Most Expensive Queries
|
The URL and total time in milliseconds for the ten queries
with the largest total computation time (that is, queue time plus Dgraph
processing time plus write time) made in the session. The queries are ordered
by processing time.
Each time a new Dgraph transaction that yields results is
completed, this tab may become updated with a new query, if it makes the list
of current top ten most expensive queries.
Each query is described with these characteristics:
Unlike in Presentation API mode, where the URL contains all of
the information about the query, in Web services mode the URL only contains the
service name. The bulk of the query is contained in the POST body. Therefore,
if the Dgraph is running in Web services mode, a serial number is appended to
the URL, as in the following example:/ws/myservice:57 .
This serial number corresponds to the HTTP Exchange ID in the
MDEX Engine Request Log. You can use it to retrieve additional information
about the contents of the query from the Request Log’s Query Body field.
|
Hotspots
|
Details on the performance of specific features, such as
clustering, record search, record filter, range filter, content spotlighting
and snippeting.
This section also contains the following page render and
record sorting statistics:
Page
render total. After the MDEX Engine knows which records and values
must be returned, this time represents the total time spent generating and
returning those results to the Presentation API. This time includes retrieving
records from memory or disk, ordering them based on the specified sort or
relevance ranking strategies, as well as other information returned to the API,
such as content spotlighting results.
Prefetching horizontal record. The cost to retrieve records from
the data layer of the MDEX Engine.
Statistics
related to various sorting strategies. The MDEX Engine examines information
about the data being returned and selects the best sorting strategy.
NoteThese statistics may change. They are used for
internal debugging and tuning of the MDEX Engine sorting selection strategy and
are not useful to the end user.
|
Results
|
The following items are listed in the Results
section. The statistics includes count, average, standard deviation, min, max
and total, where applicable:
Number of records
in result set
Result page size
in bytes
Result page format
performance in milliseconds
|
Server
|
Statistical information for the MDEX Engine server:
HTTP: Total
request time
HTTP: Time reading
request
HTTP: Time in
scheduler
HTTP: Time writing
response
HTTP: Request
bytes read (including HTTP overhead)
HTTP: Response
body size (including HTTP overhead)
Scheduler: Queue
time before processing
Scheduler:
Processing time
Scheduler: Queue
time after processing
Scheduler: Queries
queued
This metric describes the queue length.
Scheduler: Queries
in process
This metric describes the number of queries that are in
process.
Scheduler: Update
queue time
XQuery: Total time
in XQuery engine
XQuery: Total time
in XQuery external functions
NoteThis statistic only includes the time spent in the
following functions:
internal:query() ,
mdex:dimension-value-id-from-path() , and
mdex:add-navigation-descriptors() .
XQuery: Time
retrieving documents with
fn:doc()
XQuery: Time
storing documents with
fn:put()
XQuery: Result
serialization time
Most expensive MAX
invocations
Custom timing list
This metric, which can list things like expensive queries,
only appears when you implement custom metric gathering with the
ep:stats-timing pragma. See the
Web Services and XQuery Developer’s Guide for
more information.
|
Navigation
|
Information about the number of navigation
pages, as well as navigation performance, query size, and result size by
average, standard deviation, minimum, maximum, and total.
|
Record Sorting
|
The number and type of sorts performed (does
not include timing), and the percentage of those sorts for each sort type.
|
Analytics
|
Information pertaining to the analytics
features in Endeca Analytics, such as total processing time, query parsing,
time checking and evaluation times.
|
Disk usage
|
Disk usage statistics for the indices:
|
Search
|
A finer-grained analysis of the performance
of individual features. This information is used for internal analysis by
Oracle.
|
Data Layer Performance
|
Statistical information about the data layer
performance. This information is used for internal analysis by Oracle.
|