m1_rate, m5_rate and m15_rate

[Stephen Connolly]

I am seeing some strangeness with the different rates... it seems like the m1, m5 and m15 rates are the wrong way around.

Metrics 3.0.1

Starting from a rate of 0 rpm, adding 60+ rpm (since the metric reporting involves a few requests per min)

timers: 

{

• http.requests: 

  {

  • count: 717,
  • max: 12.830905000000001,
  • mean: 0.08323803207810322,
  • min: 0.00041700000000000005,
  • p50: 0.0032900000000000004,
  • p75: 0.006025,
  • p95: 0.02296930000000001,
  • p98: 0.03605908,
  • p99: 0.9900917800000395,
  • p999: 12.830905000000001,
  • values: 

    [],
  • stddev: 0.8774492983815769,
  • m15_rate: 97.84780070301642,
  • m1_rate: 39.61643622214264,
  • m5_rate: 83.52638152649571,
  • mean_rate: 153.43836709106407,
  • duration_units: "seconds",
  • rate_units: "calls/minute"

  },

 

Assuming the graph makes it through to the mailing list, you can see that the 15min average responds to the increased request load faster than the 1 min average which is slowest to respond.

After 5 minutes the m15 and m5 averages are the same:

timers: 

{

• http.requests: 

  {

  • count: 1073,
  • max: 10.389893,
  • mean: 0.034416146887159535,
  • min: 0.00041700000000000005,
  • p50: 0.002837,
  • p75: 0.004268500000000001,
  • p95: 0.019312649999999997,
  • p98: 0.026047239999999985,
  • p99: 0.034209150000000084,
  • p999: 10.388753155,
  • values: 

    [],
  • stddev: 0.5230684819414663,
  • m15_rate: 83.69251518764169,
  • m1_rate: 50.0368033558766,
  • m5_rate: 82.33652134806663,
  • mean_rate: 118.27674942776319,
  • duration_units: "seconds",
  • rate_units: "calls/minute"

  },

Am I just mis-understanding these metrics or is there an incorrect constant in the exponentially weighted moving averages?

-Stephen

[Ryan W Tenney]

This was fixed in 3.0.2

// ryan w tenney

--
You received this message because you are subscribed to the Google Groups "metrics-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to metrics-user...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Stephen Connolly]

Cool

On Tuesday, 4 March 2014, Ryan W Tenney <ry...@10e.us> wrote:

This was fixed in 3.0.2

// ryan w tenney

On Tue, Mar 4, 2014 at 12:42 PM, Stephen Connolly <stephen.al...@gmail.com> wrote:

I am seeing some strangeness with the different rates... it seems like the m1, m5 and m15 rates are the wrong way around.

Metrics 3.0.1

Starting from a rate of 0 rpm, adding 60+ rpm (since the metric reporting involves a few requests per min)

timers: 

{

• http.requests: 

  {

  • count: 717,
  • max: 12.830905000000001,
  • mean: 0.08323803207810322,
  • min: 0.00041700000000000005,
  • p50: 0.0032900000000000004,
  • p75: 0.006025,
  • p95: 0.02296930000000001,
  • p98: 0.03605908,
  • p99: 0.9900917800000395,
  • p999: 12.830905000000001,
  • values: 

    [],
  • stddev: 0.8774492983815769,
  • m15_rate: 97.84780070301642,
  • m1_rate: 39.61643622214264,

--

You received this message because you are subscribed to the Google Groups "metrics-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to metrics-user...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
You received this message because you are subscribed to the Google Groups "metrics-user" group.
To unsubscribe from this group and stop receiving emails from it, send an email to metrics-user...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

--
Sent from my phone