GI · k8s原理與實踐

[TOC] # GI GI 算子用于并行執行中，用于按照分區或者按照數據塊迭代整張表。按照迭代數據的粒度劃分，GI 算子包括 PX PARTITION ITERATOR 和 PX BLOCK ITERATOR。 ## PX PARTITION ITERATOR PX PARTITION ITERATOR 算子用于按照分區粒度迭代數據。如下示例中，2 號算子按分區粒度迭代出數據。 ~~~ obclient>CREATE TABLE t (c1 INT, c2 INT) PARTITION BY HASH(c1) PARTITIONS 4; Query OK, 0 rows affected (0.12 sec) obclient>CREATE INDEX idx ON t (c1); Query OK, 0 rows affected (0.12 sec) obclient>EXPLAIN SELECT /*+FULL(t)*/ c1 FROM t\G; *************************** 1. row *************************** Query Plan: ====================================================== |ID|OPERATOR |NAME |EST. ROWS|COST | ------------------------------------------------------ |0 |PX COORDINATOR | |400000 |427257| |1 | EXCHANGE OUT DISTR |:EX10000|400000 |247403| |2 | PX PARTITION ITERATOR| |400000 |247403| |3 | TABLE SCAN |T |400000 |247403| ====================================================== Outputs & filters: ------------------------------------- 0 - output([T.C1], [T.C2]), filter(nil) 1 - output([T.C1], [T.C2]), filter(nil), dop=1 2 - output([T.C1], [T.C2]), filter(nil) 3 - output([T.C1], [T.C2]), filter(nil), access([T.C1], [T.C2]), partitions(p[0-3]) ~~~ 上述示例的執行計劃展示中的 outputs & filters 詳細列出了 PX PARTITION ITERATOR 算子的輸出信息如下： <table data-tag="table" id="table-z8p-ulr-u97" class="table"><colgroup span="1" width="200" data-tag="col" id="col-2ah-s91-nbp" colwidth="1*" colnum="1" colname="col1" style="width:50%" class="col"></colgroup><colgroup span="1" width="540" data-tag="col" id="col-dsi-ufd-faa" colwidth="1*" colnum="2" colname="col2" style="width:50%" class="col"></colgroup><thead id="thead-38c-b02-5d2" class="thead"><tr id="tr-259-6db-u7r"><th id="td-jtf-dem-cqx"><p id="p-d5d-w57-3kl">信息名稱</p></th><th id="td-25j-y88-qz4"><p id="p-76e-ai9-e8k">含義</p></th></tr></thead><tbody data-tag="tbody" id="tbody-xoa-aut-c7g" class="tbody"><tr data-tag="tr" id="tr-p36-91p-o8b" class="tr"><td data-tag="td" id="td-1p7-l0b-deh" class="td"><p id="p-hkk-obu-qii">output</p></td><td data-tag="td" id="td-ojy-dcx-t6s" class="td"><p id="p-3y1-6ay-iiu">該算子輸出的表達式。</p></td></tr><tr data-tag="tr" id="tr-e2r-ieg-a3o" class="tr"><td data-tag="td" id="td-kv9-foc-oqz" class="td"><p id="p-7ml-iwm-f96">filter</p></td><td data-tag="td" id="td-q1a-f3z-mul" class="td"><p id="p-icf-fbf-lox">該算子上的過濾條件。</p><p id="p-iyu-qq6-lhf">由于示例中 PX PARTITION ITERATOR 算子沒有設置 filter，所以為 nil。</p></td></tr></tbody></table> ## PX BLOCK ITERATOR PX BLOCK ITERATOR 算子用于按照數據塊粒度迭代數據。相對于 PX PARTITION ITERATOR，PX BLOCK ITERATOR 算子按照數據塊迭代的方式粒度更小，能夠切分出更多的任務，支持更高的并行度。 ~~~ obclient>EXPLAIN SELECT /*+PARALLEL(4)*/ c1 FROM t\G; *************************** 1. row *************************** Query Plan: ================================================== |ID|OPERATOR |NAME |EST. ROWS|COST | -------------------------------------------------- |0 |PX COORDINATOR | |400000 |279171| |1 | EXCHANGE OUT DISTR|:EX10000|400000 |189244| |2 | PX BLOCK ITERATOR| |400000 |189244| |3 | TABLE SCAN |T(IDX) |400000 |189244| ================================================== Outputs & filters: ------------------------------------- 0 - output([T.C1]), filter(nil) 1 - output([T.C1]), filter(nil), dop=4 2 - output([T.C1]), filter(nil) 3 - output([T.C1]), filter(nil), access([T.C1]), partitions(p[0-3]) ~~~ 上述示例的執行計劃展示中的 outputs & filters 詳細列出了 PX BLOCK ITERATOR 算子的輸出信息，字段的含義與 PX PARTITION ITERATOR 算子相同。