[nanomsg] Re: nanomsg: the big picture
- From: Martin Sustrik <sustrik@xxxxxxxxxx>
- To: nanomsg@xxxxxxxxxxxxx
- Date: Thu, 05 Sep 2013 11:36:42 +0200
Hi Paul,
So, the admin looks at the graph above, sees there's an disconnection
between two intermediary devices and can actually do something about it. Is
a network connection broken? Or maybe he just set the address wrong in the
DNS? Etc.
Actually it's oversimplified. There is no such thing as disconnect in
nanomsg. It may be endpoint which has no underlying transports. But if
DNS SRV is implemented right, that thing is even more subtle.
Distinguish the user and admin roles here:
For the user TCP disconnect basically doesn't exist; it's totally
invisible and it should be.
However, from the admin point of view you want to know whether
particular TCP connection is broken, so that you can fix the underlying
problem.
Also admins do not look at the graphs in the first place. They get
SMSes or Emails when monitoring system get's error. And only if it's
not clear from monitoring data what's wrong, the graphs are helpful.
So the statistical data is a primary thing to implement. And then
graphs, on top of it.
Sorry for the wording, I've used "graph" instead of more generic
"monitoring data".
I am not going to disucss implementatation details here,
but the idea is that admins store the actual info about topology
setup in a distributed database (such as DNS) and that the library
translates topology://market-data-feed" into actual endpoint(s) to
connect to by querying the database.
So do you thing the DNS database alone is enough to build the graph?
(except actual connection counts of course)
Definitely not. DNS (or alternative) stores rules to build the graph
rather than the graph itself.
Thus, I would argue, the graph should be built from monitoring data alone.
Martin
Other related posts:
