Hello.
As mentioned before, there are two problems with the WMS GetCapabilities operation. The first problem is that one of the wrappers (OracleDatastoreWrapper or PostgisDatastoreWrapper) between GeoServer and the datastore is not passing down the visitor (MinVisitor, MaxVisitor, UniqueVisitor, etc.) and thus broke the optimization selecting granules from the datastore. The problem has been solved by adding a "accepts" method to the TransformFeatureCollection class which allows to pass down the visitor and use optimized queries (see https://github.com/geotools/geotools/pull/2009). The second problem why GetCapabilities query works not enough fast is still not solved and needs to be discussed.

Details:
Our performance problem are caused by the queries to select the values of the dimensions. We have compared queries for the dimensions which have the different values of the properties "Presentation": "List" and "Interval and resolution". It was realized that the queries selecting the values of the dimension both presentations are the same and have the following view:         PostGIS (with WrapStore=false): SELECT distinct("<dim_name>") FROM "<table_name>"
        Oracle: SELECT distinct(<dim_name>) FROM <TABLE_NAME>
        PostGIS (with WrapStore=true): SELECT distinct("<dim_name>") FROM "<table_name>"

But if the "Interval and resolution" presentation of the dimension is selected then only the min/max values are really needed, which could be calculated much faster in the database.

Solutions:
The main idea is to inform GeoTools level modules about what value of the dimension presentation is selected and to change the visitors on the fly if the "Interval and resolution" dimension presentation was selected. The following changes of the gs-main module of the Geoserver and the gt-metadata and gt-imagemosaic modules of the GeoTools were proposed.

Information about dimension configuration are saving in the object of the CoverageInfo class. Object controls and invokes visitors to extract dimensions values instantiated RasterManager class. The CoverageInfo object and RasterManager object have the common namespace in the CoverDimensionCustomizerReader.java in the method wrap. The value of the dimension presentation can be sent from the GeoServer to the GeoTools using new RenderingHints property in the Hints object. The following is the proposed changes: - in the CoverDimensionCustomizerReader.java method (GeoServer) following code should be appended:
    public static GridCoverageReader wrap(GridCoverage2DReader delegate,
        String coverageName, CoverageInfo info) {
        GridCoverage2DReader reader = delegate;
        if (coverageName != null) {
            reader = SingleGridCoverage2DReader.wrap(delegate, coverageName);
        }
        /** Check if dimension has presentation List **/
        if(info != null && delegate instanceof ImageMosaicReader) {
            RasterManager manager = ((ImageMosaicReader) delegate).getRasterManager(info.getNativeCoverageName());
            if(manager != null) {
                Map<String, Boolean> dimPresentationData = new HashMap<>();
                manager.getHints().add(new RenderingHints(Hints.DIMENSIONS_PRESENTATIONS_INFO,dimPresentationData));                 for (Map.Entry<String, Serializable> e : info.getMetadata().entrySet()) {
                    String key = e.getKey();
                    Object value = e.getValue();
                    if (key.equals(org.geoserver.catalog.ResourceInfo.TIME)) {                         DimensionInfo timeInfo = Converters.convert(value, DimensionInfo.class);
                        dimPresentationData.put(key,
                            DimensionPresentation.LIST == timeInfo.getPresentation());                     } else if (key.equals(org.geoserver.catalog.ResourceInfo.ELEVATION)) {                         DimensionInfo elevInfo =Converters.convert(value, DimensionInfo.class);
                        dimPresentationData.put(key,
                            DimensionPresentation.LIST == elevInfo.getPresentation());
                    } else if (value instanceof DimensionInfo) {
                        DimensionInfo dimInfo = (DimensionInfo) value;
                        if (dimInfo.isEnabled()) {
                            if (key.startsWith(org.geoserver.catalog.ResourceInfo.CUSTOM_DIMENSION_PREFIX)) {                                 String dimensionName = key.substring(org.geoserver
                                    .catalog
                                    .ResourceInfo
                                    .CUSTOM_DIMENSION_PREFIX
                                    .length());
dimPresentationData.put(dimensionName.toLowerCase(),
                                    DimensionPresentation.LIST == dimInfo.getPresentation());
                            }
                        }
                    }
                }
            }
        }
        /************************************************/
        if (reader instanceof StructuredGridCoverage2DReader) {
            return new CoverageDimensionCustomizerStructuredReader((StructuredGridCoverage2DReader) reader, coverageName, info);
        } else {
            return new CoverageDimensionCustomizerReader(reader, coverageName, info);
        }
    }

- in the Hints.java (GeoTools) following code should be appended:

    public static final ClassKey DIMENSIONS_PRESENTATIONS_INFO = new ClassKey("java.util.Map");

The following changes of the code allow to control value of selected presentation and allow to change the visitors to select only the minimum and the maximum values if the "Interval and resolution" dimension presentation was selected.
- in the RasterManager.java (GeoTools) following code should be appended:

 Set extractDomain(final String attribute) throws IOException {
        Query query = new Query(typeName);
        query.setPropertyNames(Arrays.asList(attribute));
        final UniqueVisitor visitor = new UniqueVisitor(attribute);
     /*** control the presentation of the dimansion ***/
        Hints rmHints = getHints();
        if (rmHints.containsKey(Hints.DIMENSIONS_PRESENTATIONS_INFO)) {
            Map<String, Boolean> dimPresentationParam =
                    (Map<String, Boolean>) rmHints.get(Hints.DIMENSIONS_PRESENTATIONS_INFO);
            if (dimPresentationParam != null) {
                Boolean isListPresentation = dimPresentationParam.get(attribute);
                if (!isListPresentation) {
                    final FeatureCalc[] visitors = {
                        new MinVisitor(attribute), new MaxVisitor(attribute)
                    };
                    Set values = new HashSet();

                    for (FeatureCalc extremVisitor : visitors) {
granuleCatalog.computeAggregateFunction(query, extremVisitor);
values.add(extremVisitor.getResult().getValue());
                    }

                    return values;
                }
            }
        }
    /*******/
        granuleCatalog.computeAggregateFunction(query, visitor);
        return visitor.getUnique();
    }

We would like to get some feedback on will these changes correct or are there another parts of GeoServer/GeoTools where GeoServer can inform GeoTools about selected presentation and GeoTools can change the visitors to obtain the "Interval and resolution" values of the dimension?

