[arin-ppml] Queue depth report?

John Curran jcurran at arin.net
Tue Sep 30 11:23:48 EDT 2014


On Sep 30, 2014, at 10:45 AM, Matthew Kaufman <matthew at matthew.at> wrote:

> I've been watching Geoff's ARIN runout prediction slip out into the future, and reading here about increased review leading to slower processing... Is runout being delayed just because legitimate applications for space are queuing up?
> 
> Can we get a report on historical and current application queue depth from staff?

Matthew - 
 
  Queue lengths for IPv4 assignment and allocations did lengthen slightly
  when we instituted team review (per Phase 1 of the countdown plan which
  was reached in April), but they've been steady since and are not what is 
  driving Geoff's runout numbers to slip...

  Geoff's outlook is based on recent allocation history (he does a least 
  squares linear best fit on this data), which means that his estimate
  will move up after a very large allocation (such as was made in this 
  past April) and then will begin to slide out if the following months
  do not maintain that predicted higher allocation rate.

  The second graph on this page shows the monthly allocation from the 
  free pool in terms of /24 equivalents; the April allocation is rather
  pronounced - <https://www.arin.net/knowledge/statistics/index.html>

  Geoff's best fit for ARIN Address Allocation Rate can be seen in the
  graph here - <http://www.potaroo.net/tools/ipv4/fig27i.png>  Note how
  it spiked after that allocation, but has been slowing down since...
  (it's a little hard to read because of the 5 year scale, but the 
  newest non-zero data on the far right side is the important part.)

  If we should make another large assignment or allocation, then Geoff's
  estimates will pull in according; Geoff has noted on several occasions
  that towards the end of the free pool, it becomes fairly difficult to 
  have a reliable prediction since it becomes more about the timing of a 
  rather small number of discrete events (e.g. the very large allocations)
  than the calculated run rate.

  If you need anything additional, please let me know... (Also, I've cc'd 
  the esteemed Geoff Huston on this email; it's only fair in any lengthly
  discussion of his great efforts. :-) )

Thanks!
/John

John Curran
President and CEO
ARIN






More information about the ARIN-PPML mailing list