On 10/11/2016 04:13 PM, Joakim wrote: > On Monday, 10 October 2016 at 08:47:54 UTC, Robert burner Schadek wrote: >> http://code.dlang.org/packages/dstatsd >> >> StatsD allows to collect statistics about any application by using >> counters, gauges and more through UDP. >> >> Usage: >> >> auto s = new StatsD("127.0.0.1", 1234, ""); // connect to statsd server >> >> s(Counter("Foo")); // increment counter "Foo" >> s.inc("Bar"); // increment counter "Foo" >> >> s(Counter("Args"), // send stats to Args, H, and timeA >> Counter("H", someIntValue), // in one UDP message >> Timer("timeA", someTimeInMS) >> ); >> >> { >> auto a = ScopeTimer("args", s); // automatic time collection >> } > > Never heard about this either, I ignore node.js stuff. I was just > reading this interesting post on tracing/profiling a couple days ago: is > it efficient enough to find some of these tail latency issues?
Stats aggregation usually has nothing to do with fine tuned performance profiling - it is application defined way to monitor relevant metrics of runtime behavior. For example, one can aggregate metrics for web app request processing latencies to monitor if those stay within expected margin - but if issue is spotted, it won't help much in debugging _why_ it has happened.
signature.asc
Description: OpenPGP digital signature