@Override
public RelOptCost computeSelfCost(RelOptPlanner planner) {
if (PrelUtil.getSettings(getCluster()).useDefaultCosting()) {
// We use multiplier 0.05 for TopN operator, and 0.1 for Sort, to make TopN a preferred
// choice.
return super.computeSelfCost(planner).multiplyBy(.1);
}
RelNode child = this.getInput();
double inputRows = RelMetadataQuery.getRowCount(child);
// int rowWidth = child.getRowType().getPrecision();
int numSortFields = this.collation.getFieldCollations().size();
double cpuCost =
DrillCostBase.COMPARE_CPU_COST
* numSortFields
* inputRows
* (Math.log(inputRows) / Math.log(2));
double diskIOCost =
0; // assume in-memory for now until we enforce operator-level memory constraints
// TODO: use rowWidth instead of avgFieldWidth * numFields
// avgFieldWidth * numFields * inputRows
double numFields = this.getRowType().getFieldCount();
long fieldWidth =
PrelUtil.getPlannerSettings(planner)
.getOptions()
.getOption(ExecConstants.AVERAGE_FIELD_WIDTH_KEY)
.num_val;
double memCost = fieldWidth * numFields * inputRows;
DrillCostFactory costFactory = (DrillCostFactory) planner.getCostFactory();
return costFactory.makeCost(inputRows, cpuCost, diskIOCost, 0, memCost);
}
@Override
public RelOptCost computeSelfCost(RelOptPlanner planner) {
if (PrelUtil.getSettings(getCluster()).useDefaultCosting()) {
return super.computeSelfCost(planner).multiplyBy(.1);
}
RelNode child = this.getChild();
double inputRows = RelMetadataQuery.getRowCount(child);
int rowWidth = child.getRowType().getFieldCount() * DrillCostBase.AVG_FIELD_WIDTH;
double hashCpuCost = DrillCostBase.HASH_CPU_COST * inputRows * distFields.size();
double svrCpuCost = DrillCostBase.SVR_CPU_COST * inputRows;
double mergeCpuCost =
DrillCostBase.COMPARE_CPU_COST * inputRows * (Math.log(numEndPoints) / Math.log(2));
double networkCost = DrillCostBase.BYTE_NETWORK_COST * inputRows * rowWidth;
DrillCostFactory costFactory = (DrillCostFactory) planner.getCostFactory();
return costFactory.makeCost(inputRows, hashCpuCost + svrCpuCost + mergeCpuCost, 0, networkCost);
}
/// TODO: this method is same as the one for ScanPrel...eventually we should consolidate
/// this and few other methods in a common base class which would be extended
/// by both logical and physical rels.
@Override
public RelOptCost computeSelfCost(final RelOptPlanner planner) {
final ScanStats stats = groupScan.getScanStats(settings);
int columnCount = getRowType().getFieldCount();
double ioCost = 0;
boolean isStarQuery =
Iterables.tryFind(
getRowType().getFieldNames(),
new Predicate<String>() {
@Override
public boolean apply(String input) {
return Preconditions.checkNotNull(input).equals("*");
}
})
.isPresent();
if (isStarQuery) {
columnCount = STAR_COLUMN_COST;
}
// double rowCount = RelMetadataQuery.getRowCount(this);
double rowCount = stats.getRecordCount();
if (rowCount < 1) {
rowCount = 1;
}
if (PrelUtil.getSettings(getCluster()).useDefaultCosting()) {
return planner
.getCostFactory()
.makeCost(rowCount * columnCount, stats.getCpuCost(), stats.getDiskCost());
}
double cpuCost =
rowCount * columnCount; // for now, assume cpu cost is proportional to row count.
// Even though scan is reading from disk, in the currently generated plans all plans will
// need to read the same amount of data, so keeping the disk io cost 0 is ok for now.
// In the future we might consider alternative scans that go against projections or
// different compression schemes etc that affect the amount of data read. Such alternatives
// would affect both cpu and io cost.
DrillCostFactory costFactory = (DrillCostFactory) planner.getCostFactory();
return costFactory.makeCost(rowCount, cpuCost, ioCost, 0);
}
