Re: Memory overhead of vector clocks…. how often are they pruned?

[Radim Kolar]

From my point vector clocks is too much overhead. If you sync clocks in 
your cluster using NTP (which you should do anyway) you will get clock 
precision < 1/1000s which is good enough.

all my machines running NTP has offset < 1/1000s. They are FreeBSD, 
Linux is not that precise in clock syncing.

     remote           local      st poll reach  delay   offset    disp
=======================================================================
*barricade.rack9 64.6.104.18      2   64  377 0.06187  0.000996 0.05093