FBU Queries (PDSW19)

1. Clone and build skyhook.

2. cd build/

3. Get the data sets.

wget https://users.soe.ucsc.edu/~kdahlgren/pdsw19/testdata/dataset_arity3_50000_rows.txt ;
wget https://users.soe.ucsc.edu/~kdahlgren/pdsw19/testdata/dataset_arity3_500000_rows.txt ;
wget https://users.soe.ucsc.edu/~kdahlgren/pdsw19/testdata/dataset_arity3_5000000_rows.txt ;
wget https://users.soe.ucsc.edu/~kdahlgren/pdsw19/testdata/dataset_arity4_42000_rows.txt ;
wget https://users.soe.ucsc.edu/~kdahlgren/pdsw19/testdata/dataset_arity4_420000_rows.txt ;
wget https://users.soe.ucsc.edu/~kdahlgren/pdsw19/testdata/dataset_arity4_4200000_rows.txt ;

4. (Re)launch the virtual cluster, if applicable.

../src/stop.sh;
make -j12 vstart;
../src/stop.sh;
../src/vstart.sh -d -n -x;
bin/rados mkpool tpchflatbuf ;
bin/ceph osd pool set tpchflatbuf size 1 ;

5. Write a data set into a ceph object. See the following syntax.

parameter	description
--filename	the name of the file containing the pipe-delimited relation to save in ceph
--write_type	rows or cols?
--debug	yes or no?
--schema_datatypes	comma-delimited list of data types. must match table schema
--schema_attnames	comma-delimited list of attribute names
--table_name	name you want to give the table
--nrows	number of rows in the input file
--ncols	number of columns in the input file
--targetoid	name of the ceph object this is going to occupy
--targetpool	name of the ceph pool to use
--writeto	ceph or disk?
--targetformat	SFT_FLATBUF_UNION_ROW or SFT_FLATBUF_UNION_COL?

# FBU_Rows arity-3 1mb (50,000 rows) => obj.0
bin/fbwriter_fbu --filename dataset_arity3_50000_rows.txt --write_type rows --debug yes --schema_datatypes int,float,string --schema_attnames att0,att1,att2 --table_name arity3_1mb --nrows 50000 --ncols 3 --targetoid obj.0 --targetpool tpchflatbuf --writeto ceph --targetformat SFT_FLATBUF_UNION_ROW ;

# FBU_Rows arity-3 1mb (500,000 rows) => obj.0
bin/fbwriter_fbu --filename dataset_arity3_500000_rows.txt --write_type rows --debug yes --schema_datatypes int,float,string --schema_attnames att0,att1,att2 --table_name arity3_100mb --nrows 500000 --ncols 3 --targetoid obj.0 --targetpool tpchflatbuf --writeto ceph --targetformat SFT_FLATBUF_UNION_ROW ;

# FBU_Rows arity-3 1mb (5,000,000 rows) => obj.0
bin/fbwriter_fbu --filename dataset_arity3_500000_rows.txt --write_type rows --debug yes --schema_datatypes int,float,string --schema_attnames att0,att1,att2 --table_name arity3_100mb --nrows 500000 --ncols 3 --targetoid obj.0 --targetpool tpchflatbuf --writeto ceph --targetformat SFT_FLATBUF_UNION_ROW ;

# FBU_Cols arity-3 1mb (50,000 rows) => obj.0
bin/fbwriter_fbu --filename dataset_arity3_50000_rows.txt --write_type cols --debug yes --schema_datatypes int,float,string --schema_attnames att0,att1,att2 --table_name arity3_1mb --nrows 50000 --ncols 3 --targetoid obj.0 --targetpool tpchflatbuf --writeto ceph --cols_per_fb 1 --targetformat SFT_FLATBUF_UNION_COL ;

# FBU_Cols arity-3 1mb (500,000 rows) => obj.0
bin/fbwriter_fbu --filename dataset_arity3_500000_rows.txt --write_type cols --debug yes --schema_datatypes int,float,string --schema_attnames att0,att1,att2 --table_name arity3_1mb --nrows 500000 --ncols 3 --targetoid obj.0 --targetpool tpchflatbuf --writeto ceph --cols_per_fb 1 --targetformat SFT_FLATBUF_UNION_COL ;

# FBU_Cols arity-3 1mb (5,000,000 rows) => obj.0
bin/fbwriter_fbu --filename dataset_arity3_5000000_rows.txt --write_type cols --debug yes --schema_datatypes int,float,string --schema_attnames att0,att1,att2 --table_name arity3_1mb --nrows 5000000 --ncols 3 --targetoid obj.0 --targetpool tpchflatbuf --writeto ceph --cols_per_fb 1 --targetformat SFT_FLATBUF_UNION_COL ;

6. Run your queries. Note Skyhook currently only queries objects with the "obj." suffix.

bin/run-query --num-objs 1 --pool tpchflatbuf --wthreads 1 --qdepth 10 --query flatbuf --select "*" ;

bin/run-query --num-objs 1 --pool tpchflatbuf --wthreads 1 --qdepth 10 --query flatbuf --select "att0,lt,25;att1,lt,25.0;"  --project-cols att0,att1,att2,att3 --data-schema "0 8 0 0 ATT0 ; 1 12 0 0 ATT1 ; 2 15 0 0 ATT2 ;" ;

bin/run-query --num-objs 1 --pool tpchflatbuf --wthreads 1 --qdepth 10 --query flatbuf --select "att0,lt,25;att1,lt,25.0;"  --project-cols att0,att2 --data-schema "0 8 0 0 ATT0 ; 1 12 0 0 ATT1 ; 2 15 0 0 ATT2 ;" ;

bin/run-query --num-objs 1 --pool tpchflatbuf --wthreads 1 --qdepth 10 --query flatbuf --select-preds ";att0,sum,0;" --table-name "atable" --data-schema "0 8 0 0 ATT0 ; 1 12 0 0 ATT1 ; 2 15 0 0 ATT2 ;" ;