Testing storage of millions of keys in Redis

gistfile1.py

#! /usr/bin/env python

import redis
import random
import pylibmc
import sys

r = redis.Redis(host = 'localhost', port = 6389)
mc = pylibmc.Client(['localhost:11222'])

REDIS_SETGET = False
REDIS_HSET = False
MC = False

NUM_ENTRIES = 1000000
MAX_VAL = 12000000

if len(sys.argv) != 2 or sys.argv[1] not in ('redis-normal', 'redis-hashes', 'memcached'):
    print 'Specify a test: redis-normal, redis-hashes, memcached'
    print 'NOTE: clear out memcached (restart) or Redis (FLUSHALL) before running'
    sys.exit(2)
if sys.argv[1] == 'redis-normal':
    REDIS_SETGET = True
elif sys.argv[1] == 'redis-hashes':
    REDIS_HSET = True
elif sys.argv[1] == 'memcached':
    MC = True

p = r.pipeline()
for i in range(0, NUM_ENTRIES):
    value = random.randint(0, MAX_VAL)
    if MC:
        mc.set(str(i), value)
    elif REDIS_SETGET:
        r.set(str(i), value)
    elif REDIS_HSET:
        bucket = int(i / 500)
        p.hset(bucket, i, value)

    if i % (NUM_ENTRIES/10) == 0:
        if REDIS_SETGET or REDIS_HSET:
            p.execute()
            p = r.pipeline()
        print i

# one final clear out
if REDIS_SETGET or REDIS_HSET:
    p.execute()

# get size
if MC:
    size = int(mc.get_stats()[0][1]['bytes'])
elif (REDIS_SETGET or REDIS_HSET):
    size = int(r.info()['used_memory'])

print '%s bytes, %s MB' % (size, size / 1024 / 1024)

I think you can save at least another 20% just by using the remainder instead of i as the key in the hash

Took 49 MB with Memcached and only 15 MB with redis hashes on my debian box with 384 MB of RAM.

I hacked together a quick Groovy script and was able to save another 20% space by doing a custom "map reduce" on the numerical keys you are using. It went from 18mb -> 14mb for me. You can find the Groovy Gist here https://gist.github.com/1337648

Oh, range() is considered harmful. http://rhodesmill.org/brandon/2012/counting-without-counting/

@pcdinh http://meyerweb.com/eric/comment/chech.html

i have used the above code on window 7 ultimate 32 bit system. i tested the code with redis-normal and redis-hashes keys. when i used redis normal it took 54 MB while redis-hashes took 69 MB. i store hash key like HSET media:1 1 7742896.https://gist.github.com/aman2626/d9170dea4c2716a2c7a07c3db9e67285

python version:3.3
redis-py version:redis-2.10.5-py3.3

i think 69 MB is the total amount of space taken by redis (54 MB with normal redis + 15 MB with hash key). am i correct?