The AQL query optimizer

AQL queries are sent through an optimizer before execution. The task of the optimizer is to create an initial execution plan for the query, look for optimization opportunities and apply them. As a result, the optimizer might produce multiple execution plans for a single query. It will then calculate the costs for all plans and pick the plan with the lowest total cost. This resulting plan is considered to be the optimal plan, which is then executed.

The optimizer is designed to only perform optimizations if they are safe, in the meaning that an optimization should not modify the result of a query. A notable exception to this is that the optimizer is allowed to change the order of results for queries that do not explicitly specify how results should be sorted.

Execution plans

The explain command can be used to query the optimal executed plan or even all plans the optimizer has generated. Additionally, explain can reveal some more information about the optimizer’s view of the query.

Inspecting plans using the explain helper

The explain method of ArangoStatement as shown in the next chapters creates very verbose output. You can work on the output programmatically, or use this handsome tool that we created to generate a more human readable representation.

You may use it like this: (we disable syntax highlighting here)

arangosh> db._create("test");
arangosh> for (i = 0; i < 100; ++i) { db.test.save({ value: i }); }
arangosh> db.test.ensureIndex({ type: "skiplist", fields: [ "value" ] });
arangosh> var explain = require("@arangodb/aql/explainer").explain;
arangosh> explain("FOR i IN test FILTER i.value > 97 SORT i.value RETURN i.value", {colors:false});
Show execution results
[ArangoCollection 64685, "test" (type document, status loaded)]
{ 
  "deduplicate" : true, 
  "fields" : [ 
    "value" 
  ], 
  "id" : "test/64890", 
  "isNewlyCreated" : true, 
  "name" : "idx_1649564012184076288", 
  "selectivityEstimate" : 1, 
  "sparse" : false, 
  "type" : "skiplist", 
  "unique" : false, 
  "code" : 201 
}
Query String (61 chars, cacheable: true):
 FOR i IN test FILTER i.value > 97 SORT i.value RETURN i.value

Execution plan:
 Id   NodeType          Est.   Comment
  1   SingletonNode        1   * ROOT
  9   IndexNode           50     - FOR i IN test   /* skiplist index scan, index only, projections: `value` */
  5   CalculationNode     50       - LET #3 = i.`value`   /* attribute expression */   /* collections used: i : test */
  8   ReturnNode          50       - RETURN #3

Indexes used:
 By   Name                      Type       Collection   Unique   Sparse   Selectivity   Fields        Ranges
  9   idx_1649564012184076288   skiplist   test         false    false       100.00 %   [ `value` ]   (i.`value` > 97)

Optimization rules applied:
 Id   RuleName
  1   move-calculations-up
  2   move-filters-up
  3   remove-redundant-calculations
  4   remove-unnecessary-calculations
  5   move-calculations-up-2
  6   move-filters-up-2
  7   use-indexes
  8   remove-filter-covered-by-index
  9   use-index-for-sort
 10   remove-unnecessary-calculations-2
 11   reduce-extraction-to-projection
Hide execution results

Execution plans in detail

Let’s have a look at the raw json output of the same execution plan using the explain method of ArangoStatement:

arangosh> stmt = db._createStatement("FOR i IN test FILTER i.value > 97 SORT i.value RETURN i.value");
arangosh> stmt.explain();
Show execution results
[object ArangoStatement]
{ 
  "plan" : { 
    "nodes" : [ 
      { 
        "type" : "SingletonNode", 
        "dependencies" : [ ], 
        "id" : 1, 
        "estimatedCost" : 1, 
        "estimatedNrItems" : 1 
      }, 
      { 
        "type" : "IndexNode", 
        "dependencies" : [ 
          1 
        ], 
        "id" : 9, 
        "estimatedCost" : 10.143856189774723, 
        "estimatedNrItems" : 50, 
        "outVariable" : { 
          "id" : 0, 
          "name" : "i" 
        }, 
        "projections" : [ 
          "value" 
        ], 
        "producesResult" : true, 
        "database" : "_system", 
        "collection" : "test", 
        "satellite" : false, 
        "isSatellite" : false, 
        "needsGatherNodeSort" : true, 
        "indexCoversProjections" : true, 
        "indexes" : [ 
          { 
            "id" : "64890", 
            "type" : "skiplist", 
            "name" : "idx_1649564012184076288", 
            "fields" : [ 
              "value" 
            ], 
            "selectivityEstimate" : 1, 
            "unique" : false, 
            "sparse" : false, 
            "deduplicate" : true 
          } 
        ], 
        "condition" : { 
          "type" : "n-ary or", 
          "typeID" : 63, 
          "subNodes" : [ 
            { 
              "type" : "n-ary and", 
              "typeID" : 62, 
              "subNodes" : [ 
                { 
                  "type" : "compare >", 
                  "typeID" : 29, 
                  "subNodes" : [ 
                    { 
                      "type" : "attribute access", 
                      "typeID" : 35, 
                      "name" : "value", 
                      "subNodes" : [ 
                        { 
                          "type" : "reference", 
                          "typeID" : 45, 
                          "name" : "i", 
                          "id" : 0 
                        } 
                      ] 
                    }, 
                    { 
                      "type" : "value", 
                      "typeID" : 40, 
                      "value" : 97, 
                      "vType" : "int", 
                      "vTypeID" : 2 
                    } 
                  ] 
                } 
              ] 
            } 
          ] 
        }, 
        "sorted" : true, 
        "ascending" : true, 
        "reverse" : false, 
        "evalFCalls" : true, 
        "limit" : 0 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          9 
        ], 
        "id" : 5, 
        "estimatedCost" : 60.14385618977472, 
        "estimatedNrItems" : 50, 
        "expression" : { 
          "type" : "attribute access", 
          "typeID" : 35, 
          "name" : "value", 
          "subNodes" : [ 
            { 
              "type" : "reference", 
              "typeID" : 45, 
              "name" : "i", 
              "id" : 0 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 4, 
          "name" : "3" 
        }, 
        "canThrow" : false, 
        "expressionType" : "attribute" 
      }, 
      { 
        "type" : "ReturnNode", 
        "dependencies" : [ 
          5 
        ], 
        "id" : 8, 
        "estimatedCost" : 110.14385618977472, 
        "estimatedNrItems" : 50, 
        "inVariable" : { 
          "id" : 4, 
          "name" : "3" 
        }, 
        "count" : true 
      } 
    ], 
    "rules" : [ 
      "move-calculations-up", 
      "move-filters-up", 
      "remove-redundant-calculations", 
      "remove-unnecessary-calculations", 
      "move-calculations-up-2", 
      "move-filters-up-2", 
      "use-indexes", 
      "remove-filter-covered-by-index", 
      "use-index-for-sort", 
      "remove-unnecessary-calculations-2", 
      "reduce-extraction-to-projection" 
    ], 
    "collections" : [ 
      { 
        "name" : "test", 
        "type" : "read" 
      } 
    ], 
    "variables" : [ 
      { 
        "id" : 6, 
        "name" : "5" 
      }, 
      { 
        "id" : 4, 
        "name" : "3" 
      }, 
      { 
        "id" : 2, 
        "name" : "1" 
      }, 
      { 
        "id" : 0, 
        "name" : "i" 
      } 
    ], 
    "estimatedCost" : 110.14385618977472, 
    "estimatedNrItems" : 50, 
    "isModificationQuery" : false 
  }, 
  "warnings" : [ ], 
  "stats" : { 
    "rulesExecuted" : 39, 
    "rulesSkipped" : 0, 
    "plansCreated" : 1 
  }, 
  "cacheable" : true 
}
Hide execution results

As you can see, the result details are very verbose so we will not show them in full in the next sections. Instead, let’s take a closer look at the results step by step.

