[Spread-users] High retrans rate
Yair Amir
yairamir at cnds.jhu.edu
Wed Oct 2 07:03:30 EDT 2002
Martin,
It seems to me that your multicast address does not work well and therefore
both your receivers loose most packets (any time you see sretrans that means that the
message had to be retransmited using multicast - which means that both
receivers missed it).
There are two programs "s" and "r" that we give with the distribution.
Build them and try running s for sender and r for receiver using the
226.0.0.6 and see what happens. s and r are just helping you to
measure the network behavior using native udp without Spread.
Cheers,
:) Yair.
Martin> Hi,
Martin> I am experiencing a very high retransmission rate when testing spflooder with
Martin> 1000 byte messages.
Martin> sent pack: 1001 recv pack : 6 retrans : 1135
Martin> We are evaluating Spread 3.17 in a setup with 1 sender and 2 receivers in the
Martin> same LAN. We use a Catalyst 5500 Switch with a static multicast address. The
Martin> spread conf looks like:
Martin> Spread_Segment 226.0.0.6:4803 {
Martin> dev901 195.232.84.170
Martin> dev911 195.232.84.188
Martin> dev912 195.232.84.189
Martin> }
Martin> I tested with:
Martin> dev912$ ./bin/spflooder -m 1000 -b 1000 -ro
Martin> dev911$ ./bin/spflooder -m 1000 -b 1000 -ro
Martin> dev901$ ./bin/spflooder -m 1000 -b 1000 -wo
Martin> Interesting seems that only one receiver (dev912) needed retrans.
Martin> ============================
Martin> Status at dev901 V 3.17. 0 (state 1, gstate 1) after 29 seconds :
Martin> Membership : 3 procs in 1 segments, leader is dev901.lab.aol.de
Martin> rounds : 516 tok_hurry : 12 memb change: 1
Martin> sent pack: 1001 recv pack : 6 retrans : 1135
Martin> u retrans: 8 s retrans : 1127 b retrans : 0
Martin> My_aru : 1008 Aru : 1008 Highest seq: 1008
Martin> Sessions : 0 Groups : 0 Window : 60
Martin> Deliver M: 1000 Deliver Pk: 1008 Pers Window: 15
Martin> Delta Mes: 1000 Delta Pack: 1005 Delta sec : 13
Martin> ==================================
Monitor>>
Martin> ============================
Martin> Status at dev912 V 3.17. 0 (state 1, gstate 1) after 26 seconds :
Martin> Membership : 3 procs in 1 segments, leader is dev901.lab.aol.de
Martin> rounds : 515 tok_hurry : 12 memb change: 1
Martin> sent pack: 3 recv pack : 1004 retrans : 156
Martin> u retrans: 156 s retrans : 0 b retrans : 0
Martin> My_aru : 1008 Aru : 1008 Highest seq: 1008
Martin> Sessions : 0 Groups : 0 Window : 60
Martin> Deliver M: 1000 Deliver Pk: 1008 Pers Window: 15
Martin> Delta Mes: 0 Delta Pack: 0 Delta sec : -3
Martin> ==================================
Monitor>>
Martin> ============================
Martin> Status at dev911 V 3.17. 0 (state 1, gstate 1) after 23 seconds :
Martin> Membership : 3 procs in 1 segments, leader is dev901.lab.aol.de
Martin> rounds : 515 tok_hurry : 12 memb change: 1
Martin> sent pack: 3 recv pack : 1433 retrans : 0
Martin> u retrans: 0 s retrans : 0 b retrans : 0
Martin> My_aru : 1008 Aru : 1008 Highest seq: 1008
Martin> Sessions : 0 Groups : 0 Window : 60
Martin> Deliver M: 1000 Deliver Pk: 1008 Pers Window: 15
Martin> Delta Mes: 0 Delta Pack: 0 Delta sec : -3
Martin> ==================================
Martin> Changing the parameters of spflooder to "-m 100000 -b 100" resulted in an
Martin> acceptable retrans rate:
Martin> ============================
Martin> Status at dev912.lab.aol.de V 3.17. 0 (state 1, gstate 1) after 173 seconds :
Martin> Membership : 3 procs in 1 segments, leader is dev901.lab.aol.de
Martin> rounds : 33123 tok_hurry : 45 memb change: 2
Martin> sent pack: 4 recv pack : 38617 retrans : 122
Martin> u retrans: 122 s retrans : 0 b retrans : 0
Martin> My_aru : 100008 Aru : 100008 Highest seq: 100008
Martin> Sessions : 0 Groups : 0 Window : 60
Martin> Deliver M: 100000 Deliver Pk: 100010 Pers Window: 15
Martin> Delta Mes: 0 Delta Pack: 0 Delta sec : 15
Martin> ==================================
Monitor>>
Martin> ============================
Martin> Status at dev901.lab.aol.de V 3.17. 0 (state 1, gstate 1) after 177 seconds :
Martin> Membership : 3 procs in 1 segments, leader is dev901.lab.aol.de
Martin> rounds : 33125 tok_hurry : 45 memb change: 2
Martin> sent pack: 37103 recv pack : 7 retrans : 2153
Martin> u retrans: 0 s retrans : 2153 b retrans : 0
Martin> My_aru : 100008 Aru : 100008 Highest seq: 100008
Martin> Sessions : 0 Groups : 0 Window : 60
Martin> Deliver M: 100000 Deliver Pk: 100010 Pers Window: 15
Martin> Delta Mes: 0 Delta Pack: 0 Delta sec : 4
Martin> ==================================
Monitor>>
Martin> ============================
Martin> Status at dev911.lab.aol.de V 3.17. 0 (state 1, gstate 1) after 168 seconds :
Martin> Membership : 3 procs in 1 segments, leader is dev901.lab.aol.de
Martin> rounds : 33014 tok_hurry : 41 memb change: 1
Martin> sent pack: 3 recv pack : 38745 retrans : 0
Martin> u retrans: 0 s retrans : 0 b retrans : 0
Martin> My_aru : 100008 Aru : 100008 Highest seq: 100008
Martin> Sessions : 0 Groups : 0 Window : 60
Martin> Deliver M: 100000 Deliver Pk: 100008 Pers Window: 15
Martin> Delta Mes: 0 Delta Pack: 0 Delta sec : -9
Martin> Can you confirm such a behaviour?
Martin> Martin
More information about the Spread-users
mailing list