JRedisPipeline comparative performance

Comparative runs using the redis-benchmark as a reference point (but its apples and oranges).

Using a single thread to continuously pipeline 10000 commands and then wait for completion of the last request. Repeat and keep running average.
Using jvm flags: -server -Xms512m -Xmx1536m
using connection spec: SocketProperty.SO_RCVBUF, 1024*256

Snippets of redis-benchmark output interspersed with JRedisPipeline stats. Pipeline response times for fire and forget are obviously better.


====== SET ======
  10031 requests completed in 0.33 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1
30305.14 requests per second

JRedisPipeline: 

   10000 SETs average response time @ 0.086 seconds 
116279.07 requests per second 

====== GET ======
  10000 requests completed in 0.34 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1
29673.59 requests per second

JRedisPipeline: 

  10000 GETs average response time @ 0.072 seconds 
138888.89 requests per second

====== INCR ======
  10007 requests completed in 0.34 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1
29090.12 requests per second

JRedisPipeline: 
  10000 INCRs average response time @ 0.070 seconds 
142857.14 requests per second

====== LPUSH ======
  10010 requests completed in 0.33 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1

30425.53 requests per second

JRedisPipeline: 
  10000 LPUSHs average response time @ 0.081 seconds 
123456.79 requests per second 

====== PING ======
  10019 requests completed in 0.33 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1

30452.89 requests per second

JRedisPipeline: 
  10000 PINGs average response time @ 0.064 seconds 
156250.00 requests per second