UT - User Tracking capacity calculating - number of endhosts

Unanswered Question
Jul 18th, 2007
User Badges:
  • Blue, 1500 points or more

starting with LMS 2.5 the 'unused up' and 'unused down' reports where introduced;


that means, UT stores information for all endhost ports in the DB, regardless if they have an endhost connected or not. So my question is if the following is correct for the initial FIRST major acquisition of UT (with prerequisite that for all devices successfull data collection was done and UT does not encounter any problems!)?

after this first major acquisition, the number of entries in the DB is (nearly) the number of available enduser ports of all devices managed by Campus Manager. Nearly, because there are (could be) ports with duplicate MACs (like IP phones with PC connected to them or some server ports). For these ports I can generate a report 'Ports with multiple MAC' to get the amount and a list of the entries.

And this is only true for the first acquisition because changing entries are not necessarily removed from the UT db (depending on the purge policy)


As a result, when estimating if one (1) installation of campus fits for the network (regarding capacity of UT which currently is 100 000 entries) he has to look at the total number of ports on all devices managed by campus to get an idea of this scenario and not only at the number of endusers on the devices because UT counts every port.


the documentation is talking of currently supporting 100 000 endhosts (250 000 endhosts in LMS 3.0) but I think they are talking of entries in the UT database (regardless if there are endhosts connected or not) and not 'real' endhosts. Is that correct or does anybody has other experiences or information on this?



  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 4.7 (3 ratings)
Loading.

Speaking from my current experience, I don't think the numbers add up either (but then again consider the source)


From one of my currently running Campus 4.06 servers I've currently got 170930 End Hosts in the database but have about half a million managed ports deployed. The 5.0 documentation lists supporting 250,000 endhosts and 2,500,000 end host records. I suspect that this is a low ball estimate too.

Martin Ermel Wed, 07/18/2007 - 04:26
User Badges:
  • Blue, 1500 points or more

thanks for your reply!

I assume you have Campus as standalone installation. Which platform and spec are you using?


Ok, so it seems UT only acts on ports it has ever categorized as a valid enduser port which means that ports that have never been used will never be listed in any UT report (even not in the 'unused down' report)

And the number of endhosts seems to be related to the process of query the necessary information (MAC and ARP table)from the devices.

So does so number of entries in the UT db affect the performance of the major/minor acquisition or only the performance of generating any reports?


Joe Clarke Wed, 07/18/2007 - 07:25
User Badges:
  • Cisco Employee,
  • Hall of Fame,

    Founding Member

UT can look at unused ports, and you can see these when you run the switch port usage reports in User Tracking. These reports are enhanced in UT 5.0 to provide you a way of specifying the number of days or a date to see how long a port as been inactive.


As for performance, while there is no hard stop in UT, TAC will refuse support if you go above the recommended maximum. The more end hosts in UT, the more disk space is used, the longer reports can take, and the more memory UTMajorAcquisition will consume. If there are too many end users, UTM will crash.

Martin Ermel Wed, 07/18/2007 - 23:28
User Badges:
  • Blue, 1500 points or more

so, does the 'switch port usage' report currently first pulls all ports for the devices from the Wbu tables and then compares this result with the UT Embu tables?


speaking of endhosts in this way means number of entries in UT db, which could highly depend on the purge pulicy, and NOT number of endhosts/endusers currently in the network.

Why does UTM memory consumption is dependent on the number of UT db entries? I think it acquires info from the *active* network and then updates the db. So I think the number of rows is not such a main thing for it as it is for the reports, but I see that the number of endusers plays a big role. If I missunderstood, please correct me.


Joe Clarke Thu, 07/19/2007 - 09:56
User Badges:
  • Cisco Employee,
  • Hall of Fame,

    Founding Member

Yes. The total number of switch ports are compared to those that have hosts/devices connected to them.


When UTM starts up, it loads all of the existing entries into memory as you do not necessarily want old entries to disappear after each acquisition.

My configuration is a cluster of 4 Sun 240's (8GB RAM/16 swap, 2 CPU, 4x 172GB local disks) servers using SSO. CS and RME on box 1, CS and Campus Box 2, CS DFM box 3 and CS IPM box 4. My redundant cluster is only two boxes but I have DFM turned off on it. But its nice because I can take down either for a day or two fix bugs, apply patches, etc without interrrupting service to my customers.


My active inventory is out of spec being some 5,500 devices composed of routers, switches, access points spread from Maine to Hawaii.

Martin Ermel Thu, 07/19/2007 - 09:32
User Badges:
  • Blue, 1500 points or more

thanks for this detailed information!


MArtin

Actions

This Discussion