Looking at possibilities for an implementation of usage-based billing, it seems that the same techniques and tools always come up. I'm looking for some feedback from the list on experiences with these tools and techniques as well as alternatives that may not be listed here. +Techniques --Flow data (Netflow, SFlow, etc) analysis to determine 95th percentile traffic levels --SNMP polling of interface counters to determine 95th percentile traffic levels Granted, there are many interpretations of how to calculate "95th percentile traffic levels" that may differ from provider to provider. Assume that we have established the method that we will use. +Tools --RTG --MRTG --Cacti --solarwinds Orion --Various, expensive PM tools such as Netcool Proviso, JDSU NetComplete, InfoVista tools, etc --flow-tools and FlowScan combo --Arbor Networks Peakflow Any follow up to this thread, either on or off list, or pointers to previous threads with good information would be appreciated! My search didn't turn up any, but that doesn't mean that they don't exist! Thanks! Regards, Mauricio Rodriguez Manager of IP/Data Engineering, FPL FiberNet Email: Mauricio.Rodriguez@fpl.com<mailto:Mauricio_Rodriguez@fpl.com>
Rodriguez, Mauricio said the following, On 3/5/2009 2:02 PM:
Looking at possibilities for an implementation of usage-based billing, it seems that the same techniques and tools always come up. I'm looking for some feedback from the list on experiences with these tools and techniques as well as alternatives that may not be listed here.
+Techniques --Flow data (Netflow, SFlow, etc) analysis to determine 95th percentile traffic levels --SNMP polling of interface counters to determine 95th percentile traffic levels
I've been using pmacctd for several years with good success. It is designed for actual usage accounting (for instance, GB per month per IP/subnet) rather then 95th percentile. The beauty of pmacctd is that it simply puts the data into an SQL database. Once the data is stored, you can write any SQL queries as required for reporting. http://www.pmacct.net/ Wim
On Thu, 5 Mar 2009, Rodriguez, Mauricio wrote:
Looking at possibilities for an implementation of usage-based billing, it seems that the same techniques and tools always come up. I'm looking for some feedback from the list on experiences with these tools and techniques as well as alternatives that may not be listed here.
+Techniques --Flow data (Netflow, SFlow, etc) analysis to determine 95th percentile traffic levels --SNMP polling of interface counters to determine 95th percentile traffic levels
I need to look into this in the near future as well. The problems I'm aware of are: 1) we have customers on policed ports, and the interface snmp counters count packets before service-policy. It doesn't seem right to bill for packets we dropped :)...so this isn't useful data for billing purposes. 2) our customer agg gear (cisco 3550s) don't do netflow. Our bigger switches the agg gear uplinks to does (6509 sup720-3bxls), but can't handle export of full netflow, so we run sampled. It's still useful for abuse tracking, but billing based on it would require some large assumptions and multipliers...unlikely to be of use. The remaining option I'm aware of is to use monitor sessions to send a copy of our traffic to a system/device which would then either generate "full" netflow data or just distill the traffic into data xfered per IP/network. What are people using for this on the several hundred mbit/s to a few gigabits/s or more range? Are there other ways? ---------------------------------------------------------------------- Jon Lewis | I route Senior Network Engineer | therefore you are Atlantic Net | _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
I use netacct - can grab data per cidr block and dumps data into mysql. I wrote scripts from there to graph in rrdtool, bill on total usage, or bill on 95th percentile. http://netacct-mysql.gabrovo.com/ -Jack Carrozzo On Thu, Mar 5, 2009 at 7:41 PM, Jon Lewis <jlewis@lewis.org> wrote:
On Thu, 5 Mar 2009, Rodriguez, Mauricio wrote:
Looking at possibilities for an implementation of usage-based billing, it seems that the same techniques and tools always come up. I'm looking for some feedback from the list on experiences with these tools and techniques as well as alternatives that may not be listed here.
+Techniques --Flow data (Netflow, SFlow, etc) analysis to determine 95th percentile traffic levels --SNMP polling of interface counters to determine 95th percentile traffic levels
I need to look into this in the near future as well. The problems I'm aware of are:
1) we have customers on policed ports, and the interface snmp counters count packets before service-policy. It doesn't seem right to bill for packets we dropped :)...so this isn't useful data for billing purposes.
2) our customer agg gear (cisco 3550s) don't do netflow. Our bigger switches the agg gear uplinks to does (6509 sup720-3bxls), but can't handle export of full netflow, so we run sampled. It's still useful for abuse tracking, but billing based on it would require some large assumptions and multipliers...unlikely to be of use.
The remaining option I'm aware of is to use monitor sessions to send a copy of our traffic to a system/device which would then either generate "full" netflow data or just distill the traffic into data xfered per IP/network. What are people using for this on the several hundred mbit/s to a few gigabits/s or more range?
Are there other ways?
---------------------------------------------------------------------- Jon Lewis | I route Senior Network Engineer | therefore you are Atlantic Net | _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
Once upon a time, Jon Lewis <jlewis@lewis.org> said:
1) we have customers on policed ports, and the interface snmp counters count packets before service-policy. It doesn't seem right to bill for packets we dropped :)...so this isn't useful data for billing purposes.