Execution nodes

In general, an execution plan can be considered to be a pipeline of processing steps. Each processing step is carried out by a so-called execution node

The nodes attribute of the explain result contains these execution nodes in the execution plan. The output is still very verbose, so here’s a shorted form of it:

arangosh> stmt.explain().plan.nodes.map(function (node) { return node.type; });
Show execution results
[ 
  "SingletonNode", 
  "IndexNode", 
  "CalculationNode", 
  "ReturnNode" 
]
Hide execution results

Note that the list of nodes might slightly change in future versions of ArangoDB if new execution node types get added or the optimizer create somewhat more optimized plans).

When a plan is executed, the query execution engine will start with the node at the bottom of the list (i.e. the ReturnNode).

The ReturnNode’s purpose is to return data to the caller. It does not produce data itself, so it will ask the node above itself, this is the CalculationNode in our example. CalculationNodes are responsible for evaluating arbitrary expressions. In our example query, the CalculationNode will evaluate the value of i.value, which is needed by the ReturnNode. The calculation will be applied for all data the CalculationNode gets from the node above it, in our example the IndexNode.

Finally, all of this needs to be done for documents of collection test. This is where the IndexNode enters the game. It will use an index (thus its name) to find certain documents in the collection and ship it down the pipeline in the order required by SORT i.value. The IndexNode itself has a SingletonNode as its input. The sole purpose of a SingletonNode node is to provide a single empty document as input for other processing steps. It is always the end of the pipeline.

Here is a summary:

  • SingletonNode: produces an empty document as input for other processing steps.
  • IndexNode: iterates over the index on attribute value in collection test in the order required by SORT i.value.
  • CalculationNode: evaluates the result of the calculation i.value > 97 to true or false
  • CalculationNode: calculates return value i.value
  • ReturnNode: returns data to the caller

Optimizer rules

Note that in the example, the optimizer has optimized the SORT statement away. It can do it safely because there is a sorted skiplist index on i.value, which it has picked in the IndexNode. As the index values are iterated over in sorted order anyway, the extra SortNode would have been redundant and was removed.

Additionally, the optimizer has done more work to generate an execution plan that avoids as much expensive operations as possible. Here is the list of optimizer rules that were applied to the plan:

arangosh> stmt.explain().plan.rules;
Show execution results
[ 
  "move-calculations-up", 
  "move-filters-up", 
  "remove-redundant-calculations", 
  "remove-unnecessary-calculations", 
  "move-calculations-up-2", 
  "move-filters-up-2", 
  "use-indexes", 
  "remove-filter-covered-by-index", 
  "use-index-for-sort", 
  "remove-unnecessary-calculations-2", 
  "reduce-extraction-to-projection" 
]
Hide execution results

Here is the meaning of these rules in context of this query:

  • move-calculations-up: moves a CalculationNode as far up in the processing pipeline as possible
  • move-filters-up: moves a FilterNode as far up in the processing pipeline as possible
  • remove-redundant-calculations: replaces references to variables with references to other variables that contain the exact same result. In the example query, i.value is calculated multiple times, but each calculation inside a loop iteration would produce the same value. Therefore, the expression result is shared by several nodes.
  • remove-unnecessary-calculations: removes CalculationNodes whose result values are not used in the query. In the example this happens due to the remove-redundant-calculations rule having made some calculations unnecessary.
  • use-indexes: use an index to iterate over a collection instead of performing a full collection scan. In the example case this makes sense, as the index can be used for filtering and sorting.
  • remove-filter-covered-by-index: remove an unnecessary filter whose functionality is already covered by an index. In this case the index only returns documents matching the filter.
  • use-index-for-sort: removes a SORT operation if it is already satisfied by traversing over a sorted index

Note that some rules may appear multiple times in the list, with number suffixes. This is due to the same rule being applied multiple times, at different positions in the optimizer pipeline.

Also see the full List of optimizer rules below.

Collections used in a query

The list of collections used in a plan (and query) is contained in the collections attribute of a plan:

arangosh> stmt.explain().plan.collections
Show execution results
[ 
  { 
    "name" : "test", 
    "type" : "read" 
  } 
]
Hide execution results

The name attribute contains the name of the collection, and type is the access type, which can be either read or write.

Variables used in a query

The optimizer will also return a list of variables used in a plan (and query). This list will contain auxiliary variables created by the optimizer itself. This list can be ignored by end users in most cases.

Cost of a query

For each plan the optimizer generates, it will calculate the total cost. The plan with the lowest total cost is considered to be the optimal plan. Costs are estimates only, as the actual execution costs are unknown to the optimizer. Costs are calculated based on heuristics that are hard-coded into execution nodes. Cost values do not have any unit.

Retrieving all execution plans

To retrieve not just the optimal plan but a list of all plans the optimizer has generated, set the option allPlans to true:

This will return a list of all plans in the plans attribute instead of in the plan attribute:

