[Codel] drop state short-circuit...

[Jeff Weeks <jweeks@sandvine.com>]

Hello,

I originally posted this in the aqm list; re-posted to this, more appropriate, list:

There exists a short-circuit into a "deeper" drop state if we were previously and recently in a drop state.  I understand the need for this, but I'm curious where the '16' came from?  How was this decided upon?

For example, here's the ns3 implementation:

          /*
           * if min went above target close to when we last went below it
           * assume that the drop rate that controlled the queue on the
           * last cycle is a good starting point to control it now.
           */
          int delta = m_count - m_lastCount;
          if (delta > 1 && CoDelTimeBefore (now - m_dropNext, 16 * Time2CoDel (m_interval)))
          {
            m_count = delta;
            NewtonStep ();
          }

If I read this correctly, it says;

If we have previously dropped for more than one interval, and have re-entered the drop state in less then 16 intervals from the previous drop state, then short circuit count to the delta... but why 16 intervals?

--Jeff