Target

This repo targets to evaluate kudu point lookup performance. The table has more than 100 columns, it DDL is like:

CREATE TABLE kudu_poc.kudu_marketing_item(
  item_id long not null,
  curnt_price double,
  price_update_time timestamp
  ...
)
USING kudu OPTIONS (
  kudu.tableHashPartitions 'item_id,512'
)

Build

In order to get the latest kudu-client 1.15.0, which is not published yet, we have to build it manually and install it to local M2:

mvn install:install-file -Dfile=`pwd`/local-maven-repo/kudu-client-1.15.0.jar -DgroupId=org.apache.kudu -DartifactId=kudu-client -Dversion=1.15.0 -Dpackaging=jar -DgeneratePom=true

mvn clean package

Run

java -jar kudu-mt-query-itemid-1.0-SNAPSHOT.jar -f item_ids.txt -k xxx0:7051,xxx1:7051,xxx2:7051 -t kudu_table -i 200

the item_ids.txt contains item_id:

114419336369
114419348892
114419348924
114419366922
114419369029
114419394436
114419395353
114419396218
114419399917
114419460482
114419463335
114419469847
114419477181
114419478283
114419502018
114419513353
114419520145

Output

iteration: 1
Table 'kudu_marketing_item_store_nov13' colums: 134
Run 1 scans take 1056 ms
The SQL is like "select curnt_price from kudu_marketing_item_store_nov13 where item_id = ?"
item: 114417465925 price: 30.0
item: 114417507056 price: 7.99
item: 114417508976 price: 190.0
item: 114417509062 price: 12.71
item: 114417512889 price: 10.1
item: 114417515816 price: 37.06
item: 114417517279 price: 14.08
...
item: 114419513353 price: 39.99
item: 114419520145 price: 52.35