arangosh> stmt.explain({ allPlans: true });
Show execution results
{ 
  "plans" : [ 
    { 
      "nodes" : [ 
        { 
          "type" : "SingletonNode", 
          "dependencies" : [ ], 
          "id" : 1, 
          "estimatedCost" : 1, 
          "estimatedNrItems" : 1 
        }, 
        { 
          "type" : "IndexNode", 
          "dependencies" : [ 
            1 
          ], 
          "id" : 9, 
          "estimatedCost" : 10.143856189774723, 
          "estimatedNrItems" : 50, 
          "outVariable" : { 
            "id" : 0, 
            "name" : "i" 
          }, 
          "projections" : [ 
            "value" 
          ], 
          "producesResult" : true, 
          "database" : "_system", 
          "collection" : "test", 
          "satellite" : false, 
          "isSatellite" : false, 
          "needsGatherNodeSort" : true, 
          "indexCoversProjections" : true, 
          "indexes" : [ 
            { 
              "id" : "64890", 
              "type" : "skiplist", 
              "name" : "idx_1649564012184076288", 
              "fields" : [ 
                "value" 
              ], 
              "selectivityEstimate" : 1, 
              "unique" : false, 
              "sparse" : false, 
              "deduplicate" : true 
            } 
          ], 
          "condition" : { 
            "type" : "n-ary or", 
            "typeID" : 63, 
            "subNodes" : [ 
              { 
                "type" : "n-ary and", 
                "typeID" : 62, 
                "subNodes" : [ 
                  { 
                    "type" : "compare >", 
                    "typeID" : 29, 
                    "subNodes" : [ 
                      { 
                        "type" : "attribute access", 
                        "typeID" : 35, 
                        "name" : "value", 
                        "subNodes" : [ 
                          { 
                            "type" : "reference", 
                            "typeID" : 45, 
                            "name" : "i", 
                            "id" : 0 
                          } 
                        ] 
                      }, 
                      { 
                        "type" : "value", 
                        "typeID" : 40, 
                        "value" : 97, 
                        "vType" : "int", 
                        "vTypeID" : 2 
                      } 
                    ] 
                  } 
                ] 
              } 
            ] 
          }, 
          "sorted" : true, 
          "ascending" : true, 
          "reverse" : false, 
          "evalFCalls" : true, 
          "limit" : 0 
        }, 
        { 
          "type" : "CalculationNode", 
          "dependencies" : [ 
            9 
          ], 
          "id" : 5, 
          "estimatedCost" : 60.14385618977472, 
          "estimatedNrItems" : 50, 
          "expression" : { 
            "type" : "attribute access", 
            "typeID" : 35, 
            "name" : "value", 
            "subNodes" : [ 
              { 
                "type" : "reference", 
                "typeID" : 45, 
                "name" : "i", 
                "id" : 0 
              } 
            ] 
          }, 
          "outVariable" : { 
            "id" : 4, 
            "name" : "3" 
          }, 
          "canThrow" : false, 
          "expressionType" : "attribute" 
        }, 
        { 
          "type" : "ReturnNode", 
          "dependencies" : [ 
            5 
          ], 
          "id" : 8, 
          "estimatedCost" : 110.14385618977472, 
          "estimatedNrItems" : 50, 
          "inVariable" : { 
            "id" : 4, 
            "name" : "3" 
          }, 
          "count" : true 
        } 
      ], 
      "rules" : [ 
        "move-calculations-up", 
        "move-filters-up", 
        "remove-redundant-calculations", 
        "remove-unnecessary-calculations", 
        "move-calculations-up-2", 
        "move-filters-up-2", 
        "use-indexes", 
        "remove-filter-covered-by-index", 
        "use-index-for-sort", 
        "remove-unnecessary-calculations-2", 
        "reduce-extraction-to-projection" 
      ], 
      "collections" : [ 
        { 
          "name" : "test", 
          "type" : "read" 
        } 
      ], 
      "variables" : [ 
        { 
          "id" : 6, 
          "name" : "5" 
        }, 
        { 
          "id" : 4, 
          "name" : "3" 
        }, 
        { 
          "id" : 2, 
          "name" : "1" 
        }, 
        { 
          "id" : 0, 
          "name" : "i" 
        } 
      ], 
      "estimatedCost" : 110.14385618977472, 
      "estimatedNrItems" : 50, 
      "isModificationQuery" : false 
    } 
  ], 
  "warnings" : [ ], 
  "stats" : { 
    "rulesExecuted" : 39, 
    "rulesSkipped" : 0, 
    "plansCreated" : 1 
  } 
}
Hide execution results

Retrieving the plan as it was generated by the parser / lexer

To retrieve the plan which closely matches your query, you may turn off most optimization rules (i.e. cluster rules cannot be disabled if you’re running the explain on a cluster coordinator) set the option rules to -all:

This will return an unoptimized plan in the plan:

arangosh> stmt.explain({ optimizer: { rules: [ "-all" ] } });
Show execution results
{ 
  "plan" : { 
    "nodes" : [ 
      { 
        "type" : "SingletonNode", 
        "dependencies" : [ ], 
        "id" : 1, 
        "estimatedCost" : 1, 
        "estimatedNrItems" : 1 
      }, 
      { 
        "type" : "EnumerateCollectionNode", 
        "dependencies" : [ 
          1 
        ], 
        "id" : 2, 
        "estimatedCost" : 102, 
        "estimatedNrItems" : 100, 
        "random" : false, 
        "indexHint" : { 
          "forced" : false, 
          "type" : "none" 
        }, 
        "outVariable" : { 
          "id" : 0, 
          "name" : "i" 
        }, 
        "projections" : [ ], 
        "producesResult" : true, 
        "database" : "_system", 
        "collection" : "test", 
        "satellite" : false, 
        "isSatellite" : false 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          2 
        ], 
        "id" : 3, 
        "estimatedCost" : 202, 
        "estimatedNrItems" : 100, 
        "expression" : { 
          "type" : "compare >", 
          "typeID" : 29, 
          "subNodes" : [ 
            { 
              "type" : "attribute access", 
              "typeID" : 35, 
              "name" : "value", 
              "subNodes" : [ 
                { 
                  "type" : "reference", 
                  "typeID" : 45, 
                  "name" : "i", 
                  "id" : 0 
                } 
              ] 
            }, 
            { 
              "type" : "value", 
              "typeID" : 40, 
              "value" : 97, 
              "vType" : "int", 
              "vTypeID" : 2 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 2, 
          "name" : "1" 
        }, 
        "canThrow" : false, 
        "expressionType" : "simple" 
      }, 
      { 
        "type" : "FilterNode", 
        "dependencies" : [ 
          3 
        ], 
        "id" : 4, 
        "estimatedCost" : 302, 
        "estimatedNrItems" : 100, 
        "inVariable" : { 
          "id" : 2, 
          "name" : "1" 
        } 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          4 
        ], 
        "id" : 5, 
        "estimatedCost" : 402, 
        "estimatedNrItems" : 100, 
        "expression" : { 
          "type" : "attribute access", 
          "typeID" : 35, 
          "name" : "value", 
          "subNodes" : [ 
            { 
              "type" : "reference", 
              "typeID" : 45, 
              "name" : "i", 
              "id" : 0 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 4, 
          "name" : "3" 
        }, 
        "canThrow" : false, 
        "expressionType" : "attribute" 
      }, 
      { 
        "type" : "SortNode", 
        "dependencies" : [ 
          5 
        ], 
        "id" : 6, 
        "estimatedCost" : 1066.3856189774724, 
        "estimatedNrItems" : 100, 
        "elements" : [ 
          { 
            "inVariable" : { 
              "id" : 4, 
              "name" : "3" 
            }, 
            "ascending" : true 
          } 
        ], 
        "stable" : false, 
        "limit" : 0, 
        "strategy" : "standard" 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          6 
        ], 
        "id" : 7, 
        "estimatedCost" : 1166.3856189774724, 
        "estimatedNrItems" : 100, 
        "expression" : { 
          "type" : "attribute access", 
          "typeID" : 35, 
          "name" : "value", 
          "subNodes" : [ 
            { 
              "type" : "reference", 
              "typeID" : 45, 
              "name" : "i", 
              "id" : 0 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 6, 
          "name" : "5" 
        }, 
        "canThrow" : false, 
        "expressionType" : "attribute" 
      }, 
      { 
        "type" : "ReturnNode", 
        "dependencies" : [ 
          7 
        ], 
        "id" : 8, 
        "estimatedCost" : 1266.3856189774724, 
        "estimatedNrItems" : 100, 
        "inVariable" : { 
          "id" : 6, 
          "name" : "5" 
        }, 
        "count" : true 
      } 
    ], 
    "rules" : [ ], 
    "collections" : [ 
      { 
        "name" : "test", 
        "type" : "read" 
      } 
    ], 
    "variables" : [ 
      { 
        "id" : 6, 
        "name" : "5" 
      }, 
      { 
        "id" : 4, 
        "name" : "3" 
      }, 
      { 
        "id" : 2, 
        "name" : "1" 
      }, 
      { 
        "id" : 0, 
        "name" : "i" 
      } 
    ], 
    "estimatedCost" : 1266.3856189774724, 
    "estimatedNrItems" : 100, 
    "isModificationQuery" : false 
  }, 
  "warnings" : [ ], 
  "stats" : { 
    "rulesExecuted" : 2, 
    "rulesSkipped" : 37, 
    "plansCreated" : 1 
  }, 
  "cacheable" : true 
}
Hide execution results

