I would normally agree, and although I dont have all the answers I do know
our network is somewhat different. in this environment the heaviest talkers
are machines that send tons, and tons of tiny packets (basically exactly
what you describe below :) ). What are you using as a collector? As i
mentioned maybe ntop can handle this, but using flow-tools and
flowscan/cuflow definately couldnt hang, which is all perl based.

On Thu, Feb 12, 2009 at 3:42 PM, Gary Gatten <[email protected]> wrote:

>  Something is not right! 500Mb/s is "nothing" if the traffic is typical
> and not 500Mb/s of 64byte packets all with unique source and destination
> info – Ie: some sort of DoS or test environment.
>
>
>
> The POS box I have sees peaks of 700Mb/s and still only uses maybe 60%
> cpu.  What are your rrd configs?  Maybe that's what's slowing everything
> down?  If you're using "full" and saving hosts, interfaces, etc. etc. – that
> could be it.  You're 2.4GHz system should EASILY handle 500Mb/s using
> netflow and not even break a sweat.
>
>
>
> G
>
>
>
>
>  ------------------------------
>
> *From:* [email protected] [mailto:[email protected]] *On Behalf Of
> *Kyle McLerren
> *Sent:* Thursday, February 12, 2009 5:30 PM
>
> *To:* [email protected]
> *Subject:* Re: [Ntop] Ntop NetFlow Question
>
>
>
> Well hopefully I can look forward to such a feature in the future. I'll
> play more with tweaking some settings, but in an environment like ours,
> processing flows on over 500Mbps worth of traffic is... interesting :) Even
> on a quad-core 2.4ghz collector with 4gb of ram and 10k SAS disks the flows
> were taking 10 minutes a piece to process with sampling turned off.
>
> Might just have to bite the bullet and move to some expensive but robust
> solution :) thanks again.
>
> On Thu, Feb 12, 2009 at 3:20 PM, Gary Gatten <[email protected]> wrote:
>
> Seems like it would be simple – multiply rx stats by sample rate before
> storing/graphing?  I don't do much development so who knows.
>
>
>
> I would maybe try without sampling, but maybe set your active/inactive
> timers to 120/60?  It's not as real-time as it could be, but if you have a
> ton of dynamic traffic it will slow down the flow rate.
>
>
>
> I'm only seeing traffic of around 100Kpps, but I'm on an old P-III and it's
> only using 25% during peaks.  Surely a fast box can keep up with netflow
> exports from really high util – unless every packet is a different "flow" –
> like during some sort of DoS attack.
>
>
>
> G
>
>
>  ------------------------------
>
> *From:* [email protected] [mailto:[email protected]] *On Behalf Of
> *Kyle McLerren
> *Sent:* Thursday, February 12, 2009 4:03 PM
>
>
> *To:* [email protected]
> *Subject:* Re: [Ntop] Ntop NetFlow Question
>
>
>
> Thanks for the response. We are doing a boat load of traffic. We had to
> enable sampling because our previous collector could no longer cope with the
> sheer volume of flows it was trying to process. I do admit I dont know if
> ntop can process the flows better as I just started to use it. Previous we
> were using flow-tools and flowscan/cuflow. I should not have said
> "accurate," as they are prefectly accurate. It just others get confuse when
> looking at them and it gets old telling people to keep in mind the data is
> accurate, its just the numbers are "smaller" then they actually really are
> :)
>
> so it looks like the answer is no, there isnt a way to set the sample
> rate.. seems like a really basic and easy feature to implement, would be
> great to see it!
>
> thanks again.
>
> On Thu, Feb 12, 2009 at 1:54 PM, Gary Gatten <[email protected]> wrote:
>
> How'd anyone ever function without Google?
>
>
>
> http://www.mail-archive.com/[email protected]/msg11605.html
>
>
>
>
>
>
>  ------------------------------
>
> *From:* [email protected] [mailto:[email protected]] *On Behalf Of
> *Gary Gatten
> *Sent:* Thursday, February 12, 2009 3:49 PM
> *To:* [email protected]
> *Subject:* Re: [Ntop] Ntop NetFlow Question
>
>
>
> I have seen this asked and I think answered a couple times, but since I
> don't do sampling I wasn't paying much attention.  I'm sure if you searched
> the list you'd see some answers from Luca and/or Burton.  I searched the FAQ
> and man page and couldn't find anything.  Also checked the "Preferences" and
> couldn't see anything there either.  MAYBE check the globals-define.h;
> there's lots of cool stuff in there but you have to recompile with changes
> L
>
>
>
> Of course the easy answer is to NOT sample.  Unless you have a $HIT load of
> traffic it will be OK, especially if you set the active/inactive export
> timers to something reasonable.
>
>
>
> BTW: What is "not accurate"?
>
>
>
>
>
>
>  ------------------------------
>
> *From:* [email protected] [mailto:[email protected]] *On Behalf Of
> *Kyle McLerren
> *Sent:* Thursday, February 12, 2009 3:13 PM
> *To:* [email protected]
> *Subject:* [Ntop] Ntop NetFlow Question
>
>
>
> Hi All,
>
> Im sure this has been answered before, but I couldnt find an answer
> anywhere. I use sampled netflow, and I simply wanted to know if theres an
> option with the ntop netflow plugin to configure the sample rate? Otherwise,
> my results arent acurate. Im sending 1 out of 100 from my router.
>
> thanks!
>
> "This email is intended to be reviewed by only the intended recipient and
> may contain information that is privileged and/or confidential. If you are
> not the intended recipient, you are hereby notified that any review, use,
> dissemination, disclosure or copying of this email and its attachments, if
> any, is strictly prohibited. If you have received this email in error,
> please immediately notify the sender by return email and delete this email
> from your system."
>
> "This email is intended to be reviewed by only the intended recipient and
> may contain information that is privileged and/or confidential. If you are
> not the intended recipient, you are hereby notified that any review, use,
> dissemination, disclosure or copying of this email and its attachments, if
> any, is strictly prohibited. If you have received this email in error,
> please immediately notify the sender by return email and delete this email
> from your system."
>
>
> _______________________________________________
> Ntop mailing list
> [email protected]
> http://listgateway.unipi.it/mailman/listinfo/ntop
>
>
>
> "This email is intended to be reviewed by only the intended recipient and
> may contain information that is privileged and/or confidential. If you are
> not the intended recipient, you are hereby notified that any review, use,
> dissemination, disclosure or copying of this email and its attachments, if
> any, is strictly prohibited. If you have received this email in error,
> please immediately notify the sender by return email and delete this email
> from your system."
>
>
> _______________________________________________
> Ntop mailing list
> [email protected]
> http://listgateway.unipi.it/mailman/listinfo/ntop
>
>
>   "This email is intended to be reviewed by only the intended recipient
> and may contain information that is privileged and/or confidential. If you
> are not the intended recipient, you are hereby notified that any review,
> use, dissemination, disclosure or copying of this email and its attachments,
> if any, is strictly prohibited. If you have received this email in error,
> please immediately notify the sender by return email and delete this email
> from your system."
>
> _______________________________________________
> Ntop mailing list
> [email protected]
> http://listgateway.unipi.it/mailman/listinfo/ntop
>
>
_______________________________________________
Ntop mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop

Reply via email to