*How fast is Redis?

Redis includes the redis-benchmark utility that simulates running commands done by N clients at the same time sending M total queries (it is similar to the Apache's ab utility). Below you'll find the full output of a benchmark executed against a Linux box.

The following options are supported:

Usage: redis-benchmark [-h <host>] [-p <port>] [-c <clients>] [-n <requests]> [-k <boolean>]

 -h <hostname>      Server hostname (default
 -p <port>          Server port (default 6379)
 -s <socket>        Server socket (overrides host and port)
 -c <clients>       Number of parallel connections (default 50)
 -n <requests>      Total number of requests (default 10000)
 -d <size>          Data size of SET/GET value in bytes (default 2)
 -k <boolean>       1=keep alive 0=reconnect (default 1)
 -r <keyspacelen>   Use random keys for SET/GET/INCR, random values for SADD
  Using this option the benchmark will get/set keys
  in the form mykey_rand:000000012456 instead of constant
  keys, the <keyspacelen> argument determines the max
  number of values for the random number. For instance
  if set to 10 only rand:000000000000 - rand:000000000009
  range will be allowed.
 -P <numreq>        Pipeline <numreq> requests. Default 1 (no pipeline).
 -q                 Quiet. Just show query/sec values
 --csv              Output in CSV format
 -l                 Loop. Run the tests forever
 -t <tests>         Only run the comma-separated list of tests. The test
                    names are the same as the ones produced as output.
 -I                 Idle mode. Just open N idle connections and wait.

You need to have a running Redis instance before launching the benchmark. A typical example would be:

redis-benchmark -q -n 100000

Using this tool is quite easy, and you can also write your own benchmark, but as with any benchmarking activity, there are some pitfalls to avoid.

*Running only a subset of the tests

You don't need to run all the default tests every time you execute redis-benchmark. The simplest thing to select only a subset of tests is to use the -t option like in the following example:

$ redis-benchmark -t set,lpush -n 100000 -q
SET: 74239.05 requests per second
LPUSH: 79239.30 requests per second

In the above example we asked to just run test the SET and LPUSH commands, in quiet mode (see the -q switch).

It is also possible to specify the command to benchmark directly like in the following example:

$ redis-benchmark -n 100000 -q script load "redis.call('set','foo','bar')"
script load redis.call('set','foo','bar'): 69881.20 requests per second

*Selecting the size of the key space

By default the benchmark runs against a single key. In Redis the difference between such a synthetic benchmark and a real one is not huge since it is an in-memory system, however it is possible to stress cache misses and in general to simulate a more real-world work load by using a large key space.

This is obtained by using the -r switch. For instance if I want to run one million SET operations, using a random key for every operation out of 100k possible keys, I'll use the following command line:

$ redis-cli flushall

$ redis-benchmark -t set -r 100000 -n 1000000
====== SET ======
  1000000 requests completed in 13.86 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1

99.76% `<=` 1 milliseconds
99.98% `<=` 2 milliseconds
100.00% `<=` 3 milliseconds
100.00% `<=` 3 milliseconds
72144.87 requests per second

$ redis-cli dbsize
(integer) 99993

*Using pipelining

By default every client (the benchmark simulates 50 clients if not otherwise specified with -c) sends the next command only when the reply of the previous command is received, this means that the server will likely need a read call in order to read each command from every client. Also RTT is paid as well.

Redis supports /topics/pipelining, so it is possible to send multiple commands at once, a feature often exploited by real world applications. Redis pipelining is able to dramatically improve the number of operations per second a server is able do deliver.

This is an example of running the benchmark in a Macbook air 11" using a pipelining of 16 commands:

$ redis-benchmark -n 1000000 -t set,get -P 16 -q
SET: 403063.28 requests per second
GET: 508388.41 requests per second

Using pipelining results in a significant increase in performance.

*Pitfalls and misconceptions

The first point is obvious: the golden rule of a useful benchmark is to only compare apples and apples. Different versions of Redis can be compared on the same workload for instance. Or the same version of Redis, but with different options. If you plan to compare Redis to something else, then it is important to evaluate the functional and technical differences, and take them in account.

A common misconception is that redis-benchmark is designed to make Redis performances look stellar, the throughput achieved by redis-benchmark being somewhat artificial, and not achievable by a real application. This is actually plain wrong.

