ruby is slow about ruby-redis HOT 6 CLOSED

Oh, bee tee dubs, those benchmarks are shockingly inaccurate. Here are some real benchmarks:

C Redis:

PING: 47846.89 requests per second
PING (multi bulk): 42372.88 requests per second
SET: 46314.81 requests per second
GET: 45477.27 requests per second
INCR: 43103.45 requests per second
LPUSH: 45054.05 requests per second
LPOP: 44444.45 requests per second
SADD: 45871.56 requests per second
SPOP: 47851.67 requests per second
LPUSH (again, in order to bench LRANGE): 43859.65 requests per second
LRANGE (first 100 elements): 7072.14 requests per second
LRANGE (first 300 elements): 2442.00 requests per second
LRANGE (first 450 elements): 1663.34 requests per second
LRANGE (first 600 elements): 1204.38 requests per second

Ruby Redis:

PING: 27472.53 requests per second
PING (multi bulk): 24154.59 requests per second
SET: 7610.35 requests per second
GET: 22273.94 requests per second
INCR: 23041.47 requests per second
LPUSH: 210.77 requests per second
LPOP: 15898.25 requests per second
SADD: 6558.03 requests per second
SPOP: 26251.97 requests per second
LPUSH (again, in order to bench LRANGE): 210.57 requests per second
LRANGE (first 100 elements): 1354.41 requests per second
LRANGE (first 300 elements): 497.69 requests per second
LRANGE (first 450 elements): 332.33 requests per second
LRANGE (first 600 elements): 254.10 requests per second

At best ruby-redis is about half as slow as the real redis, at worst it's more than 200 times slower.

from ruby-redis.

The readme benchmarks are from a 32-bit platform with 8k payload and 1000 parallel clients. Your write operations seem defective. My laptop just ran set=>19417, get=>18018, lpush=>15174, sadd=>16583. Please post more about your setup. What version of Ruby? Did you tune the benchmark options?

from ruby-redis.

I don't think the point is to have a fast implementation. It is rather to have an alternative implementation of the same protocol, which is always good. It can be used for learning, experimentation and behavior comparison. There's no reason to use it in production instead of the real Redis. Am I mistaken?

from ruby-redis.

@dturnbull:

I'm running 64 bit OS X on a Core 2 Quad with Ruby 1.9.2. I ran the default settings in redis-benchmark for both.

@catwell:

I realise that, I just don't believe the benchmarks given are very accurate.

from ruby-redis.

charliesome, there is something wrong with your Ruby install. There is no reason writing a string to a hash takes significantly longer than reading it back. From my MacBook Pro:

Using the default OSX Ruby 1.8.7-p174: set=>7639, get=>10869, lpush=>9718, sadd=>11579
Using MacPorts Ruby 1.9.2-p180: set=>19417, get=>18018, lpush=>15174, sadd=>16583

Roughly, you should expect Ruby Redis will perform 80% as well as C Redis on 32-bit platforms and about 50% on 64-bit.

from ruby-redis.

no longer see benchmarks in readme, trololol

from ruby-redis.