Not sure how you are policing, but I belive both Juniper and Cisco have MIBs that show the policed traffic. For example, when we used Cisco CAR to limit traffic on some ports, I set up Cricket to monitor both the base port and the CAR stats (so we could see how much traffic was actually passed). I haven't got around to doing it for Juniper firewall policers, but I pretty sure the info is in a MIB. -- Chris Adams <cmadams@hiwaay.net> Systems and Network Administrator - HiWAAY Internet Services I don't speak for anybody but myself - that's enough trouble.
Jon Lewis wrote:
On Thu, 5 Mar 2009, Rodriguez, Mauricio wrote:
Looking at possibilities for an implementation of usage-based billing, it seems that the same techniques and tools always come up. I'm looking for some feedback from the list on experiences with these tools and techniques as well as alternatives that may not be listed here.
+Techniques --Flow data (Netflow, SFlow, etc) analysis to determine 95th percentile traffic levels --SNMP polling of interface counters to determine 95th percentile traffic levels
I need to look into this in the near future as well. The problems I'm aware of are:
1) we have customers on policed ports, and the interface snmp counters count packets before service-policy. It doesn't seem right to bill for packets we dropped :)...so this isn't useful data for billing purposes.
Torrus (www.torrus.org) can use the Cisco MIBs to graph pre and post-policy packets. http://www.torrus.org/plugins/tp-cisco-cbqos.pod.html Sam
On Mon, 9 Mar 2009, Sam Stickland wrote:
I need to look into this in the near future as well. The problems I'm aware of are:
1) we have customers on policed ports, and the interface snmp counters count packets before service-policy. It doesn't seem right to bill for packets we dropped :)...so this isn't useful data for billing purposes.
Torrus (www.torrus.org) can use the Cisco MIBs to graph pre and post-policy packets.
I just checked a few 3550s and 3560s, and I don't see any evidence that they support CISCO-CLASS-BASED-QOS-MIB. Walking .1.3.6.1.4.1.9.9.166, I get nothing. On our 6500s, walking .1.3.6.1.4.1.9.9.166, I do get lots of data, but I don't see anything obvious in it that tells me which numbers go with which interfaces. Is anyone actually making use of CISCO-CLASS-BASED-QOS-MIB on 3550s to graph/bill traffic passed after per-interface policing service policies are applied? ---------------------------------------------------------------------- Jon Lewis | I route Senior Network Engineer | therefore you are Atlantic Net | _________ http://www.lewis.org/~jlewis/pgp for PGP public key_________
I'd like to add: --flow-tools and FlowViewer ( http://ensight.eos.nasa.gov/FlowViewer ) Keeps max, mean, and 95th pct. for up to three years for any predefined customer (defined by a flow-tool filter and stored in RRDtool). Can group customers for visual comparison. Joe "Rodriguez, Mauricio" <Mauricio.Rodriguez@fpl.com> wrote on 03/05/2009 05:02:02 PM:
Looking at possibilities for an implementation of usage-based billing, it seems that the same techniques and tools always come up. I'm looking for some feedback from the list on experiences with these tools and techniques as well as alternatives that may not be listed here.
+Techniques --Flow data (Netflow, SFlow, etc) analysis to determine 95th percentile traffic levels --SNMP polling of interface counters to determine 95th percentile traffic levels
Granted, there are many interpretations of how to calculate "95th percentile traffic levels" that may differ from provider to provider. Assume that we have established the method that we will use.
+Tools --RTG --MRTG --Cacti --solarwinds Orion --Various, expensive PM tools such as Netcool Proviso, JDSU NetComplete, InfoVista tools, etc --flow-tools and FlowScan combo --Arbor Networks Peakflow
Any follow up to this thread, either on or off list, or pointers to previous threads with good information would be appreciated! My search didn't turn up any, but that doesn't mean that they don't exist!
Thanks!
Regards, Mauricio Rodriguez Manager of IP/Data Engineering, FPL FiberNet Email: Mauricio.Rodriguez@fpl.com<mailto:Mauricio_Rodriguez@fpl.com>
On 5 Mar 2009, at 22:02, Rodriguez, Mauricio wrote:
Looking at possibilities for an implementation of usage-based billing, [...] +Techniques --SNMP polling of interface counters to determine 95th percentile traffic levels
+Tools --RTG --MRTG --Cacti
You can be pretty sure that there are probably a lot of billing systems based loosely around the tools mentioned here, and that in their particular environments they are robust enough to be successful. If you go down this route, or a similar route, you need to remember that you MUST update the default RRD storage behaviours so that the most granular data is not aggregated until four or five months after the billing month, otherwise you will find that the averaging causes different results to fall out of your scripts when you look into customer queries on their bill ! Best wishes Andy
participants (8)
-
Andy Davidson
-
Chris Adams
-
Jack Carrozzo
-
Joe Loiacono
-
Jon Lewis
-
Rodriguez, Mauricio
-
Sam Stickland
-
Wim Kerkhoff