Overview

A sandbox playgroud with Jedis. Testing different aproach about how to write/read data to Redis.

HOW-TO

mvn clean compile assembly:single
java -jar ./target/redis-playground-1.0-SNAPSHOT-jar-with-dependencies.jar

Test results

test machine : ubuntu 16.04, Intel Core i7-4510U, 6Gb

Redis: v. 3.2.8, default configuraion from sources

case 1: write 10000 records SET, json value

Run 1: avg 236ms.
Run 2: avg 232ms.
Run 3: avg 233ms.

case 2: write 10000 records SET (pipeline), json value

Run 1: avg 59ms.
Run 2: avg 51ms.
Run 3: avg 51ms.

case 3: write 10000 records as one key, json value

Run 1: avg 56ms.
Run 2: avg 48ms.
Run 3: avg 49ms.

case 4: write 10000 recods as HASH per instrument (pipeline) : fields plain values

Run 1: avg 64ms.
Run 2: avg 52ms.
Run 3: avg 56ms.

note: a resource monitor shows more extensive CPU usage. It is easy explain that we do 10x more requests vs cases where we save a recond as a single json string

case 4.1: write 10000 recods as a single HASH where instruments are json fields

Run 1: avg 57ms.
Run 2: avg 51ms.
Run 3: avg 51ms.

case 5: read 10000 records GET (from case 1)

Run 1: avg 226ms.
Run 2: avg 229ms.
Run 3: avg 228ms.

case 6: read 10000 records as one key (from case 3)

Run 1: avg 50ms.
Run 2: avg 50ms.
Run 3: avg 47ms.

case 7: read 10000 records as HASH (from case 4)

Run 1: avg 1546ms.
Run 2: avg 1550ms.
Run 3: avg 1533ms.

case 8: read 10000 records as MGET (from case 1)

Run 1: avg 59ms.
Run 2: avg 55ms.
Run 3: avg 57ms.