a
ten-minute period then you need to do the same and synchronise the start
of each monitoring period. The calculation of the 95th percentile is
straightforward but you need to use the same sample interval and the same
number of samples to get the same answer. If you use the default 5-minute
load figures from mrtg and calculate the 95th percentile for the month
then what you have is the 95th percentile of the 5-minute loadings for the
month: this may be very different from the 95th percentile of 1-hour
loadings for the month or the 95th percentile of 1-second loadings for the
month.

On a similar theme, CIR/EIR calculations are (usually?) done per second so
the 5-minute intervals shown (by default) in mrtg can only give you an
indication of what traffic might be in excess of the CIR if your traffic
profile is reasonably constant second by second over the whole 5-minute
period.

Martin

--
Martin Ansdell-Smith
Network Analyst http://www.ansdell.demon.co.uk/

--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
George D. Nincehelser
1999-09-28 22:46:38 UTC
Permalink
I thought about this some more, and have convinced myself that in most cases
the sampling rate and sychronization isn't all that critical given the way
the information is being used for billing purposes.

The numbers you get, and the numbers your providers get may not be exactly
the same, but they will lie pretty close to each other. Given that the
providers I am looking at have 0.5Mbps billing bands, there's quite a bit of
room to play with.

My guess is that if you are well within a band, there's no problem. But if
you are hovering around the border between bands, the $$$ involved can be
significant...perhaps $1000 in a month depending on if you are above or
below the line. I can see where that could cause some arguments about when
and how often samples are taken.

But it must work farily well, since it seems the 95th percentile is an
industry standard...or maybe it's just the best solution available?

George
Post by Martin Ansdell-Smith
This highlights a point to consider when calculating percentiles: you need
to be using the same sampling period. If the provider is monitoring over a
ten-minute period then you need to do the same and synchronise the start
of each monitoring period. The calculation of the 95th percentile is
straightforward but you need to use the same sample interval and the same
number of samples to get the same answer. If you use the default 5-minute
load figures from mrtg and calculate the 95th percentile for the month
then what you have is the 95th percentile of the 5-minute loadings for the
month: this may be very different from the 95th percentile of 1-hour
loadings for the month or the 95th percentile of 1-second loadings for the
month.
On a similar theme, CIR/EIR calculations are (usually?) done per second so
the 5-minute intervals shown (by default) in mrtg can only give you an
indication of what traffic might be in excess of the CIR if your traffic
profile is reasonably constant second by second over the whole 5-minute
period.
Martin
--
Martin Ansdell-Smith
Network Analyst http://www.ansdell.demon.co.uk/
--
* To unsubscribe from the mrtg mailing list, send a message with the
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
--
* To unsubscribe from the mrtg mailing list, send a message with the
subject: unsubscribe to mrtg-***@list.ee.ethz.ch
* The mailing list archive is at http://www.ee.ethz.ch/~slist/mrtg
Continue reading on narkive:
Loading...