On Wed, Apr 7, 2010 at 4:36 AM, Simon Riggs <[email protected]> wrote:
> On Wed, 2010-04-07 at 01:23 -0700, Ryan Tomayko wrote:
>> But we also get occasional 200ms response times in those runs. Here's
>> the max response times for the same memslap runs graphed above:
>
> It wouldn't be uncommon to see the occasional spike, though that graph
> shows the very worst response time in the whole profile. It would be
> useful to see all of the requests above a threshold, so we can see if a
> pattern exists and to see how whether this occurs in groups.

Yeah. I was hoping to tease that out of the memslap result data but I
can't find an easy way to get it. I'm starting in on a set of custom
non-libmemcached tests now so I'll see what I can come up with.

> Are the connections all equally and constantly busy, or are some
> connections idle for much of the time?

All connections are constantly busy. The results in my previous
messages are from memslap
<http://docs.tangent.org/libmemcached/memslap.html>. It basically just
pounds the server with a configurable number of connections for a
given time period unless you pass the --tps arg, which I haven't been
able to get working. Here's the results from a quick run over the
network at 50 concurrent connections:

$ memslap -s memcache5 --threads=1 --time=5s --concurrency=50
servers : memcache5
threads count: 1
concurrency: 50
run time: 5s
windows size: 10k
set proportion: set_prop=0.10
get proportion: get_prop=0.90

Get Statistics (297962 events)
   Min:       165
   Max:    204267
   Avg:       751
   Geo:    676.71
   Std:   1743.18
   Log2 Dist:
       8:      775    92879   162248    40450
      12:     1077      358      154        0
      16:        0        0       21

Set Statistics (33134 events)
   Min:       155
   Max:    203581
   Avg:       777
   Geo:    688.75
   Std:   2258.45
   Log2 Dist:
       8:       76    10018    17598     5245
      12:      124       52       17        0
      16:        0        0        4

Total Statistics (331096 events)
   Min:       155
   Max:    204267
   Avg:       753
   Geo:    677.91
   Std:   1801.66
   Log2 Dist:
       8:      851   102897   179846    45695
      12:     1201      410      171        0
      16:        0        0       25

cmd_get: 298009
cmd_set: 33137
get_misses: 0
written_bytes: 57470410
read_bytes: 331002892
object_bytes: 36053056

Run time: 5.0s Ops: 331146 TPS: 66202 Net_rate: 74.1M/s

Thanks,
Ryan


-- 
To unsubscribe, reply using "remove me" as the subject.

Reply via email to