spf-discuss
[Top] [All Lists]

SPF timings and statistics.

2004-02-23 13:02:33
With the recent talk about SPF being slow and tying up SMTP sessions, I
thought I'd toss in some of our statistics.

The following is a SPF statistical report from a brief run on a
production box:

14:48:49 ecelerity> spf_v1 dns stats
SPF induced DNS Queries:
        TXT: 6424
        PTR: 0
         MX: 11
          A: 55
  arpa IPv4: 17
  arpa IPv6: 0

SPF DNS responses:
      Total: 6507
        Bad: 6290

SPF DNS cache:
    Lookups: 16657
   Hit Rate: 60.9353%

14:49:03 ecelerity> spf_v1 spf stats
      Result:      Total [percent]  SPF check time
=====================================================
       error:          0 [  0.00%]  avg:      0ms
        pass:        697 [  8.16%]  avg:      0ms
     unknown:       7790 [ 91.19%]  avg:   1338ms
        fail:         53 [  0.62%]  avg:   1477ms
    softfail:          3 [  0.04%]  avg:    112ms
   aggregate:       8543 [100.00%]  avg:   1229ms

The above results show, at least in our set up, that the cache helps SPF
passes tremendously -- given the 0ms average lookup time.

Obviously these statistics are of limited usefulness without STD dev. 
If I get sometime, perhaps I can show a distribution graph of SMTP
session latency introduced by enabling SPF on receipt.

I should reiterate that I think the latency is unimportant when it is so
reasonable bounded (under 60 seconds).  As it doesn't really cause more
CPU or memory consumption on the server side.  We don't see any
throughput decreases at all with SPF enabled -- just more concurrent
connections.  But, as latency seemed to be a hot topic here and I
thought the numbers presented were unreasonably high, I thought I'd add
my two cents.

-- 
// Theo Schlossnagle
// Principal Engineer -- http://www.omniti.com/~jesus/
// Postal Engine -- http://www.postalengine.com/
// Ecelerity: fastest MTA on earth



<Prev in Thread] Current Thread [Next in Thread>
  • SPF timings and statistics., Theo Schlossnagle <=