...

1. Improved query latency, as we will be able to skip much more partitions than now (only backup partitions are skipped for now)
2. Improved thin client latency - it will be possible to send requests to target node, thus saving one network hop.
3. Decreased page cache pressure - less data to read, less data to evict, less number of page locks
4. Improved system throughput, as less total CPU and IO operations will be required to execute optimized query
5. Improved thin client latency - it will be possible to send requests to target node, thus saving one network hop.

Partition pruning is already implemented in Apache Ignite in very simplified form [1]. Only WHERE condition with equality is considered and only for SQL queries without joins. We should expand it further.

[1] https://issues.apache.org/jira/browse/IGNITE-4509

...

In the following sections we first explain how partitions could be extracted from SQL parts, and how certain query rewrite techniques could help us with it. Then we will describe how extracted partition info is assembled in a form of tree. Then we discuss that partition extraction should be performed two times - before split for the whole query, and after split for query parts. Finally, we explain how partition info will be passed to thin clients, and how users will be able to control and monitor partition pruning.

...

(P1) OR (P2) => (P1, P2)
(P1) OR (ALL) => (ALL)
(P1) OR () => (P1)
(P1, P2) OR (P2, P3) => (P1, P2, P3)


(:1) OR (:2) => (:1, :2)
(P1, :1) OR (P2, :2) => (P1, P2, :1, :2)

Joins

Query Rewrite

Subqueries

TODO

OR

TODO

Theory

This technique is used excessively by both distributed databases and classical RDBMS vendors (as a part of their partitioning/sharding solutions). First, query is analyzed and optionally re-written to allow for better partition pruning. Second, information of interesting partitions are extracted from query for every table (cache) and subquery (for complex multi-stage distributed queries). 

Query Rewrite

WHERE conditions and JOIN-s are relatively simple targets for partition extraction. The main difficulty is subqueries, as they potentially introduce unlimited nesting that complicates analysis a lot. So the main goal of query rewrite is to minimize number of subqueries. Main techniques are JOIN conversion, correlation and materialization. Interestingly, all these techniques not only simplify partition extraction, but also allow optimizer to choose better execution plan for local execution, as JOINs are much easier to be analyzed and reordered than subqueries. As we shall see below, efforts are usually focused on non-correlated subqueries as they are the most costly for straightforward execution.

JOIN conversion

The goal is to convert a subquery to join if possible. Important prerequisite is that number of resulting rows is not changed. Outer(FK) → inner(PK) joins are possible. E.g. subqueries in SELECT (aka "derived table") and FROM clauses.

Example 1: JOIN conversion for derived table

SELECT emp.name, (SELECT dept.name FROM dept WHERE emp.dept_id=dept.id)
FROM emp
WHERE emp.salary>1000

SELECT emp.name, dept.name
FROM emp, dept
WHERE emp.salary>1000 AND emp.dept_id=dept.id

Example 2: JOIN conversion for FROM clause

SELECT emp.name, dept_subquery.name
FROM emp, (SELECT * FROM dept WHERE state='CA') dept_subquery
WHERE emp.salary>1000 AND emp.dept_id=dept_subquery.id

SELECT emp.name, dept.name
FROM emp, dept
WHERE emp.salary>1000 AND emp.dept_id=dept.id AND dept.state='CA'

Also, it is possible for some IN-clauses. MariaDB calls it "table pullout optimization" [1]

Joins are very common, so it is crucial to support partition extraction for them as well. General solution might be extremely complex, so we need to define reasonable bounds where optimization is applicable, and improve them iteratively in future.  We start with query AST obtained from parser. Proposed flow to extract partitions is explained below. Some of these steps could be merged to improve performance.

1. Look for non-equality JOIN conditions. When one is found, exit. This way join type space is reduced to equijoins.
2. Build co-location tree, which is another tree explaining how PARTITIONED tables are joined together
   1. Copy current JOIN AST into separate tree
   2. If table is REPLICATED and do not have node filter, then mark it as "ANY" and remove from the tree, as it doesn't affect JOIN outcome. Otherwise - exit, no need to bother with custom filters.
   3. If CROSS JOIN is found, then exit (might be improved in future)
   4. If tables are joined on their affinity columns and has equal affinity functions, then mark them as belonging to the same co-location group. Otherwise - assign them to different co-location groups. Repeat this for all tables and joins in the tree. Functions are equal if and only if the following is true:
      1. Affinity function is deterministic (e.g. RendezvousAffintiyFunction is deterministic, while FairAffinityFunction is not)
      2. Both affinity functions are equal
      3. There are no custom node filters
      4. There are no custom affinity key mappers
   5. Every subquery is assigned it's own co-location group unconditionally (may be improved in future)
   6. At this point we have a co-location tree with only PARTITIONED caches, only equi-joins, where every table is assigned to a single co-location group.
3. Extract partitions from expression tree with two additional rules:
   1. Every group of partitions is assigned respective co-location group from co-location tree

   2. REPLICATED caches with "ANY" policy should be eliminated as follows:

      Code Block
      language sql title ANY algebra linenumbers true
      (P1, :2) AND (ANY) => (P1, :2) (P1, :2) OR (ANY) => (P1, :2)

      
      

   3. If partition tree contain rules from different co-location groups, then exit.

4. At this point we have partition tree over a single co-location group. All outstanding arguments could be passed through the same affinity function to get target partitions.

Subquery rewrite

It is not easy to extract partitions from subqueries. But we can rewrite certain subqueries to joins with a technique called "join conversion". Important prerequisite is that number of resulting rows is not changed.

