Timers

Timers record the time it takes to do things. They’re a bit like histograms where the value being recorded is time.

Timers should be a fairly intuitive concept. They can tell you things like:

75% of all searches took 0.5 seconds or less. 95% of all searches took 1.0 seconds or less.

Timers also track the rate of the timed events, so it’s like they have a meter metric built-in for convenience.

Creating

Create your timer:

(require '[metrics.core :refer [new-registry]])
(require '[metrics.timers :refer [timer]])

(def image-processing-time (timer "image-processing-time"))

The timer function is idempotent, which means that you don’t need to keep a local reference to the timer. Once a timer has been registered, a call to (timer reg "image-processing-time") will return the existing timer.

You can also use the deftimer macro to create a timer and bind it to a var in one concise, easy step:

(require '[metrics.timers :refer [deftimer]])

(deftimer image-processing-time)

All the def[metric] macros do some magic to the metric title to make it easier to define.

Writing

Once you have a timer you can record times to it in three different ways.

time!

You can record the time it takes to evaluate one or more expressions with the time! macro:

(require '[metrics.timers :refer [time!]])

(time! image-processing-time
       (process-image-part-1 ...)
       (process-image-part-2 ...))

Or if you haven’t held a reference to image-processing-time, you can do the following:

(time! (timer reg "image-processing-time")
       (process-image-part-1 ...)
       (process-image-part-2 ...))

time-fn!

time! is a macro. If you need a function instead (e.g.: for map’ing over a list), you can use time-fn!, but you’ll need to pass it a function instead of a body:

(require '[metrics.timers :refer [time-fn!]])

(time-fn! image-processing-time
          (fn []
            (process-image-part-1 ...)
            (process-image-part-2 ...)))

start/stop

You can also use the start and stop functions in metrics.timers, assuming you hang onto the Timer$Context instance that is returned.:

(require '[metrics.timers :as tmr])

(tmr/deftimer my-tmr)

(let [a (tmr/start my-tmr)
      b (tmr/start my-tmr)
      c (tmr/start my-tmr)]
    (Thread/sleep 1000)
    (println (tmr/stop c))
    (println (tmr/stop b))
    (println (tmr/stop a))
#_ => 1000266000 ; nanoseconds this instance ran for.
      1000726000
      1000908000
      nil

Reading

percentiles

You can use percentiles to find the percentage of actions that take less than or equal to a certain amount of time:

(require '[metrics.timers :refer (percentiles)])

(percentiles image-processing-time)
=> { 0.75  232.00
     0.95  240.23
     0.99  280.01
     0.999 400.232
     1.0   903.1 }

This returns a map of the percentiles you probably care about. The keys are the percentiles (doubles between 0 and 1 inclusive) and the values are the maximum time taken for that percentile. In this example:

  • 75% of images were processed in 232 nanoseconds or less

  • 95% of images were processed in 240 nanoseconds or less

  • … etc

If you want a different set of percentiles just pass them as a sequence:

(require '[metrics.timers :refer [percentiles]])

(percentiles image-processing-time [0.50 0.75])
=> { 0.50 182.11
     0.75 232.00 }

number-recorded

To get the number of data points recorded over the entire lifetime of this timers:

(require '[metrics.timers :refer [number-recorded]])

(number-recorded image-processing-time)
=> 12882

smallest

To get the smallest data point recorded over the entire lifetime of this timer:

(require '[metrics.timers :refer [smallest]])

(smallest image-processing-time)
=> 80.66

largest

To get the largest data point recorded over the entire lifetime of this timer:

(require '[metrics.timers :refer [largest]])

(largest image-processing-time)
=> 903.1

mean

To get the mean of the data points recorded over the entire lifetime of this timer:

(require '[metrics.timers :refer [mean]])

(mean image-processing-time)
=> 433.12

std-dev

To get the standard deviation of the data points recorded over the entire lifetime of this timer:

(require '[metrics.histograms :only [std-dev]])

(std-dev image-processing-time)
=> 300.51

sample

You can get the current sample points the timer is using with sample, but you almost certainly don’t care about this. If you use it make sure you know what you’re doing.

(require '[metrics.timers :refer [sample]])

(sample image-processing-time)
=> [803.234 102.223 ...]

TODO: Rates