Github user zuyu commented on a diff in the pull request:
https://github.com/apache/incubator-quickstep/pull/179#discussion_r99908259
--- Diff: storage/AggregationOperationState.hpp ---
@@ -156,6 +152,29 @@ class AggregationOperationState {
const CatalogDatabaseLite &database);
/**
+ * @brief Get the number of partitions to be used for initializing the
+ * aggregation.
+ *
+ * @return The number of partitions to be used for initializing the
aggregation.
+ **/
+ std::size_t getNumInitializationPartitions() const;
+
+ /**
+ * @brief Get the number of partitions to be used for finalizing the
+ * aggregation.
+ *
+ * @return The number of partitions to be used for finalizing the
aggregation.
+ **/
+ std::size_t getNumFinalizationPartitions() const;
--- End diff --
I get your point, but for the distributed version, my concern is that we
have to schedule the finalize work order with a partition to the same node that
executes the initialize work order. Thus, your approach makes it harder to
implement in the distributed version.
On the other hand, we could measure the overheads of the initialize work
order, and I guess it is related small, compared to the whole aggregation
processing time.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