Best regards,
Ivan

Dňa 09.08.2018 o 18:52 Andrea Aime napísal(a):
Ah,
by looking at it, it would seems the store wrapper in the mosaic module is not correctly
delegating down the visitor and thus breaking database optimizations.
Yep, pull requests are quite welcomed, please pay attention to the rules to contribute,
detailed here:
https://github.com/geotools/geotools/blob/master/CONTRIBUTING.md

In particular, note code formatting (just running maven on the command
line will reformat the code as expected), presence of tests, and contribution agreement

Cheers
Andrea


On Thu, Aug 9, 2018 at 6:30 PM György Tomcsányi <gyorgy.tomcsa...@microstep-mis.com <mailto:gyorgy.tomcsa...@microstep-mis.com>> wrote:

    Hello all,

    we are using GeoServer to display a large number of GeoTIFFs using
    ImageMosaic data stores. The data has several dimensions (time and
    custom). We are adding new data to these stores periodically and the
    goal is to be able to display years of data (millions of
    granules). We
    have encountered performance problems with the WMS GetCapabilities
    operation. It is most noticeable when using Oracle database (or
    PostGIS
    with parameter WrapStore=true). My colleague Ivan (in CC) implemented
    changes which improved the performance for our use case
    significantly.
    We would like to publish these changes and hopefully merge them
    into the
    official repo.

    Details:
    our performance problems are caused by the queries for values and
    defaults for dimensions. The currently used queries vary depending on
    the selected presentation for the dimension:

    1. query for values:
       "List":
         * PostGIS (with WrapStore=false): SELECT distinct("<dim_name>")
    FROM "<table_name>"
         * Oracle: SELECT FID, <DIM_NAME> FROM <table_name>
         * PostGIS (with WrapStore=true): SELECT "fid","<dim_name>" FROM
    "<table_name>"
         First version is clearly the fastest, because it loads only
    distinct values.

    "Interval and resolution":
       The queries are the same as for "List", but in this case we only
    really need the min/max values, which could be calculated much
    faster in
    the database.

    2. query for defaults (when using smallest/biggest domain value):
       * PostGIS (with WrapStore=false): SELECT <min | max>
    ("<dim_name>")
    FROM "<table_name>"
       * Oracle: SELECT FID, <DIM_NAME> FROM <table_name>
       * PostGIS (with WrapStore=true): SELECT "fid","<dim_name>" FROM
    "<table_name>"
       Again the first PostGIS variant is clearly the fastest. Others
    load
    all the values to the application where the min/max values are
    calculated.

    Our proposed solution creates optimized queries depending on which
    presentation and default setting is used for the dimension. The
    needed
    additional information from the settings is sent to the lower level
    functions using new RenderingHints. The goal of the optimized
    queries is
    to load only the needed values from the database (usually what the
    current PostGIS query does). Unit tests were also added.

    We would like to get some feedback on this. Can I submit merge
    requests
    (to both GeoTools and GeoServer) to review the code?
    We are looking forward to any comments.

    best regards,
    Gyorgy Tomcsanyi

    
------------------------------------------------------------------------------
    Check out the vibrant tech community on one of the world's most
    engaging tech sites, Slashdot.org! http://sdm.link/slashdot
    _______________________________________________
    GeoTools-Devel mailing list
    GeoTools-Devel@lists.sourceforge.net
    <mailto:GeoTools-Devel@lists.sourceforge.net>
    https://lists.sourceforge.net/lists/listinfo/geotools-devel



--

Regards, Andrea Aime == GeoServer Professional Services from the experts! Visit http://goo.gl/it488V for more information. == Ing. Andrea Aime @geowolf Technical Lead GeoSolutions S.A.S. Via di Montramito 3/A 55054 Massarosa (LU) phone: +39 0584 962313 fax: +39 0584 1660272 mob: +39 339 8844549 http://www.geo-solutions.it http://twitter.com/geosolutions_it ------------------------------------------------------- /Con riferimento alla normativa sul trattamento dei dati personali (Reg. UE 2016/679 - Regolamento generale sulla protezione dei dati “GDPR”), si precisa che ogni circostanza inerente alla presente email (il suo contenuto, gli eventuali allegati, etc.) è un dato la cui conoscenza è riservata al/i solo/i destinatario/i indicati dallo scrivente. Se il messaggio Le è giunto per errore, è tenuta/o a cancellarlo, ogni altra operazione è illecita. Le sarei comunque grato se potesse darmene notizia. This email is intended only for the person or entity to which it is addressed and may contain information that is privileged, confidential or otherwise protected from disclosure. We remind that - as provided by European Regulation 2016/679 “GDPR” - copying, dissemination or use of this e-mail or the information herein by anyone other than the intended recipient is prohibited. If you have received this email by mistake, please notify us immediately by telephone or e-mail./



------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot


_______________________________________________
GeoTools-Devel mailing list
GeoTools-Devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/geotools-devel

------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
GeoTools-Devel mailing list
GeoTools-Devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/geotools-devel

Reply via email to