Example 1: JOIN conversion for derived tableExample 3: Table pullout optimization

SELECT emp.name
FROM emp
WHERE emp.salary>1000 AND, (SELECT dept.name FROM dept WHERE emp.dept_id IN (SELECT id FROM dept WHERE state='CA')=dept.id)
FROM emp
WHERE emp.salary>1000

SELECT emp.name, dept.name
FROM emp, dept
WHERE emp.salary>1000 AND emp.dept_id=dept.id AND dept.state='CA'

Another example where inner subquery return at most one row is outer(PK) → inner(FK GROUP BY PK) case. Subquery is not eliminated in this case, but IN expression is replaced with more comfortable JOIN.

Example 4: Replacing IN with JOIN

Example 2: JOIN conversion for FROM clause

SELECT emp.name, dept_subquery.name
FROM dept
WHEREemp, (SELECT * FROM dept.id IN (SELECT emp.dept_id FROM emp GROUP BYWHERE state='CA') dept_subquery
WHERE emp.salary>1000 AND emp.dept_id HAVING COUNT(*)>10)=dept_subquery.id

SELECT emp.name, dept.name
FROM emp, dept
WHERE JOIN (SELECTemp.salary>1000 AND emp.dept_id FROM emp GROUP BY emp.dept_id HAVING COUNT(*)>10) emp_subquery ON dept.id=emp_subquery.dept_id

Correlation

Uncorrelated subquery may need to scan huge result set. If underlying cache is PARTITIONED, then multiple nodes need to be scanned. If underlying cache is REPLICATED, then scan is local. If previous optimizations are not applicable, we can try to push-down condition from outer table to subquery. If condition is equality and both outer and inner columns are co-located, then potentially distributed query is reduced to co-located query. The downside is that subquery must be evaluated multiple times, once for every tuple from outer relation. This might be less than optimal comparing to subquery materialization explained below.  MySQL refer to this as IN-to-EXISTS optimization [2].

...

=dept.id AND dept.state='CA'

Also, it is possible for some IN-clauses. MariaDB calls it "table pullout optimization" [1]

Example 3: Table pullout optimization

SELECT emp.name
FROM emp
WHERE emp.salary>1000 AND emp.dept_id IN (SELECT id FROM dept WHERE state='CA')

SELECT emp.name
FROM emp, dept
WHERE emp.salary>1000 AND emp.dept_id IN (SELECT =dept.id FROMAND dept WHERE .state='CA')

SELECT emp.name
FROM emp
WHERE EXISTS (SELECT id FROM dept WHERE dept.id=emp.dept_id AND state='CA')

Materialization

If subquery is not correlated, then it can be executed only once. Result is cached in temporary table and then re-used. If expected number of returned records is small, then it makes sense to execute the request before other query parts, and send obtained result to other map nodes. If result is scalar, then entire query is replaced with a single result.

This way some queries which were too complex to extract target partitions from may be reduced to simpler form, where extraction is possible.

[1] https://mariadb.com/kb/en/library/table-pullout-optimization/
[2] https://dev.mysql.com/doc/refman/8.0/en/subquery-optimization-with-exists.html

Partition Extraction

TODO

Proposed Changes

[1] https://mariadb.com/kb/en/library/table-pullout-optimization/
[2] https://dev.mysql.com/doc/refman/8.0/en/subquery-optimization-with-exists.html

Thin Clients

Partitions could be calculated not only on the server side, but on clients as well. This way thin clients could be able to send requests directly to affected nodes, reducing latency. To achieve this we first define serializable partition tree model:

interface PartitionNode {
    Collection<Integer> apply(Object[] args);
}

class PartitionGroup implements PartitionNode {
    Collection<Object> parts; // Concrete partitions, arguments or both.
}

class PartitionExpression implements PartitionNode {
    PartitionNode left;
    PartitionNode right;
}

Partition tree is enriched with AffinityTopologyVersion it was built on, and affinity function descriptor. Descriptor can only be defined for well-known affinity functions, such as RendezvousAffinityFunction. 

class PartitionInfo {
    PartitionNode tree;
    AffintiyTopologyVersion affTopVer;
    AffinityFunctionDecriptor affFunc;
}

When client executes request for the first time, the following sequence of actions happen:

1. Query request is received by the server
2. Partition tree is built using specific affinity topology version.
3. Query is executed as usual
4. If partition tree exists and is built with well-known affinity function, then it is attached to query response.

Client saves partition tree and re-use it for further requests as follows:

1. Query arguments are applied on the client. 
2. Target node is determined from the list of partitions. We assume that partition distribution for the given affinity topology versions has been requested in advance similarly how we do that for C++ thin client. 
3. If only one node is resolved, send request to it. If several nodes are resolved - send request to random node from the list. 
4. Request is executed on the server and current affinity topology version is attached to the response. If it differs from the one received from the client, new partition tree is built and attached.
5. Client checks if current affinity topology version differs. If yes - old partition tree is invalidated.

Optimizations

If partition tree is extracted form the query successfully, then two types of optimizations are possible:

1. If tree evaluation returned empty partition set, return empty result set immediately without actual query execution
2. If tree evaluation returned one partition, then all data reside on a single node. Convert query to "local" and execute it on target node without two-phase flow
3. If tree evaluation returned several partitions, and all of them appear to be on the same node, then try to execute query speculatively on a single node, provided that partitions are still on that node. Fallback to normal execution mode in case of concurrent eviction.

Management and Monitoring

It is very important to let user know if partition pruning is applicable to query for performance tuning. For every cached two-step query we may expose the following information:

1. Whether partition pruning is applicable
2. Formatted partition tree
3. Affinity topology version of the plan
4. If not applicable - explain why (e.g. non-equijoin, incompatible affinity functions, etc.)

 Also we need to let user disable this optimization. Otherwise a bug in implementation may lead to incorrect results with no workarounds. System on configuration property could be used for that.TODO

Tickets