For this quick start tutorial, you need:

Trying it out

You can play with it a bit using redis-cli:

Connect to redis.

$ redis-cli -p 12543>

Run these commands:> BF.ADD bloom kirk
 1) (integer) 1> BF.ADD bloom redis
 1) (integer) 1> BF.EXISTS bloom kirk
 (integer) 1> BF.EXISTS bloom redis
 (integer) 1> BF.EXISTS bloom nonexist
 (integer) 0> BF.EXISTS bloom que?
 (integer) 0>> BF.MADD bloom elem1 elem2 elem3
 1) (integer) 1
 2) (integer) 1
 3) (integer) 1> BF.MEXISTS bloom elem1 elem2 elem3
 1) (integer) 1
 2) (integer) 1
 3) (integer) 1

You can also create a custom Bloom filter. The BF.ADD command creates a new Bloom filter suitable for a small-ish number of items. This consumes less memory but may not be ideal for large filters. In that case:> BF.RESERVE largebloom 0.0001 1000000
 OK> BF.ADD largebloom kirk
 1) (integer) 1

Using Cuckoo filters in Redis Enterprise Software

Cuckoo filters can also be used as part of the RedisBloom module. You can play with it using redis-cli:> CF.ADD newcuckoo redis
(integer) 1> CF.EXISTS newcuckoo redis
(integer) 1> CF.EXISTS newcuckoo notpresent
(integer) 0> CF.DEL newcuckoo redis
(integer) 1

Debugging Bloom filters

Finally, I added a BF.DEBUG command, to see exactly how the filter is being utilized:> BF.DEBUG test
 1) "size:987949"
 2) "bytes:239627 bits:1917011 hashes:14 capacity:100000 size:100000 ratio:0.0001"
 3) "bytes:551388 bits:4411101 hashes:16 capacity:200000 size:200000 ratio:2.5e-05"
 4) "bytes:1319180 bits:10553436 hashes:19 capacity:400000 size:400000 ratio:3.125e-06"
 5) "bytes:3215438 bits:25723497 hashes:23 capacity:800000 size:287949 ratio:1.95313e-07"

This outputs the total number of elements as the first result, and then a list of details for each filter in the chain. As you can see, whenever a new filter is added, its capacity grows exponentially and the strictness for errors increases.

Note that this filter chain also uses a total of 5MB. This is still much more space efficient than alternative solutions, since we’re still at about 5 bytes per element, and the uppermost filter is only at about 12% utilization. Had the initial capacity been greater, more space would have been saved and lookups would have been quicker.

More info