Note that some optimizations are already done at parse time (i.e. evaluate simple constant calculation as 1 + 1)

Turning specific optimizer rules off

Optimizer rules can also be turned on or off individually, using the rules attribute. This can be used to enable or disable one or multiple rules. Rules that shall be enabled need to be prefixed with a +, rules to be disabled should be prefixed with a -. The pseudo-rule all matches all rules.

Rules specified in rules are evaluated from left to right, so the following works to turn on just the one specific rule:

arangosh> stmt.explain({ optimizer: { rules: [ "-all", "+use-index-range" ] } });
Show execution results
{ 
  "plan" : { 
    "nodes" : [ 
      { 
        "type" : "SingletonNode", 
        "dependencies" : [ ], 
        "id" : 1, 
        "estimatedCost" : 1, 
        "estimatedNrItems" : 1 
      }, 
      { 
        "type" : "EnumerateCollectionNode", 
        "dependencies" : [ 
          1 
        ], 
        "id" : 2, 
        "estimatedCost" : 102, 
        "estimatedNrItems" : 100, 
        "random" : false, 
        "indexHint" : { 
          "forced" : false, 
          "type" : "none" 
        }, 
        "outVariable" : { 
          "id" : 0, 
          "name" : "i" 
        }, 
        "projections" : [ ], 
        "producesResult" : true, 
        "database" : "_system", 
        "collection" : "test", 
        "satellite" : false, 
        "isSatellite" : false 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          2 
        ], 
        "id" : 3, 
        "estimatedCost" : 202, 
        "estimatedNrItems" : 100, 
        "expression" : { 
          "type" : "compare >", 
          "typeID" : 29, 
          "subNodes" : [ 
            { 
              "type" : "attribute access", 
              "typeID" : 35, 
              "name" : "value", 
              "subNodes" : [ 
                { 
                  "type" : "reference", 
                  "typeID" : 45, 
                  "name" : "i", 
                  "id" : 0 
                } 
              ] 
            }, 
            { 
              "type" : "value", 
              "typeID" : 40, 
              "value" : 97, 
              "vType" : "int", 
              "vTypeID" : 2 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 2, 
          "name" : "1" 
        }, 
        "canThrow" : false, 
        "expressionType" : "simple" 
      }, 
      { 
        "type" : "FilterNode", 
        "dependencies" : [ 
          3 
        ], 
        "id" : 4, 
        "estimatedCost" : 302, 
        "estimatedNrItems" : 100, 
        "inVariable" : { 
          "id" : 2, 
          "name" : "1" 
        } 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          4 
        ], 
        "id" : 5, 
        "estimatedCost" : 402, 
        "estimatedNrItems" : 100, 
        "expression" : { 
          "type" : "attribute access", 
          "typeID" : 35, 
          "name" : "value", 
          "subNodes" : [ 
            { 
              "type" : "reference", 
              "typeID" : 45, 
              "name" : "i", 
              "id" : 0 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 4, 
          "name" : "3" 
        }, 
        "canThrow" : false, 
        "expressionType" : "attribute" 
      }, 
      { 
        "type" : "SortNode", 
        "dependencies" : [ 
          5 
        ], 
        "id" : 6, 
        "estimatedCost" : 1066.3856189774724, 
        "estimatedNrItems" : 100, 
        "elements" : [ 
          { 
            "inVariable" : { 
              "id" : 4, 
              "name" : "3" 
            }, 
            "ascending" : true 
          } 
        ], 
        "stable" : false, 
        "limit" : 0, 
        "strategy" : "standard" 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          6 
        ], 
        "id" : 7, 
        "estimatedCost" : 1166.3856189774724, 
        "estimatedNrItems" : 100, 
        "expression" : { 
          "type" : "attribute access", 
          "typeID" : 35, 
          "name" : "value", 
          "subNodes" : [ 
            { 
              "type" : "reference", 
              "typeID" : 45, 
              "name" : "i", 
              "id" : 0 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 6, 
          "name" : "5" 
        }, 
        "canThrow" : false, 
        "expressionType" : "attribute" 
      }, 
      { 
        "type" : "ReturnNode", 
        "dependencies" : [ 
          7 
        ], 
        "id" : 8, 
        "estimatedCost" : 1266.3856189774724, 
        "estimatedNrItems" : 100, 
        "inVariable" : { 
          "id" : 6, 
          "name" : "5" 
        }, 
        "count" : true 
      } 
    ], 
    "rules" : [ ], 
    "collections" : [ 
      { 
        "name" : "test", 
        "type" : "read" 
      } 
    ], 
    "variables" : [ 
      { 
        "id" : 6, 
        "name" : "5" 
      }, 
      { 
        "id" : 4, 
        "name" : "3" 
      }, 
      { 
        "id" : 2, 
        "name" : "1" 
      }, 
      { 
        "id" : 0, 
        "name" : "i" 
      } 
    ], 
    "estimatedCost" : 1266.3856189774724, 
    "estimatedNrItems" : 100, 
    "isModificationQuery" : false 
  }, 
  "warnings" : [ ], 
  "stats" : { 
    "rulesExecuted" : 2, 
    "rulesSkipped" : 37, 
    "plansCreated" : 1 
  }, 
  "cacheable" : true 
}
Hide execution results

