On 5/7/09 4:25 AM, Bongjae Chang wrote:
> Hi.
> I have been trying to abstract out the transport layer in Shoal
> experimentally.
> In the process of reviewing JXTA dependency in Shoal, I have a
> question about JXTA's OutputPipe.
> JXTA's OutputPipe is mainly used in the following 3 parts.
> 1. ClusterManager: for sending AppMessage and DSC's Message
> 2. MasterNode: for sending master's message
> 3. HealthMonitor: for sending Health's message
> And above 3 parts include a pipe cache map with one another.
> So, I am curious to know why above 3 modules have OutputPipes with one
> another.
> Are any problems likely to happen in JXTA if above 3 modules share
> only one OutputPipe?
These pipes have different endpoints to each of the modules, we can use
a single endpoint which can be demuxed, but why do that, as this already
provides this separation.
> I think that one OutputPipe is enough because peerid is already needed
> for sending a message, but I wonder if I miss any important point in
> using JXTA.
this also maximizes message flow, as the output pipe allow one sender at
a time. Using different pipes reduces this contention.
> Please advice me.
> Thanks.
> --
> Bongjae Chang