The redis-benchmark program is a quick and useful way to get some figures and evaluate the performance of a Redis instance on a given hardware. However, by default, it does not represent the maximum throughput a Redis instance can sustain. Actually, by using pipelining and a fast client (hiredis), it is fairly easy to write a program generating more throughput than redis-benchmark. The default behavior of redis-benchmark is to achieve throughput by exploiting concurrency only (i.e. it creates several connections to the server). It does not use pipelining or any parallelism at all (one pending query per connection at most, and no multi-threading).

To run a benchmark using pipelining mode (and achieve higher throughput), you need to explicitly use the -P option. Please note that it is still a realistic behavior since a lot of Redis based applications actively use pipelining to improve performance.

Finally, the benchmark should apply the same operations, and work in the same way with the multiple data stores you want to compare. It is absolutely pointless to compare the result of redis-benchmark to the result of another benchmark program and extrapolate.

For instance, Redis and memcached in single-threaded mode can be compared on GET/SET operations. Both are in-memory data stores, working mostly in the same way at the protocol level. Provided their respective benchmark application is aggregating queries in the same way (pipelining) and use a similar number of connections, the comparison is actually meaningful.

This perfect example is illustrated by the dialog between Redis (antirez) and memcached (dormando) developers.

antirez 1 - On Redis, Memcached, Speed, Benchmarks and The Toilet

dormando - Redis VS Memcached (slightly better bench)

antirez 2 - An update on the Memcached/Redis benchmark

You can see that in the end, the difference between the two solutions is not so staggering, once all technical aspects are considered. Please note both Redis and memcached have been optimized further after these benchmarks.

Finally, when very efficient servers are benchmarked (and stores like Redis or memcached definitely fall in this category), it may be difficult to saturate the server. Sometimes, the performance bottleneck is on client side, and not server-side. In that case, the client (i.e. the benchmark program itself) must be fixed, or perhaps scaled out, in order to reach the maximum throughput.

*Factors impacting Redis performance

There are multiple factors having direct consequences on Redis performance. We mention them here, since they can alter the result of any benchmarks. Please note however, that a typical Redis instance running on a low end, untuned box usually provides good enough performance for most applications.

Data size impact

NUMA chart

connections chart

*Other things to consider

One important goal of any benchmark is to get reproducible results, so they can be compared to the results of other tests.

*Benchmark results on different virtualized and bare-metal servers.

Intel(R) Xeon(R) CPU E5520 @ 2.27GHz (with pipelining)

$ ./redis-benchmark -r 1000000 -n 2000000 -t get,set,lpush,lpop -P 16 -q
SET: 552028.75 requests per second
GET: 707463.75 requests per second
LPUSH: 767459.75 requests per second
LPOP: 770119.38 requests per second

Intel(R) Xeon(R) CPU E5520 @ 2.27GHz (without pipelining)

$ ./redis-benchmark -r 1000000 -n 2000000 -t get,set,lpush,lpop -q
SET: 122556.53 requests per second
GET: 123601.76 requests per second
LPUSH: 136752.14 requests per second
LPOP: 132424.03 requests per second

Linode 2048 instance (with pipelining)

$ ./redis-benchmark -r 1000000 -n 2000000 -t get,set,lpush,lpop -q -P 16
SET: 195503.42 requests per second
GET: 250187.64 requests per second
LPUSH: 230547.55 requests per second
LPOP: 250815.16 requests per second

Linode 2048 instance (without pipelining)

$ ./redis-benchmark -r 1000000 -n 2000000 -t get,set,lpush,lpop -q
SET: 35001.75 requests per second
GET: 37481.26 requests per second
LPUSH: 36968.58 requests per second
LPOP: 35186.49 requests per second

*More detailed tests without pipelining

$ redis-benchmark -n 100000

====== SET ======
  100007 requests completed in 0.88 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1

58.50% <= 0 milliseconds
99.17% <= 1 milliseconds
99.58% <= 2 milliseconds
99.85% <= 3 milliseconds
99.90% <= 6 milliseconds
100.00% <= 9 milliseconds
114293.71 requests per second

====== GET ======
  100000 requests completed in 1.23 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1

