You will see IPs and MACs, so yeah, you should be able to distinguish
amongst machines as long as you don't have NAT or something going on.

Another solution would be to mirror the uplink interfaces from the
non-sflow switch.

Definitely look here for info about sFlow:

http://sflow.org/about/index.php

It'll show you what you do get from sFlow (an awful lot, as it turns
out) and pay special attention to the sampling stuff. You can at the
list back traffic for more discussion, but the basics are as follows:

1) sFlow is designed around sampling and most implementations do so at
less than 1:1 by default.

2) Ntop doesn't really handle the sampling at this point. Well - it does
- it will report what sampling rate sFlow is telling it that it is
using. But what it doesn't do is the multiplication. So if sFlow is
sampling, say, 1 in every 4 packets Ntop will only report approx. (see
link for sampling theory stuff) 1/4 of the total actual traffic
traversing the link. At one point Luca was looking at adding
functionality like this, but I'm not sure where he left it. You'll see
from the sampling theory that simply multiplying by the sampling rate
will usually work over the long term but is not necessarily accurate
when you're looking for small traffic flows.

Regards,

Chris

_________________________________________
Chris Moore
Senior Network Engineer
Guardian Mortgage Documents

303-942-2019
emergency (GMD Help Desk): 303-942-2002
[EMAIL PROTECTED]




-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of
Burton Strauss
Sent: Wednesday, May 17, 2006 7:07 AM
To: [email protected]
Subject: RE: [Ntop] SFlow and chain linked switches

Well, experience in general w/ sFlow seems pretty thin all over :-)

If the ProCurve sees the packets, it should report on them via sFlow to
ntop. You won't have a complete picture of the sub networks, only the
subset
of traffic that the ProCurve sees.

-----Burton

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of
Victor Rodriguez Cortes
Sent: Wednesday, May 17, 2006 6:30 AM
To: [email protected]
Subject: RE: [Ntop] SFlow and chain linked switches

        Hello,

> This really isn't an ntop question ... It belongs with the switch
> vendor. Because it's dependent on their implementation of sFlow and
> (more importantly) on what 'chain-linked' is and how it's implemented.

        Sorry for the OT, I just thought that someone here may have
experience with this kind of setup.

> Within the realm of standard Ethernet and tcp/ip, the 'procurve'
> switch in your diagram will never see packets local to either of the
> 'non-sFlow' switches, so it wouldn't be able to report ANYTHING about
> them.

        That's correct: if the traffic remains within any "non-sFlow"
switches I won't see anything. But if the traffic is like this:

Computer1 -->
                non s-Flow switch -> Procurve Switch -> gateway
Computer2 -->

        Would I be able to distinguishing from which computer the
traffic
comes from using sFlow at the Procurve switch?

        Thanks a lot,
        Victor.


> -----Burton
>
> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf
> Of Victor Rodriguez Cortes Sent: Tuesday, May 16, 2006 10:21 AM To:
> [EMAIL PROTECTED] Subject: [Ntop] SFlow and chain linked
> switches
>
>  Hello,
>
>  I'm new to the list and maybe someone has some experience with this
> and could give me some advice. I sent this questions to the Misc list,

> but I didn't received any reply yet. I hope that someone here may help

> me.
>
>  I would like to know what kind of information would SFlow give me
> from a chain- linked switch that does NOT support SFlow. The network
> would be something like this (sorry, I'm too bad drawing with
> letters!):
>
>
>        Gateway
>    |
>   Procurve 2800
>    |   |
> Non-sflow switchA     Non-slow switchB
>  |                    |
> Network devices       Network devices
>
>
>  For some systems (printers, voip phones, some departments...) I do
> not need the performance of one Procurve's Gb port, so I would use
> those ports for other systems and use a different switch to give
> network access to less performance-hungry devices. It's basically a
> matter of cost per port, trying to avoid buying 2 or more ProCurve and

> reuse some basic switches that I have right now.
>
>  - Would I see real information for traffic that comes from / goes to
> systems connected to the Non-sflow switches? Obviously all traffic
> would be detected on the Procurve port that has switchA and switchB
> connected to, but I suppose that I should get the IP address of each
> system and be able of identify each one through it (as is there was
> just one system with many IP addresses).
>
>  - Would I be able of getting real traffic stats from those systems?
>
>  - Would all this work if the switches are connected with a 2Gb trunk?
>
>  Many thanks in advance,
>  Victor.
>


_______________________________________________
Ntop mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop

_______________________________________________
Ntop mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop

**********************************************************************
Confidential/Proprietary Note

The information in this email is confidential and may be legally privileged.  
Access to this email by anyone other than the intended addressee is 
unauthorized.  If you are not the intended recipient of this message, any 
review, disclosure, copying, distribution, retention, or any action taken or 
omitted to be taken in reliance on it is prohibited and may be unlawful.  If 
you are not the intended recipient, please reply to or forward a copy of this 
message to the sender and delete the message, any attachments, and any copies 
thereof from your system.  Thank you.
Guardian Mtg Documents, Inc.
225 Union Boulevard, Suite 200
Lakewood, CO 80228.
**********************************************************************
_______________________________________________
Ntop mailing list
[email protected]
http://listgateway.unipi.it/mailman/listinfo/ntop

Reply via email to