Update queries to support ROLLUP (#705)

src/CloudDataWarehouseBenchmark/Cloud-DWB-Derived-from-TPCDS/README.md

@@ -14,16 +14,13 @@ In order to address the enormous range of query types and user behaviors encount
 
 The Cloud DW benchmark consists of the 99 queries from TPC-DS. These have a wide variation of complexity, amount of data scanned, answer set size and elapsed time. Each query asks a business question and includes the corresponding query to answer the question. We have generated TPC-DS datasets and corresponding queries using the official TPC-DS kit from tpc.org. We have also provided sample DDL for each of the tables in the Cloud DW benchmark. This gives you an easy and complete way of setting up your own Cloud DW benchmark on Redshift and running the same queries that we use in our own testing.  While the queries we provide use the default random seed as provided by TPC-DS, you can use the tools in the TPC-DS Benchmark Kit to generate different versions of these queries that have different parameter values.
 
-TPC-DS also permits the use of and provides official query variants for several queries. As Amazon Redshift does not currently support the ROLLUP or GROUPING operators, this benchmark uses the approved SQL variants provided by TPC-DS for queries 5, 14, 18, 22, 27, 35, 36, 67, 70, 77, 80, and 86.
-
 ## Repository Structure
 
 This repository contains a directory for each data scale that we have generated.  For example, the `3TB` directory includes the queries, table definitions, and load SQL for the 3 Terabyte TPC-DS dataset.  The data itself is stored in Amazon S3.
 
 The files in each data scale directory are as follows:
 * `ddl.sql`:  This SQL file creates the TPC-DS tables and then loads the data via COPY from S3. Prior to executing, you will need to replace the strings `<USER_ACCESS_KEY_ID>` and `<USER_SECRET_ACCESS_KEY>` with any valid S3 credentials.  The file also executes `SELECT COUNT(*)` against all tables at the end to show that all data was loaded.
 * `queries/query_0.sql`:  This file contains the 99 queries that are part of the TPC-DS "Power Run".  The "Power Run" is a single-concurrency execution of the official TPC-DS queries.  Note that 4 of the TPC-DS "queries" have 2 parts, so executing the file will result in 103 SQL queries being executed.
-* `queries/query_1.sql` through `queries/query_10.sql`:  These files contain the same 99 queries, but each in a different random order.  The random order is defined by the TPC-DS specification.  These files are used for the TPC-DS "Throughput Run". The "Throughput Run" is a multi-concurrency execution. So for example, to execute a 5-user Throughput Run, the files `query_1.sql` through `query_5.sql` should be executed at the same time against the database.
 
 
 ## Instructions For Use
@@ -37,12 +34,6 @@ To execute the TPC-DS Power Run at a particular data scale (e.g. 3TB), perform t
 
 3. `ddl.sql` will execute `SELECT COUNT(*)` at the end against each table.  Verify that the resulting counts match the values at the end of `ddl.sql`.
 
-4. Execute `query_0.sql` several times sequentially (e.g. five times) and record the run time of each execution.
+4. Execute `query_0.sql` three times and record the best run time.
 
 Note that the S3 data is located in US-EAST-1 so you may get faster load times if your test cluster is located in that region.
-
-### Throughput Run
-
-Once you have the data loaded and have executed the Power Run, you can then execute the Throughput Run using the same data:
-
-5. Simultaneously execute the query files that correspond to the number of concurrent streams you want.  For example, to execute 2 concurrent streams, simultaneously execute `query_1.sql` and `query_2.sql`.  To execute 5 concurrent streams simultaneously execute `query_1.sql` through `query_5.sql`.  The Throughput Run time will be the time elapsed from the start of the first query for the streams to the end of the last query of any stream.

src/CloudDataWarehouseBenchmark/Cloud-DWB-Derived-from-TPCH/README.md

@@ -36,7 +36,7 @@ To execute the TPC-H Power Run at a particular data scale (e.g. 3TB), perform th
 
 2. Connect to the database you created and execute `ddl.sql`.  This may take several hours, depending on the data scale and data warehouse size.
 
-3. Execute `query_0.sql` several times sequentially (e.g. five times) and record the run time of each execution.
+3. Execute `query_0.sql` three times and record the best run time.
 
 Note that the S3 data is located in US-EAST-1 so you may get faster load times if your test cluster is located in that region.