43.12% <= 0 milliseconds
96.82% <= 1 milliseconds
98.62% <= 2 milliseconds
100.00% <= 3 milliseconds
81234.77 requests per second

====== INCR ======
  100018 requests completed in 1.46 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1

32.32% <= 0 milliseconds
96.67% <= 1 milliseconds
99.14% <= 2 milliseconds
99.83% <= 3 milliseconds
99.88% <= 4 milliseconds
99.89% <= 5 milliseconds
99.96% <= 9 milliseconds
100.00% <= 18 milliseconds
68458.59 requests per second

====== LPUSH ======
  100004 requests completed in 1.14 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1

62.27% <= 0 milliseconds
99.74% <= 1 milliseconds
99.85% <= 2 milliseconds
99.86% <= 3 milliseconds
99.89% <= 5 milliseconds
99.93% <= 7 milliseconds
99.96% <= 9 milliseconds
100.00% <= 22 milliseconds
100.00% <= 208 milliseconds
88109.25 requests per second

====== LPOP ======
  100001 requests completed in 1.39 seconds
  50 parallel clients
  3 bytes payload
  keep alive: 1

54.83% <= 0 milliseconds
97.34% <= 1 milliseconds
99.95% <= 2 milliseconds
99.96% <= 3 milliseconds
99.96% <= 4 milliseconds
100.00% <= 9 milliseconds
100.00% <= 208 milliseconds
71994.96 requests per second

Notes: changing the payload from 256 to 1024 or 4096 bytes does not change the numbers significantly (but reply packets are glued together up to 1024 bytes so GETs may be slower with big payloads). The same for the number of clients, from 50 to 256 clients I got the same numbers. With only 10 clients it starts to get a bit slower.

You can expect different results from different boxes. For example a low profile box like Intel core duo T5500 clocked at 1.66 GHz running Linux 2.6 will output the following:

$ ./redis-benchmark -q -n 100000
SET: 53684.38 requests per second
GET: 45497.73 requests per second
INCR: 39370.47 requests per second
LPUSH: 34803.41 requests per second
LPOP: 37367.20 requests per second

Another one using a 64-bit box, a Xeon L5420 clocked at 2.5 GHz:

$ ./redis-benchmark -q -n 100000
PING: 111731.84 requests per second
SET: 108114.59 requests per second
GET: 98717.67 requests per second
INCR: 95241.91 requests per second
LPUSH: 104712.05 requests per second
LPOP: 93722.59 requests per second

*Example of benchmark results with optimized high-end server hardware

Using a unix domain socket:

$ numactl -C 6 ./redis-benchmark -q -n 100000 -s /tmp/redis.sock -d 256
PING (inline): 200803.22 requests per second
PING: 200803.22 requests per second
MSET (10 keys): 78064.01 requests per second
SET: 198412.69 requests per second
GET: 198019.80 requests per second
INCR: 200400.80 requests per second
LPUSH: 200000.00 requests per second
LPOP: 198019.80 requests per second
SADD: 203665.98 requests per second
SPOP: 200803.22 requests per second
LPUSH (again, in order to bench LRANGE): 200000.00 requests per second
LRANGE (first 100 elements): 42123.00 requests per second
LRANGE (first 300 elements): 15015.02 requests per second
LRANGE (first 450 elements): 10159.50 requests per second
LRANGE (first 600 elements): 7548.31 requests per second

Using the TCP loopback:

$ numactl -C 6 ./redis-benchmark -q -n 100000 -d 256
PING (inline): 145137.88 requests per second
PING: 144717.80 requests per second
MSET (10 keys): 65487.89 requests per second
SET: 142653.36 requests per second
GET: 142450.14 requests per second
INCR: 143061.52 requests per second
LPUSH: 144092.22 requests per second
LPOP: 142247.52 requests per second
SADD: 144717.80 requests per second
SPOP: 143678.17 requests per second
LPUSH (again, in order to bench LRANGE): 143061.52 requests per second
LRANGE (first 100 elements): 29577.05 requests per second
LRANGE (first 300 elements): 10431.88 requests per second
LRANGE (first 450 elements): 7010.66 requests per second
LRANGE (first 600 elements): 5296.61 requests per second