Re: [Nfsen-discuss] SFlow or Graphing Issue?

[James W. <jam...@gm...>]

Reporting back, as I think this is solved.  I have reduced the
sampling rate from 1 in 1000 to 1 in 10000.   I am now getting what
appear to be appropriate statistics.


Thanks,
James

On Tue, Jun 25, 2013 at 2:59 PM, James Wright <jam...@gm...> wrote:
> Just to share further investigation, it appears that the graphs are
> correct, based on the stored flow data.  Using nfdump shows the same
> avg bps as what I see in the graph.  I still don't know why this is
> though.  It seems that the issue is the switch or the way it samples
> flows.  The switches in question are 3Com 4800G (in case anyone has
> experience with them and sflow).
>
>
> Thanks,
> James
>
> On Thu, Jun 20, 2013 at 2:35 PM, James Wright <jam...@gm...> wrote:
>> Hello Folks,
>>
>> I've been happily collecting netflow for a while and decided to try
>> collecting from some devices providing sflow.  Once setup the
>> collection seems to work well.  My issue is that the traffic (Bits/s)
>> graph seems to hit a ceiling of ~125.8 Mb/s.  At first I thought
>> something was wrong with the switches, except that showing the stats
>> on the interface that I am collecting sflow from seems to indicate
>> that I am moving much more data than this (all over, but upwards to
>> and above 200 Mb/s).
>>
>> It may be that I am not correctly understanding sflow, but I see
>> reference to it being used for metering/billing, which suggests some
>> accuracy exists.  Currently I am collecting ingress/egress on the
>> uplink ports of the switches (some are 1G, some 10G), a sampling size
>> of 1000, and an interval of 2.  It seems to me that this would be a
>> bit accurate.
>>
>> The other graphs/statistics appear to be correct, and they do not
>> appear to be hitting a ceiling.
>>
>> I am continuing to dig and learn, but has anyone else seen this before?
>>
>>
>> Thanks,
>> James