database/doris
2
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

[Enhancement](topn) support two phase read for topn query (#15642)

Browse Source 
This PR optimize topn query like `SELECT * FROM tableX ORDER BY columnA ASC/DESC LIMIT N`.

TopN is is compose of SortNode and ScanNode, when user table is wide like 100+ columns the order by clause is just a few columns.But ScanNode need to scan all data from storage engine even if the limit is very small.This may lead to lots of read amplification.So In this PR I devide TopN query into two phase:
1. The first phase we just need to read `columnA`'s data from storage engine along with an extra RowId column called `__DORIS_ROWID_COL__`.The other columns are pruned from ScanNode.
2. The second phase I put it in the ExchangeNode beacuase it's the central node for topn nodes in the cluster.The ExchangeNode will spawn a RPC to other nodes using the RowIds(sorted and limited from SortNode) read from the first phase and read row by row from storage engine.

After the second phase read, Block will contain all the data needed for the query
This commit is contained in:
lihangyu
2023-01-19 10:01:33 +08:00
committed by GitHub
parent c7a72436e6
commit 3894de49d2
53 changed files with 829 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

32
be/src/vec/exec/vexchange_node.cpp
View File

@ -17,12 +17,15 @@
#include "vec/exec/vexchange_node.h"
#include "common/consts.h"
#include "exec/rowid_fetcher.h"
#include "pipeline/exec/exchange_source_operator.h"
#include "pipeline/pipeline.h"
#include "pipeline/pipeline_fragment_context.h"
#include "runtime/exec_env.h"
#include "runtime/runtime_state.h"
#include "runtime/thread_context.h"
#include "util/defer_op.h"
#include "vec/runtime/vdata_stream_mgr.h"
#include "vec/runtime/vdata_stream_recvr.h"
@ -45,10 +48,15 @@ Status VExchangeNode::init(const TPlanNode& tnode, RuntimeState* state) {
if (!_is_merging) {
return Status::OK();
}
RETURN_IF_ERROR(_vsort_exec_exprs.init(tnode.exchange_node.sort_info, _pool));
_is_asc_order = tnode.exchange_node.sort_info.is_asc_order;
_nulls_first = tnode.exchange_node.sort_info.nulls_first;
if (tnode.exchange_node.__isset.nodes_info) {
_nodes_info = _pool->add(new DorisNodesInfo(tnode.exchange_node.nodes_info));
}
_use_two_phase_read = tnode.exchange_node.sort_info.__isset.use_two_phase_read &&
tnode.exchange_node.sort_info.use_two_phase_read;
return Status::OK();
}
@ -87,6 +95,19 @@ Status VExchangeNode::open(RuntimeState* state) {
return Status::OK();
}
Status VExchangeNode::_second_phase_fetch_data(RuntimeState* state, Block* final_block) {
auto row_id_col = final_block->get_by_position(final_block->columns() - 1);
auto tuple_desc = _row_descriptor.tuple_descriptors()[0];
RowIDFetcher id_fetcher(tuple_desc, state);
RETURN_IF_ERROR(id_fetcher.init(_nodes_info));
MutableBlock materialized_block(_row_descriptor.tuple_descriptors(), final_block->rows());
// fetch will sort block by sequence of ROWID_COL
RETURN_IF_ERROR(id_fetcher.fetch(row_id_col.column, &materialized_block));
// Notice swap may change the structure of final_block
final_block->swap(materialized_block.to_block());
return Status::OK();
}
Status VExchangeNode::get_next(RuntimeState* state, Block* block, bool* eos) {
INIT_AND_SCOPE_GET_NEXT_SPAN(state->get_tracer(), _get_next_span, "VExchangeNode::get_next");
SCOPED_TIMER(runtime_profile()->total_time_counter());
@ -97,6 +118,12 @@ Status VExchangeNode::get_next(RuntimeState* state, Block* block, bool* eos) {
_is_ready = true;
return Status::OK();
}
if (_use_two_phase_read) {
// Block structure may be changed by calling _second_phase_fetch_data() before.
// So we should clear block before _stream_recvr->get_next, since
// blocks in VSortedRunMerger may not compatible with this block.
block->clear();
}
auto status = _stream_recvr->get_next(block, eos);
if (block != nullptr) {
if (!_is_merging) {
@ -119,6 +146,9 @@ Status VExchangeNode::get_next(RuntimeState* state, Block* block, bool* eos) {
}
COUNTER_SET(_rows_returned_counter, _num_rows_returned);
}
if (_use_two_phase_read && block->rows() > 0) {
RETURN_IF_ERROR(_second_phase_fetch_data(state, block));
}
return status;
}