Benching Riak's performance with the Python client.

riak_bench.py

import datetime
import json
import random
import riak
import time
import uuid


URLS = (
    '/',
    '/news/',
    '/news/2010/08/13/first-post/',
    '/news/2010/08/13/second-post/',
    '/weblogs/slc_punk/2010/08/14/the_show/',
)

ACCESSED = (
    datetime.datetime(2010, 8, 13, 7, 35, 0),
    datetime.datetime(2010, 8, 13, 9, 16, 0),
    datetime.datetime(2010, 8, 14, 10, 47, 0),
    datetime.datetime(2010, 8, 14, 18, 10, 0),
    datetime.datetime(2010, 8, 15, 8, 0, 0),
)

IPS = (
    '127.0.0.1',
    '192.168.0.1',
    '107.123.4.7',
    '244.34.156.5',
    '10.0.1.2',
)

TOTAL_HITS = 10000
# TOTAL_HITS = 10000000


def make_timestamp(date):
    return time.mktime(date.timetuple())

def get_access():
    return {
        'url': random.choice(URLS),
        'accessed': make_timestamp(random.choice(ACCESSED)),
        'ip': random.choice(IPS),
    }

def time_it(func):
    start_time = time.time()
    result = func()
    return (time.time() - start_time, result)


# Riak implementation
# HTTP Transport
# client = riak.RiakClient()
# Protobuf Transport
client = riak.RiakClient(port=8087, transport_class=riak.RiakPbcTransport)
bucket = client.bucket('test_pageviews')

def write_docs():
    for i in xrange(0, TOTAL_HITS):
        if i % 1000 == 0:
            print '  Inserting %s' % i
        
        the_id = uuid.uuid1()
        obj = bucket.new(str(the_id), data=get_access())
        obj.store()

def read_1000():
    query = riak.RiakMapReduce(client).add('test_pageviews')
    query.map("function(v){ return [JSON.parse(v.values[0].data).id]; }")
    return len(query.run()[:1000])

def query_by_url():
    query = riak.RiakMapReduce(client).add('test_pageviews')
    query.map('function(v) { var data = JSON.parse(v.values[0].data); if(data.url == "/news/2010/08/13/first-post/") { return [1]; } else { return [0]; } }').reduce('Riak.reduceSum')
    result = query.run()
    if len(result):
        return result[0]
    else:
        return 0

def query_by_date():
    start_date = make_timestamp(datetime.date(2010, 8, 13))
    end_date = make_timestamp(datetime.date(2010, 8, 14))
    query = riak.RiakMapReduce(client).add('test_pageviews')
    query.map('function(v) { var data = JSON.parse(v.values[0].data); if(data.accessed >= %s && data.accessed < %s) { return [1]; } return [0]; }' % (start_date, end_date)).reduce('Riak.reduceSum')
    result = query.run()
    if len(result):
        return result[0]
    else:
        return 0

def query_by_url_and_date():
    start_date = make_timestamp(datetime.date(2010, 8, 13))
    end_date = make_timestamp(datetime.date(2010, 8, 14))
    query = riak.RiakMapReduce(client).add('test_pageviews')
    query.map('function(v) { var data = JSON.parse(v.values[0].data); if(data.accessed >= %s && data.accessed < %s) { if(data.url == "/news/2010/08/13/first-post/") { return [1]; } } return [0]; }' % (start_date, end_date)).reduce('Riak.reduceSum')
    result = query.run()
    if len(result):
        return result[0]
    else:
        return 0

def nuke_it():
    print "  Getting keys to nuke at %s" % datetime.datetime.now()
    keys = bucket.get_keys()
    print "  Got keys to nuke by %s" % datetime.datetime.now()
    
    for count, key in enumerate(keys):
        if count % 1000 == 0:
            print '  Nuking %s' % count
        
        the_obj = bucket.get(key)
        the_obj.delete()


if __name__ == '__main__':
    print "Writing %s records: %.2f seconds" % (TOTAL_HITS, time_it(write_docs)[0])
    print "Read 1000: %.2f seconds" % time_it(read_1000)[0]
    print "Query by URL: %.2f seconds (%s docs)" % time_it(query_by_url)
    print "Query by Date: %.2f seconds (%s docs)" % time_it(query_by_date)
    print "Query by URL & Date: %.2f seconds (%s docs)" % time_it(query_by_url_and_date)
    print "Nuke it: %.2f seconds" % time_it(nuke_it)[0]

RiakHttp10k.txt

[daniel@Europa:Desktop]: python riak_test.py 
  Inserting 0
  Inserting 1000
  Inserting 2000
  Inserting 3000
  Inserting 4000
  Inserting 5000
  Inserting 6000
  Inserting 7000
  Inserting 8000
  Inserting 9000
Writing 10000 records: 36.22 seconds
Read 1000: 4.81 seconds
Query by URL: 5.04 seconds (2031 docs)
Query by Date: 4.80 seconds (3949 docs)
Query by URL & Date: 4.92 seconds (797 docs)
  Getting keys to nuke at 2010-10-28 19:43:20.862265
  Got keys to nuke by 2010-10-28 19:44:11.839581
  Nuking 0
  Nuking 1000
  Nuking 2000
  Nuking 3000
  Nuking 4000
  Nuking 5000
  Nuking 6000
  Nuking 7000
  Nuking 8000
  Nuking 9000
Nuke it: 120.11 seconds

RiakPB10k.txt

[daniel@Europa:Desktop]: python riak_test.py 
  Inserting 0
  Inserting 1000
  Inserting 2000
  Inserting 3000
  Inserting 4000
  Inserting 5000
  Inserting 6000
  Inserting 7000
  Inserting 8000
  Inserting 9000
Writing 10000 records: 12.38 seconds
Read 1000: 5.86 seconds
Query by URL: 4.65 seconds (1969 docs)
Query by Date: 4.85 seconds (4016 docs)
Query by URL & Date: 4.73 seconds (790 docs)
  Getting keys to nuke at 2010-10-28 22:06:18.746901
  Got keys to nuke by 2010-10-28 22:06:18.967677
  Nuking 0
  Nuking 1000
  Nuking 2000
  Nuking 3000
  Nuking 4000
  Nuking 5000
  Nuking 6000
  Nuking 7000
  Nuking 8000
  Nuking 9000
Nuke it: 23.81 seconds