data size for performance test in tpch-v2.md

[jeeffy]

The docs/benchmark/tpch-v2.md is splendid, it shows that the performance of 2.1 has been greatly improved. But data size in each test table makes me confused. Can you show the following data?

select count(*) customer from customer;
select count(*) region from region;
select count(*) nation from nation;
select count(*) supplier from supplier;
select count(*) part from part;
select count(*) partsupp from partsupp;
select count(*) orders from orders;
select count(*) lineitem from lineitem;

[zz-jason]

@jeeffy the scale is 50, https://github.com/pingcap/docs/blob/master/benchmark/tpch-v2.md#tidb-tpc-h-50g-performance-test-report-v21

[jeeffy]

@zz-jason Thank you for reply. I know data size is 50G, but the amount of data is what I care about, which is displayed as Table size in https://github.com/pingcap/docs/blob/master/benchmark/sysbench.md.

[winoros]

scale factor is one factor of TPC-H's official data generation tool DBGEN.
Once the factor is defined, the size of each table is always the same.

This img is the screenshot of TPC-H's official document. It tells the table size of scale = 1.
When scale=50, the size of the large tables like LINEITEM will just be 50 times with small accuracy error.
So we usually use scale to indicate the data size rather than COUNT(*) statement