By default, all rules are turned on. To turn off just a few specific rules, use something like this:

arangosh> stmt.explain({ optimizer: { rules: [ "-use-index-range", "-use-index-for-sort" ] } });
Show execution results
{ 
  "plan" : { 
    "nodes" : [ 
      { 
        "type" : "SingletonNode", 
        "dependencies" : [ ], 
        "id" : 1, 
        "estimatedCost" : 1, 
        "estimatedNrItems" : 1 
      }, 
      { 
        "type" : "IndexNode", 
        "dependencies" : [ 
          1 
        ], 
        "id" : 9, 
        "estimatedCost" : 10.143856189774723, 
        "estimatedNrItems" : 50, 
        "outVariable" : { 
          "id" : 0, 
          "name" : "i" 
        }, 
        "projections" : [ 
          "value" 
        ], 
        "producesResult" : true, 
        "database" : "_system", 
        "collection" : "test", 
        "satellite" : false, 
        "isSatellite" : false, 
        "needsGatherNodeSort" : false, 
        "indexCoversProjections" : true, 
        "indexes" : [ 
          { 
            "id" : "64890", 
            "type" : "skiplist", 
            "name" : "idx_1649564012184076288", 
            "fields" : [ 
              "value" 
            ], 
            "selectivityEstimate" : 1, 
            "unique" : false, 
            "sparse" : false, 
            "deduplicate" : true 
          } 
        ], 
        "condition" : { 
          "type" : "n-ary or", 
          "typeID" : 63, 
          "subNodes" : [ 
            { 
              "type" : "n-ary and", 
              "typeID" : 62, 
              "subNodes" : [ 
                { 
                  "type" : "compare >", 
                  "typeID" : 29, 
                  "subNodes" : [ 
                    { 
                      "type" : "attribute access", 
                      "typeID" : 35, 
                      "name" : "value", 
                      "subNodes" : [ 
                        { 
                          "type" : "reference", 
                          "typeID" : 45, 
                          "name" : "i", 
                          "id" : 0 
                        } 
                      ] 
                    }, 
                    { 
                      "type" : "value", 
                      "typeID" : 40, 
                      "value" : 97, 
                      "vType" : "int", 
                      "vTypeID" : 2 
                    } 
                  ] 
                } 
              ] 
            } 
          ] 
        }, 
        "sorted" : true, 
        "ascending" : true, 
        "reverse" : false, 
        "evalFCalls" : true, 
        "limit" : 0 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          9 
        ], 
        "id" : 5, 
        "estimatedCost" : 60.14385618977472, 
        "estimatedNrItems" : 50, 
        "expression" : { 
          "type" : "attribute access", 
          "typeID" : 35, 
          "name" : "value", 
          "subNodes" : [ 
            { 
              "type" : "reference", 
              "typeID" : 45, 
              "name" : "i", 
              "id" : 0 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 4, 
          "name" : "3" 
        }, 
        "canThrow" : false, 
        "expressionType" : "attribute" 
      }, 
      { 
        "type" : "SortNode", 
        "dependencies" : [ 
          5 
        ], 
        "id" : 6, 
        "estimatedCost" : 342.33666567851094, 
        "estimatedNrItems" : 50, 
        "elements" : [ 
          { 
            "inVariable" : { 
              "id" : 4, 
              "name" : "3" 
            }, 
            "ascending" : true 
          } 
        ], 
        "stable" : false, 
        "limit" : 0, 
        "strategy" : "standard" 
      }, 
      { 
        "type" : "ReturnNode", 
        "dependencies" : [ 
          6 
        ], 
        "id" : 8, 
        "estimatedCost" : 392.33666567851094, 
        "estimatedNrItems" : 50, 
        "inVariable" : { 
          "id" : 4, 
          "name" : "3" 
        }, 
        "count" : true 
      } 
    ], 
    "rules" : [ 
      "move-calculations-up", 
      "move-filters-up", 
      "remove-redundant-calculations", 
      "remove-unnecessary-calculations", 
      "move-calculations-up-2", 
      "move-filters-up-2", 
      "use-indexes", 
      "remove-filter-covered-by-index", 
      "remove-unnecessary-calculations-2", 
      "reduce-extraction-to-projection" 
    ], 
    "collections" : [ 
      { 
        "name" : "test", 
        "type" : "read" 
      } 
    ], 
    "variables" : [ 
      { 
        "id" : 6, 
        "name" : "5" 
      }, 
      { 
        "id" : 4, 
        "name" : "3" 
      }, 
      { 
        "id" : 2, 
        "name" : "1" 
      }, 
      { 
        "id" : 0, 
        "name" : "i" 
      } 
    ], 
    "estimatedCost" : 392.33666567851094, 
    "estimatedNrItems" : 50, 
    "isModificationQuery" : false 
  }, 
  "warnings" : [ ], 
  "stats" : { 
    "rulesExecuted" : 38, 
    "rulesSkipped" : 1, 
    "plansCreated" : 1 
  }, 
  "cacheable" : true 
}
Hide execution results

The maximum number of plans created by the optimizer can also be limited using the maxNumberOfPlans attribute:

