[HIVE-4147] Slow Hive JDBC in concurrency mode to create/drop table

• Type: Improvement
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.10.0
• Component/s: HiveServer2, JDBC
• Assignee: Brock Noland
• Reporter: Alexander Alten-Lorenz
• Created: 11/Mar/13 06:54
• Updated: 15/Mar/13 18:09

Description

It's very slow using hive jdbc in concurrency mode to create/drop table, which is 20 times slower than using HiveMetatstoreClient.

test steps:
1. create 100 different hive table one by one by using hive jdbc: "create table ......"
2. drop table one by one by using hive jdbc: "drop table ......" and timing
3. create 100 different hive table one by one by using hive jdbc: "create table ......"
4. drop tables one by one by using new HiveMetatstoreClient().dropTable("default", "table_name") and timing

results
step 2 is 20 times slower than step 4.
basically hive jdbc is 20 times slower than HiveMetatstoreClient not only create/table, but also the same kind of calls.

Dropping tables via this low level API could cause issues if there are any clients concurrently querying the table.