[Varnish] #613: Efficiency figure for varnishstat

[Varnish]

#613: Efficiency figure for varnishstat
-------------------------+--------------------------------------------------
 Reporter:  slink        |       Owner:  phk  
     Type:  enhancement  |      Status:  new  
 Priority:  normal       |   Milestone:       
Component:  varnishstat  |     Version:  trunk
 Severity:  normal       |    Keywords:       
-------------------------+--------------------------------------------------
 In some deployment scenarios, the (cache) hit ratio of varnishstat
 suggests that the cache was providing great benefit and really the cache
 itself probably is, but still (too) many requests hit the backend in the
 form of passes or even piped sessions.

 I am suggesting to add an efficiency measure to varnishstat, defined as
 the ratio of client requests minus backend requests by client requests.

 efficiency = client requests - passes - pipes - fetches / client requests

 While one could argue that cache hits by client requests would do as well,
 at least as of now I believe that this measure is more accurate because it
 also takes into account that one client request may trigger multiple
 backend requests.

 With the attached patch, varnishstat will look something like this

 {{{
 2+12:13:01
 HOSTNAME
    Hitrate ratio:          10      100     1000
    Hitrate avg:        0.9721   0.9721   0.9731
 Efficiency ratio:          10      100     1000
 Efficiency avg:        0.9505   0.9522   0.9533

     55697963       200.97       256.93 Client connections accepted
    402992210      1518.81      1858.98 Client requests received
    390022582      1471.82      1799.15 Cache hits
         1549         0.00         0.01 Cache hits for pass
      9053637        22.00        41.76 Cache misses
 }}}

 Two notes on the patch
  * I've chosen to implement a macro rather than a function, but I'll be
 happy to make this a function if core team members prefer
  * One might want to exclude the additional "ratio" line because users may
 find it redundant. And in fact it's probably not very likely that the two
 sets of numbers will ever differ. At any rate, I think state for the
 number of samples should be kept even if it's not printed.

-- 
Ticket URL: <http://varnish.projects.linpro.no/ticket/613>
Varnish <http://varnish.projects.linpro.no/>
The Varnish HTTP Accelerator