[Spread-users] spread daemon

Ryan Caudy caudy at jhu.edu
Thu May 15 11:04:37 EDT 2003

The difference is in how the communication takes place.  Clients on the 
same host as their daemon can communicate with it using Unix domain 
sockets (IPC); otherwise they use TCP.  Daemons communicate with other 
daemons over multicast/broadcast channels if possible (unicast if that 
fails) using UDP.  So, if you run four daemons, transmissions will only 
need to go on the network once (plus or minus some amount of traffic for 
reliability and ordering).  If you run one daemon, it will receive each 
message from a client via unicast, and send it out to each client via 
unicast; for the remote clients, this will be over TCP.  The data will 
be transferred over the network something like n times, where n is the 
number of clients (with some additional traffic caused by TCP reliability).

The other issue is that, with a spread cluster of four daemons, the 
behavior of spread when experiencing network/host faults is more 
meaningful.  With only one daemon, there can be at most one partition of 
the spread network.


Cédric Coulon wrote:

>I use spread in a cluster of PC with 4 nodes.
>I would know the differences beetwen launching a spread's daemon on each
>node and only one spread's daemon on one node?
>Cédric Coulon                                          Tel:   02 51 12 59 63
>IRIN, 2, rue de la Houssiniere, BP 92208    Fax: 02 51 12 58 12
>44322 NANTES cedex 3
>FRANCE                     mailto:Cedric.Coulon at irin.univ-nantes.fr
>Spread-users mailing list
>Spread-users at lists.spread.org

More information about the Spread-users mailing list