tree cca554738cf59758274950667fca13cbfd27827e
parent 0ce77c56ce07b657ddd4f58f5f36692e3c31490d
author ajantha-bhat <ajanthabhat@gmail.com> 1595416395 +0530
committer akashrn5 <akashnilugal@gmail.com> 1597741425 +0530

[CARBONDATA-3919] Improve concurrent query performance

Why is this PR needed?
1. when 500 queries executed concurrently.
checkIfRefreshIsNeeded method was synchronized. so only one thread was working at a time.
But actually synchronization is required only when schema modified to drop tables. Not
for whole function.
2. TokenCache.obtainTokensForNamenodes was causing a performance bottleneck for concurrent
queries.so, removed it

What changes were proposed in this PR?
1. Synchronize only remove table part. Observed 500 query total performance improved from
10s to 3 seconds in cluster.
2. Avoid calling the API.

This closes #3858