|
From: Ingen S. J. v. (ICTS) <j.v...@ut...> - 2011-04-07 12:58:34
|
Hi Ivan, all, On Thu, 2011-04-07 at 14:24 +0200, Ivan Brunello wrote: > You pointed out a good possible flaw in my assumptions. > > I would not bother to an high level of detail. > AFAIK, DHCP allocation (on NOT expired leases) on our DHCP servers is > done one a "lowest free ip address first". > Not checked documentation (they are mostly Microsoft), but this is > what seems to happen. OK. I gave it a try with our data and the numbers match fairly well with what I'd expect (including subnets for wireless). > This would give me a quite good (good enough for me) snapshot of used address. > And our wifi relies heavily on DHCP and REALLY SHORT LEASES :-) > > I would check a 7 days arp count: long enough to keep track to weekly > peaks, and short enough to not contain too many allocated but not > currently used addresses. > > > My need is to check e.g. whether a /24 always contains no more than 40 > address, and this would lead to a leaner /26. > We're consolidating our ip addressing to avoid overlap w/ other > departments, and we need to make a lot of capacity planning. Well, let's try it in SQL first before implementing this in a report. Data is sorted by subnet size first, then by percentage used (so you can easily see where you can reclaim the most space). select net as "Subnet", power(2, (32-masklen(net))) as "Subnet size", count(distinct ip) as "Count active last week", round(100 * count(distinct ip) / power(2, (32-masklen(net)))) as "Pct used" from subnets, node_ip where ip <<= net and active and time_last > now() - interval '7 days' and last_discover > now() - interval '7 days' group by net order by 2 desc, 4, net; Let me know how this works out for you; if this is what you're looking for, I think I can implement it in a report in the near future. Regards, Jeroen van Ingen ICT Service Centre University of Twente, P.O.Box 217, 7500 AE Enschede, The Netherlands |