We have Hadoop cluster contains Phoenix table over HBase service and supporting 10k QPS. While running analytic query in the same table like count, group by for some operational purpose it's causing huge fluctuation in QPS(some time's 0 also). So to avoid can we enable any queue for analytical queries so that it won't disturb OLATP resources.
cluster size is 18 nodes( each 128 GB RAM + 32 vCPU) and data size is 300 GB(around 3 billion records).
hbase is not specifically optimized for statistical analysis.
If it is a complex algorithm distributed computing, just the task scheduling will have a lot of time consumption.
You might want to take a closer look at the data physical IO requirements of different business code to identify performance bottlenecks,