Measure latency of app.breathe() #777
Conversation
This commit adds a histogram facility that can record the distribution of different sampled magnitudes. In particular, it is useful for recording distributions of times. Histograms can be mapped into /var/run/snabb, where they can be analyzed by a separate process. This commit also wires up the app.breathe() loop to record latencies for its breath cycles into a well-known file (/var/run/snabb/PID/engine/latency), and wires up "snabb top" to do some basic statistics on this data.
|
By analyzing the blame information on this pull request, we identified @eugeneia, @lukego and @javierguerragiraldez to be potential reviewers |
|
Originally from Igalia#276. |
|
I'd like to nominate @eugeneia to be the reviewer/upstream on this one, if he's willing. |
| if options.measure_latency or options.measure_latency == nil then | ||
| local histogram = require('lib.histogram') | ||
| local latency = histogram.create('engine/latency', 1e-6, 1e0) | ||
| print('hi') |
There was a problem hiding this comment.
I assume this is a debug leftover?
|
Super cool feature, love to see @lukego I think the |
|
Snabb top has been really useful to us -- for example if we are experiencing packet loss, we need to find where it happens: between which apps, or if not between any apps then on ingress. It helped us have insight on whether to apply backpressure or not (usually we don't, now, following #656). It's good stuff :) I can look at switching to use core.shm.map; good suggestion. I guess by default core.shm.map should clear the file if a previous PID had it open, though. |
Reduced the number of globals
This commit adds a histogram facility that can record the distribution
of different sampled magnitudes. In particular, it is useful for
recording distributions of times. Histograms can be mapped into
/var/run/snabb, where they can be analyzed by a separate process.
This commit also wires up the app.breathe() loop to record latencies for
its breath cycles into a well-known
file (/var/run/snabb/PID/engine/latency), and wires up "snabb top" to do
some basic statistics on this data.