When the data is dynamic, the performance is much worse

[Lloyd-Pottiger]

Enhancement

These are the result of running CHBenchmark as below:

TP 100 concurrency

concurrency  | 8 | 16 | 32 | 64 | 128 -- | -- | -- | -- | -- | -- q3 | 329.9 | 1236.4 | 3259.7 | 3780.4 | 2901.2 q10 | 7388.8 | 17403.3 | 46178.9 | 121042.4 | 235424.8 QphH | 3037.8 | 2614.1 | 2162.1 | 1758.5 | 1856.6

TP 10 concurrency

 concurrency | 8 | 16 | 32 | 64 | 128 -- | -- | -- | -- | -- | -- q3 | 316.8 | 1017.8 | 2362.2 | 3118.8 | 2871.2 q10 | 7475 | 16022.4 | 40390.9 | 98414 | 233739.8 QphH | 2989.3 | 2894.6 | 2456.7 | 2148.9 | 1848.6

TP 0 concurrency

concurrency | 8 | 16 | 32 | 64 | 128 -- | -- | -- | -- | -- | -- q3 | 195 | 317.6 | 361.9 | 519.8 | 1017.3 q10 | 7445.8 | 13492.1 | 29838.8 | 60010.6 | 120043.6 QphH | 3123.2 | 3737.4 | 3627.7 | 3675.1 | 3663.2

It appears that data updates have a significant impact on the performance of concurrent query.

[Lloyd-Pottiger]

/assign @CalvinNeo