arangosh> stmt.explain({ maxNumberOfPlans: 1 });
Show execution results
{ 
  "plan" : { 
    "nodes" : [ 
      { 
        "type" : "SingletonNode", 
        "dependencies" : [ ], 
        "id" : 1, 
        "estimatedCost" : 1, 
        "estimatedNrItems" : 1 
      }, 
      { 
        "type" : "IndexNode", 
        "dependencies" : [ 
          1 
        ], 
        "id" : 9, 
        "estimatedCost" : 10.143856189774723, 
        "estimatedNrItems" : 50, 
        "outVariable" : { 
          "id" : 0, 
          "name" : "i" 
        }, 
        "projections" : [ 
          "value" 
        ], 
        "producesResult" : true, 
        "database" : "_system", 
        "collection" : "test", 
        "satellite" : false, 
        "isSatellite" : false, 
        "needsGatherNodeSort" : true, 
        "indexCoversProjections" : true, 
        "indexes" : [ 
          { 
            "id" : "64890", 
            "type" : "skiplist", 
            "name" : "idx_1649564012184076288", 
            "fields" : [ 
              "value" 
            ], 
            "selectivityEstimate" : 1, 
            "unique" : false, 
            "sparse" : false, 
            "deduplicate" : true 
          } 
        ], 
        "condition" : { 
          "type" : "n-ary or", 
          "typeID" : 63, 
          "subNodes" : [ 
            { 
              "type" : "n-ary and", 
              "typeID" : 62, 
              "subNodes" : [ 
                { 
                  "type" : "compare >", 
                  "typeID" : 29, 
                  "subNodes" : [ 
                    { 
                      "type" : "attribute access", 
                      "typeID" : 35, 
                      "name" : "value", 
                      "subNodes" : [ 
                        { 
                          "type" : "reference", 
                          "typeID" : 45, 
                          "name" : "i", 
                          "id" : 0 
                        } 
                      ] 
                    }, 
                    { 
                      "type" : "value", 
                      "typeID" : 40, 
                      "value" : 97, 
                      "vType" : "int", 
                      "vTypeID" : 2 
                    } 
                  ] 
                } 
              ] 
            } 
          ] 
        }, 
        "sorted" : true, 
        "ascending" : true, 
        "reverse" : false, 
        "evalFCalls" : true, 
        "limit" : 0 
      }, 
      { 
        "type" : "CalculationNode", 
        "dependencies" : [ 
          9 
        ], 
        "id" : 5, 
        "estimatedCost" : 60.14385618977472, 
        "estimatedNrItems" : 50, 
        "expression" : { 
          "type" : "attribute access", 
          "typeID" : 35, 
          "name" : "value", 
          "subNodes" : [ 
            { 
              "type" : "reference", 
              "typeID" : 45, 
              "name" : "i", 
              "id" : 0 
            } 
          ] 
        }, 
        "outVariable" : { 
          "id" : 4, 
          "name" : "3" 
        }, 
        "canThrow" : false, 
        "expressionType" : "attribute" 
      }, 
      { 
        "type" : "ReturnNode", 
        "dependencies" : [ 
          5 
        ], 
        "id" : 8, 
        "estimatedCost" : 110.14385618977472, 
        "estimatedNrItems" : 50, 
        "inVariable" : { 
          "id" : 4, 
          "name" : "3" 
        }, 
        "count" : true 
      } 
    ], 
    "rules" : [ 
      "move-calculations-up", 
      "move-filters-up", 
      "remove-redundant-calculations", 
      "remove-unnecessary-calculations", 
      "move-calculations-up-2", 
      "move-filters-up-2", 
      "use-indexes", 
      "remove-filter-covered-by-index", 
      "use-index-for-sort", 
      "remove-unnecessary-calculations-2", 
      "reduce-extraction-to-projection" 
    ], 
    "collections" : [ 
      { 
        "name" : "test", 
        "type" : "read" 
      } 
    ], 
    "variables" : [ 
      { 
        "id" : 6, 
        "name" : "5" 
      }, 
      { 
        "id" : 4, 
        "name" : "3" 
      }, 
      { 
        "id" : 2, 
        "name" : "1" 
      }, 
      { 
        "id" : 0, 
        "name" : "i" 
      } 
    ], 
    "estimatedCost" : 110.14385618977472, 
    "estimatedNrItems" : 50, 
    "isModificationQuery" : false 
  }, 
  "warnings" : [ ], 
  "stats" : { 
    "rulesExecuted" : 38, 
    "rulesSkipped" : 1, 
    "plansCreated" : 1 
  }, 
  "cacheable" : true 
}
Hide execution results

Optimizer statistics

The optimizer will return statistics as a part of an explain result.

The following attributes will be returned in the stats attribute of an explain result:

  • plansCreated: total number of plans created by the optimizer
  • rulesExecuted: number of rules executed (note: an executed rule does not indicate a plan was actually modified by a rule)
  • rulesSkipped: number of rules skipped by the optimizer

Warnings

For some queries, the optimizer may produce warnings. These will be returned in the warnings attribute of the explain result:

arangosh> var stmt = db._createStatement("FOR i IN 1..10 RETURN 1 / 0")
arangosh> stmt.explain().warnings;
Show execution results
[ 
  { 
    "code" : 1562, 
    "message" : "division by zero" 
  } 
]
Hide execution results

There is an upper bound on the number of warnings a query may produce. If that bound is reached, no further warnings will be returned.

Optimization in a cluster

When you are running AQL in the cluster, the parsing of the query is done on the Coordinator. The Coordinator then chops the query into snippets, which are either to remain on the Coordinator or need to be distributed to the shards on the DBServers over the network. The cutting sites are interconnected via Scatter-, Gather- and RemoteNodes. These nodes mark the network borders of the snippets.

The optimizer strives to reduce the amount of data transferred via these network interfaces by pushing FILTERs out to the shards, as it is vital to the query performance to reduce that data amount to transfer over the network links.

Some hops between Coordinators and DBServers are unavoidable. An example are user-defined functions (UDFs), which have to be executed on the Coordinator. If you cannot modify your query to have a lower amount of back and forth between sites, then try to lower the amount of data that has to be transferred between them. In case of UDFs, use effective FILTERs before calling them.

Using a cluster, there is a Site column if you explain a query. Snippets marked with DBS are executed on DBServers, COOR ones are executed on the respective Coordinator.

Query String (57 chars, cacheable: false):
 FOR doc IN test UPDATE doc WITH { updated: true } IN test

Execution plan:
 Id   NodeType          Site     Est.   Comment
  1   SingletonNode     DBS         1   * ROOT
  3   CalculationNode   DBS         1     - LET #3 = { "updated" : true }   
 13   IndexNode         DBS   1000000     - FOR doc IN test   /* primary index scan, index only, projections: `_key`, 5 shard(s) */    
  4   UpdateNode        DBS         0       - UPDATE doc WITH #3 IN test 
  7   RemoteNode        COOR        0       - REMOTE
  8   GatherNode        COOR        0       - GATHER 

List of execution nodes

The following execution node types will appear in the output of explain:

  • CalculationNode: evaluates an expression. The expression result may be used by other nodes, e.g. FilterNode, EnumerateListNode, SortNode etc.

  • CollectNode: aggregates its input and produces new output variables. This will appear once per COLLECT statement.

  • EnumerateCollectionNode: enumeration over documents of a collection (given in its collection attribute) without using an index.

  • EnumerateListNode: enumeration over a list of (non-collection) values.

  • EnumerateViewNode: enumeration over documents of a View.

  • FilterNode: only lets values pass that satisfy a filter condition. Will appear once per FILTER statement.

  • IndexNode: enumeration over one or many indexes (given in its indexes attribute) of a collection. The index ranges are specified in the condition attribute of the node.

  • InsertNode: inserts documents into a collection (given in its collection attribute). Will appear exactly once in a query that contains an INSERT statement.

  • KShortestPathsNode: indicates a traversal for k Shortest Paths (K_SHORTEST_PATHS in AQL).

  • LimitNode: limits the number of results passed to other processing steps. Will appear once per LIMIT statement.

  • MaterializeNode: the presence of this node means that the query is not fully covered by indexes and therefore needs to involve the storage engine.

  • RemoveNode: removes documents from a collection (given in its collection attribute). Will appear exactly once in a query that contains a REMOVE statement.

  • ReplaceNode: replaces documents in a collection (given in its collection attribute). Will appear exactly once in a query that contains a REPLACE statement.

