On Fri Oct 27, 2006 at 07:55:23 +0200, Jaime Nebrera wrote: > I answer this one first. > > You fill 200GB regulary? Then most probably you are not aggregating properly > nor removing older data.
The traffic that is being collected by pmacct is a significant amount. We are aggregating properly though but archival of old data is something of an issue. Some data needs to stick around for a while for various purposes, however this is irrelevant to the discussion of sFlow. > As somebody already has pointed you, ISP dont bill on 100% precise data. > Actually, Cisco designed nF with this in mind so its something enough for > billing. > >> I guess my question at this point is, what do you do when you have say >> 1GBps >> of traffic to record and you need 100% accuracy? > > In this particular cases they do the following: > > 1) Pray > 2) Prepare a deeeep pocket > 3) Sample !!! > > Or use cards like the ones from Endace: http://www.endace.com OK I'm starting to see the logic behind sFlow and how it is used in reality. It seems like the logical step for us to take with our pmacct system. It would appear that the sFlow configuration in ProCurve 3400cl switches is only maintained temporarily in the MIB. There is a timeout value after which the sFlow configuration essentially reverts to original empty values. Naturally sfacctd cares not about this! It is easy to set the correct values using the sFlowenable script from InMon, but what happens when the timeout expires? Obviously the script has to run again to set the correct values. This concerns me that we need something continually setting the values in the switch for sFlow to continue operating. Has anyone else run into this problem? -- Regards, Oliver Hookins _______________________________________________ pmacct-discussion mailing list http://www.pmacct.net/#mailinglists
