yunbalabs / redis_proxy

A distributed proxy for redis in Erlang.

redis_proxy

A distributed proxy for redis in Erlang. It's based on distributed_proxy.

Usage

Build

$ make generate

Start

$ ./rel/redis_proxy/bin/redis_proxy start

Play with redis-cli

$ redis-cli -p 6380
127.0.0.1:6380> SET 1 123
OK
127.0.0.1:6380> GET 1
"123"

Supported Commands

SET key value

set 1 1

GET key

get 1

MGET key [key ...]

mget 1 2 3

DBSIZE

DBSIZE

Cluster

Join a cluster

$ ./rel/redis_proxy/bin/redis_proxy_admin.sh join any_live_node_in_the_cluster@hostname

Cluster map

$ ./rel/redis_proxy/bin/redis_proxy_admin.sh map

Status

Node status

$ ./rel/redis_proxy/bin/redis_proxy_admin.sh status

Replica status

$ ./rel/redis_proxy/bin/redis_proxy_admin.sh replicas               %% all replicas status
$ ./rel/redis_proxy/bin/redis_proxy_admin.sh replica replica_id     %% a specific replica status
$ ./rel/redis_proxy/bin/redis_proxy_admin.sh locate key             %% replica status about a specific key

Request/Response status

1. Update config (sys.config).

   • Export stat data through exomter_core

       %% exometer core config
       {exometer_core, [
           {report, [
               {reporters, [
                   {exometer_report_influxdb, [{protocol, udp},
                       {host, <<"localhost">>},
                       {port, 8089},
                       {db, <<"udp">>}]}
               ]}
           ]}
       ]}
   

   • Enable stat

       {redis_proxy, [
           ...
           {enable_stat, true}
       ]}
   

2. Restart the application. The stat data will appear in InfluxDB:

   measurement name	data type
   frontend_request	count of the frontend request
   frontend_response	count of the frontend response
   backend_request	count of the backend request
   latency	time(milliseconds) of request

Benchmark

Intel(R) Xeon(R) CPU E5-2630 0 @ 2.30GHz, 2300 MHz x 8 + 32G RAM

• Erlang R16B02

slot_num: 1, replica_size: 1

$ redis-benchmark -t set,get -q -n 10000 -r 10000 -p 6380 -h test.host -c 200
SET: 5293.81 requests per second
GET: 5437.74 requests per second

slot_num: 8, replica_size: 1

$ redis-benchmark -t set,get -q -n 100000 -r 100000 -p 6380 -h test.host -c 200
SET: 16742.01 requests per second
GET: 17543.86 requests per second

slot_num: 16, replica_size: 1

$ redis-benchmark -t set,get -q -n 100000 -r 100000 -p 6380 -h test.host -c 200
SET: 17006.80 requests per second
GET: 17537.71 requests per second

slot_num: 32, replica_size: 1

$ redis-benchmark -t set,get -q -n 100000 -r 100000 -p 6380 -h test.host -c 200
SET: 18653.24 requests per second
GET: 19409.94 requests per second

slot_num: 64, replica_size: 1

$ redis-benchmark -t set,get -q -n 100000 -r 100000 -p 6380 -h test.host -c 200
SET: 18145.53 requests per second
GET: 18667.16 requests per second

slot_num: 32, replica_size: 2

$ redis-benchmark -t set,get -q -n 100000 -r 100000 -p 6380 -h test.host -c 200
SET: 17247.33 requests per second
GET: 19346.10 requests per second

slot_num: 1, replica_size: 1, redis_client: eredis_pool

$ redis-benchmark -t set,get -q -n 100000 -r 100000 -p 6380 -h test.host -c 200
SET: 17479.46 requests per second
GET: 18497.96 requests per second

TODO

Optimize performance

1. eredis_parser:parse
2. distributed_proxy_util:pmap
3. redis_proxy_util:locate_key