yairamir at cnds.jhu.edu
Tue Jun 22 14:32:32 EDT 2004
P.> Hello Yair,
P.> Thank you for the prompt response. I read the 2 pager.
P.> Spread maintains only one ring of spread daemons, and the seq# is global
P.> across all groups.
P.> Looking @ Figure 1.4 which plots mesg latency versus number of groups.
P.> If I look at the curve for 10000B packets, the latency is about 2.8ms.
P.> The xmission latency due to pkt over 100Mb link is 0.8ms - what is the
P.> breakup of the remaining 2ms? How much of it is due to CPU (P3 in this
P.> expt) and how much is due to the ordering protocol - i.e., token based
Most of it is due to flow control and the need to provide reliability.
I think this is a good result considering that there are 20 computers playing
(that can potentially send messages at any time). I speculate that making the
CPU stronger is not likely to provide considerable improvement in this regards
(but may help the overall system including your application of course).
P.> I am just trying to get a gauge of performance expectation if we use a
P.> different setup for example, faster CPUs but continue using 100Mb/s
More information about the Spread-users