  • ReturnNode: returns data to the caller. Will appear in each read-only query at least once. Subqueries will also contain ReturnNodes.

  • SingletonNode: the purpose of a SingletonNode is to produce an empty document that is used as input for other processing steps. Each execution plan will contain exactly one SingletonNode as its top node.

  • ShortestPathNode: indicates a traversal for a Shortest Path (SHORTEST_PATH in AQL).

  • SortNode: performs a sort of its input values.

  • SubqueryEndNode: end of a spliced (inlined) subquery.

  • SubqueryNode: executes a subquery.

  • SubqueryStartNode: beginning of a spliced (inlined) subquery.

  • TraversalNode: indicates a regular graph traversal, as opposed to a shortest path(s) traversal.

  • UpdateNode: updates documents in a collection (given in its collection attribute). Will appear exactly once in a query that contains an UPDATE statement.

  • UpsertNode: upserts documents in a collection (given in its collection attribute). Will appear exactly once in a query that contains an UPSERT statement.

For queries in the cluster, the following nodes may appear in execution plans:

  • DistributeNode: used on a coordinator to fan-out data to one or multiple shards, taking into account a collection’s shard key.

  • GatherNode: used on a coordinator to aggregate results from one or many shards into a combined stream of results. Parallelizes work for certain types of queries when there are multiple database servers involved (shown as GATHER /* parallel */ in query explain).

  • RemoteNode: a RemoteNode will perform communication with another ArangoDB instances in the cluster. For example, the cluster coordinator will need to communicate with other servers to fetch the actual data from the shards. It will do so via RemoteNodes. The data servers themselves might again pull further data from the coordinator, and thus might also employ RemoteNodes. So, all of the above cluster relevant nodes will be accompanied by a RemoteNode.

  • ScatterNode: used on a coordinator to fan-out data to one or multiple shards.

  • SingleRemoteOperationNode: used on a coordinator to directly work with a single document on a DB-Server that was referenced by its _key.

List of optimizer rules

The following optimizer rules may appear in the rules attribute of a plan:

  • fuse-filters: will appear if the optimizer merges adjacent FILTER nodes together into a single FILTER node

  • geo-index-optimizer: will appear when a geo index is utilized.

  • handle-arangosearch-views: appears whenever an ArangoSearch View is accessed in a query.

  • inline-subqueries: will appear when a subquery was pulled out in its surrounding scope, e.g. FOR x IN (FOR y IN collection FILTER y.value >= 5 RETURN y.test) RETURN x.a would become FOR tmp IN collection FILTER tmp.value >= 5 LET x = tmp.test RETURN x.a

  • interchange-adjacent-enumerations: will appear if a query contains multiple FOR statements whose order were permuted. Permutation of FOR statements is performed because it may enable further optimizations by other rules.

  • late-document-materialization: tries to read from collections as late as possible if the involved attributes are covered by regular indexes.

  • late-document-materialization-arangosearch: tries to read from the underlying collections of a View as late as possible if the involved attributes are covered by the View index.

  • move-calculations-down: will appear if a CalculationNode was moved down in a plan. The intention of this rule is to move calculations down in the processing pipeline as far as possible (below FILTER, LIMIT and SUBQUERY nodes) so they are executed as late as possible and not before their results are required.

  • move-calculations-up: will appear if a CalculationNode was moved up in a plan. The intention of this rule is to move calculations up in the processing pipeline as far as possible (ideally out of enumerations) so they are not executed in loops if not required. It is also quite common that this rule enables further optimizations to kick in.

  • move-filters-into-enumerate: moves filters on non-indexed collection attributes into IndexNode or EnumerateCollectionNode to allow early pruning of non-matching documents. This optimization can help to avoid a lot of temporary document copies.

  • move-filters-up: will appear if a FilterNode was moved up in a plan. The intention of this rule is to move filters up in the processing pipeline as far as possible (ideally out of inner loops) so they filter results as early as possible.

  • optimize-subqueries: will appear when optimizations are applied to a subquery. The optimizer rule will add a LIMIT statement to qualifying subqueries to make them return less data. Another optimization performed by this rule is to modify the result value of subqueries in case only the number of subquery results is checked later. This saves copying the document data from the subquery to the outer scope and may enable follow-up optimizations.

  • optimize-traversals: will appear if either the edge or path output variable in an AQL traversal was optimized away, or if a FILTER condition from the query was moved in the TraversalNode for early pruning of results.

  • patch-update-statements: will appear if an UpdateNode or ReplaceNode was patched to not buffer its input completely, but to process it in smaller batches. The rule will fire for an UPDATE or REPLACE query that is fed by a full collection scan or an index scan only, and that does not use any other collections, indexes, subqueries or traversals.

  • propagate-constant-attributes: will appear when a constant value was inserted into a filter condition, replacing a dynamic attribute value.

  • reduce-extraction-to-projection: will appear when an EnumerationCollectionNode or an IndexNode that would have extracted an entire document was modified to return only a projection of each document. Projections are limited to at most 5 different document attributes. This optimizer rule is specific for the RocksDB storage engine.

  • remove-collect-variables: will appear if an INTO clause was removed from a COLLECT statement because the result of INTO is not used. May also appear if a result of a COLLECT statement’s AGGREGATE variables is not used.

  • remove-data-modification-out-variables: avoids setting the pseudo-variables OLD and NEW if not used in data modification queries.

  • remove-filter-covered-by-index: will appear if a FilterNode was removed or replaced because the filter condition is already covered by an IndexNode.

  • remove-filter-covered-by-traversal: will appear if a FilterNode was removed or replaced because the filter condition is already covered by an TraversalNode.

  • remove-redundant-calculations: will appear if redundant calculations (expressions with the exact same result) were found in the query. The optimizer rule will then replace references to the redundant expressions with a single reference, allowing other optimizer rules to remove the then-unneeded CalculationNodes.

  • remove-redundant-or: will appear if multiple OR conditions for the same variable or attribute were combined into a single condition.

  • remove-redundant-path-var: avoids computing the variables emitted by traversals if they are unused in the query, significantly reducing overhead.

  • remove-redundant-sorts: will appear if multiple SORT statements can be merged into fewer sorts.

  • remove-sort-rand: will appear when a SORT RAND() expression is removed by moving the random iteration into an EnumerateCollectionNode. This optimizer rule is specific for the MMFiles storage engine.

  • remove-unnecessary-calculations: will appear if CalculationNodes were removed from the query. The rule will removed all calculations whose result is not referenced in the query (note that this may be a consequence of applying other optimizations).

  • remove-unnecessary-filters: will appear if a FilterNode was removed or replaced. FilterNodes whose filter condition will always evaluate to true will be removed from the plan.

  • replace-function-with-index: will appear when a deprecated index function such as FULLTEXT(), NEAR(), WITHIN() or WITHIN_RECTANGLE() is replaced with a regular subquery.

