[Spread-users] interpretation of high retrans rate over WAN

[Stanisław Trytek]

Hello all,
I have some problems with interpretation of some of spmonitor  
informations, mainly high retrans rate.
I am working with spread over WAN, the composition of demons is now very  
simple: 3 demons in 3 segments, messages - mainly FIFO.

The retransmission rate seems to me be incredibly high, in order to deal  
with it I would like to know more about protocol used in spread.
I have found out from the mailng list archives that packets are sent by  
udp among segments still using RING protocol.
So I assume that my three daemons form RING  ex. PROC1(leader of that  
ring)->PROC2->PROC-3->.... and the token revolves round it.
In what way data packets are sent? In ring or point-to-point in that  
manner ex. PROC1->PROC2, PROC1->PROC3 and respectively for each other?

I understand why s retrans is 0 (it is mcast or broadcast address in one  
segment) , but not why b retrans is more than 0. What brodcast means  
considering communication across segments?

Delta sec is the time between status queries. Why is it so big and even  
negative once? The same question about Delta Mes?

Could someone help me with that issue.

Stanislaw Trytek


Example of status:

Monitor> Monitor: send status query

============================
Status at PROC1 V 3.17. 3 (state 1, gstate 1) after 56267 seconds :
Membership  :  3  procs in 3 segments, leader is PROC1
rounds   :   44513      tok_hurry :   10836     memb change:      42
sent pack:    8166      recv pack :    1986     retrans    :   37982
u retrans:   37067      s retrans :       0     b retrans  :     915
My_aru   :    2490      Aru       :    2490     Highest seq:    2490
Sessions :      10      Groups    :      10     Window     :      60
Deliver M:   10703      Deliver Pk:   10917     Pers Window:      15
Delta Mes:    5684      Delta Pack:       3     Delta sec  :   42338
==================================

Monitor>
============================
Status at PROC2 V 3.17. 3 (state 1, gstate 1) after 7675 seconds :
Membership  :  3  procs in 3 segments, leader is PROC1
rounds   :    9924      tok_hurry :     662     memb change:      13
sent pack:     137      recv pack :    3418     retrans    :       2
u retrans:       2      s retrans :       0     b retrans  :       0
My_aru   :    2490      Aru       :    2490     Highest seq:    2490
Sessions :       3      Groups    :      10     Window     :      60
Deliver M:    2981      Deliver Pk:    3045     Pers Window:      15
Delta Mes:   -7722      Delta Pack:       0     Delta sec  :  -48592
==================================

Monitor>
============================
Status at PROC3 V 3.17. 3 (state 1, gstate 1) after 13939 seconds :
Membership  :  3  procs in 3 segments, leader is PROC1
rounds   :   18759      tok_hurry :    1321     memb change:      17
sent pack:     646      recv pack :   13190     retrans    :     772
u retrans:     439      s retrans :       0     b retrans  :     333
My_aru   :    2490      Aru       :    2490     Highest seq:    2490
Sessions :       2      Groups    :      10     Window     :      60
Deliver M:    5022      Deliver Pk:    5188     Pers Window:      15
Delta Mes:    2041      Delta Pack:       0     Delta sec  :    6264
==================================


-- 
_________________________________________
PRZEMYSŁOWY INSTYTUT TELEKOMUNIKACJI
TELECOMMUNICATIONS RESEARCH INSTITUTE
ul. Poligonowa 30,  04-051 Warszawa,  Poland
tel:. +48 (22) 48 65 232
e-mail: tryteks at pit.edu.pl