[ndnSIM] a problem in 'ndn-simple-with-pit-operation-stats'
Alex Afanasyev
alexander.afanasyev at ucla.edu
Mon Nov 4 23:42:19 PST 2013
Hi Amin,
This is entirely effect of "Lru" PIT entry "replacement" policy. This is actually **not** a default one. I think, if you try to run stats with default policy (ns3::ndn::pit::Persistent::AggregateStats), you may get results that you're expecting.
With Lru, every time a new interest coming in, some old element is getting evicted, if capacity is reached. When Persistent policy is used, whenever PIT capacity is reached, new items will be rejected, not existing items evicted.
---
Alex
On Oct 2, 2013, at 3:45 AM, Amin Karami <amin at ac.upc.edu> wrote:
> Hi Alex,
> I ran the ./waf --run=ndn-simple-with-pit-operation-stats by Lru implementation for PIT table.
> I set PIT Size to 50: ndnHelper.SetPit ("ns3::ndn::pit::Lru::AggregateStats","MaxSize","50");
> Interest rate to 1000: consumerHelper.SetAttribute ("Frequency", StringValue ("1000"));
> Finally, the output is:
> Time NodeId Updates Inserts Lookups Erases
> 1 1 0 990 3 940
> 2 1 0 999 0 999
> 3 1 0 998 0 998
> 4 1 0 1000 0 1000
> 5 1 0 1000 0 1000
> 6 1 0 1000 0 1000
> 7 1 0 1000 0 1000
> 8 1 0 1000 0 1000
> 9 1 0 1000 0 1000
> 10 1 0 1000 0 1000
>
> As you see, the number of Insert/Lookup/Erase is based on the rate of the Consumer (1000 interests per second), not based on the PIT maximum size. I think maximum insert/Erase should be fixed to 50 (PIT size) not to 1000 (consumer rate).
> Is there any problem?
>
>
> Best Regards,
> Amin
More information about the ndnSIM
mailing list