  • replace-or-with-in: will appear if multiple OR-combined equality conditions on the same variable or attribute were replaced with an IN condition.

  • simplify-conditions: will appear if the optimizer replaces parts in a CalculationNode’s expression with simpler expressions

  • sort-in-values: will appear when the values used as right-hand side of an IN operator will be pre-sorted using an extra function call. Pre-sorting the comparison array allows using a binary search in-list lookup with a logarithmic complexity instead of the default linear complexity in-list lookup.

  • sort-limit: will appear when a SortNode is followed by a LimitNode with no intervening nodes that may change the element count (e.g. a FilterNode which could not be moved before the sort, or a source node like EnumerateCollectionNode). This is used to make the SortNode aware of the limit and offset from the LimitNode to enable some optimizations internal to the SortNode which allow for reduced memory usage and and in many cases, improved sorting speed. The optimizer may choose not to apply the rule if it decides that it will offer little or no benefit. In particular it will not apply the rule if the input size is very small or if the output from the LimitNode is similar in size to the input. In exceptionally rare cases, this rule could result in some small slowdown. If observed, one can disable the rule for the affected query at the cost of increased memory usage.

  • splice-subqueries: will appear when a subquery has been spliced into the surrounding query. Only suitable subqueries can be spliced. A subquery becomes unsuitable if it contains a LIMIT node or a COLLECT WITH COUNT INTO … construct (but not due to a COLLECT var = <expr> WITH COUNT INTO …). A subquery also becomes unsuitable if it is contained in a (sub)query containing unsuitable parts after the subquery.

    This optimization is applied after all other optimizations, and reduces overhead for executing subqueries by inlining the execution. This mainly benefits queries which execute subqueries very often that only return a few results at a time.

  • use-index-for-sort: will appear if an index can be used to avoid a SORT operation. If the rule was applied, a SortNode was removed from the plan.

  • use-indexes: will appear when an index is used to iterate over a collection. As a consequence, an EnumerateCollectionNode was replaced with an IndexNode in the plan.

Some rules are applied a second time at a different optimization stage. These rules show in plans with an appended -2 to their name.

The following optimizer rules may appear in the rules attribute of cluster plans:

  • cluster-one-shard (Enterprise Edition only): will appear for eligible queries in OneShard deployment mode as well as for queries that only involve collection(s) with a single shard (and identical sharding in case of multiple collections, e.g. via distributeShardsLike). Queries involving V8 / JavaScript (e.g. user-defined AQL functions) can not be optimized.

    Offloads the entire query to the DBServer (except the client communication via a Coordinator). This saves all the back and forth that normally exists in regular cluster queries, benefitting traversals and joins in particular.

  • collect-in-cluster: will appear when a CollectNode on a coordinator is accompanied by extra CollectNodes on the database servers, which will do the heavy processing and allow the CollectNode on the coordinator to a light-weight aggregation only.

  • distribute-filtercalc-to-cluster: will appear when filters are moved up in a distributed execution plan. Filters are moved as far up in the plan as possible to make result sets as small as possible as early as possible.

  • distribute-in-cluster: will appear when query parts get distributed in a cluster. This is not an optimization rule, and it cannot be turned off.

  • distribute-sort-to-cluster: will appear if sorts are moved up in a distributed query. Sorts are moved as far up in the plan as possible to make result sets as small as possible as early as possible.

  • optimize-cluster-single-document-operations: it may appear if you directly reference a document by its _key; in this case no AQL will be executed on the DB-Servers, instead the coordinator will directly work with the documents on the DBServers.

  • parallelize-gather: will appears if an optimization to execute Coordinator GatherNodes in parallel was applied.

  • remove-satellite-joins (Enterprise Edition only): optimizes Scatter-, Gather- and RemoteNodes for Satellite Collections away. Depends on remove-unnecessary-remote-scatter rule.

  • remove-unnecessary-remote-scatter: will appear if a RemoteNode is followed by a ScatterNode, and the ScatterNode is only followed by calculations or the SingletonNode. In this case, there is no need to distribute the calculation, and it will be handled centrally.

  • restrict-to-single-shard: will appear if a collection operation (IndexNode or a data-modification node) will only affect a single shard, and the operation can be restricted to the single shard and is not applied for all shards. This optimization can be applied for queries that access a collection only once in the query, and that do not use traversals, shortest path queries and that do not access collection data dynamically using the DOCUMENT, FULLTEXT, NEAR or WITHIN AQL functions. Additionally, the optimizer will only pull off this optimization if can safely determine the values of all the collection’s shard keys from the query, and when the shard keys are covered by a single index (this is always true if the shard key is the default _key).

  • scatter-arangosearch-view-in-cluster: will appear when scatter, gather, and remote nodes are inserted into a distributed View query. This is not an optimization rule, and it cannot be turned off.

  • scatter-in-cluster: will appear when scatter, gather, and remote nodes are inserted into a distributed query. This is not an optimization rule, and it cannot be turned off.

  • smart-joins (Enterprise Edition only): will appear when the query optimizer can reduce an inter-node join to a server-local join. This rule is only active in the Enterprise Edition of ArangoDB, and will only be employed when joining two collections with identical sharding setup via their shard keys.

  • undistribute-remove-after-enum-coll: will appear if a RemoveNode can be pushed into the same query part that enumerates over the documents of a collection. This saves inter-cluster roundtrips between the EnumerateCollectionNode and the RemoveNode. From v3.6.0 on, it includes simple UPDATE and REPLACE operations that modify multiple documents and do not use LIMIT.

Note that some rules may appear multiple times in the list, with number suffixes. This is due to the same rule being applied multiple times, at different positions in the optimizer pipeline.

Additional optimizations applied

If a query iterates over a collection (for filtering or counting) but does not need the actual document values later, the optimizer can apply a “scan-only” optimization for EnumerateCollectionNodes and IndexNodes. In this case, it will not build up a result with the document data at all, which may reduce work significantly especially with the RocksDB storage engine. In case the document data is actually not needed later on, it may be sensible to remove it from query strings so the optimizer can apply the optimization.

If the optimization is applied, it will show up as “scan only” in an AQL query’s execution plan for an EnumerateCollectionNode or an IndexNode.

Additionally, the optimizer can apply an “index-only” optimization for AQL queries that can satisfy the retrieval of all required document attributes directly from an index.

This optimization will be triggered for the RocksDB engine if an index is used that covers all required attributes of the document used later on in the query. If applied, it will save retrieving the actual document data (which would require an extra lookup in RocksDB), but will instead build the document data solely from the index values found. It will only be applied when using up to 5 attributes from the document, and only if the rest of the document data is not used later on in the query.

The optimization is currently available for the RocksDB engine for the index types primary, edge, hash, skiplist and persistent.

If the optimization is applied, it will show up as “index only” in an AQL query’s execution plan for an IndexNode.