GeoServer on steroids - GeoSolutions Demo Server

GeoServer on steroids All you wanted to know about how to make GeoServer faster but you never asked (or you did and no one answered)

Ing. Andrea Aime, GeoSolutions Ing. Simone Giannecchini, GeoSolutions

FOSS4G 2011, Denver 12th-16th September 2011

GeoSolutions 

Founded in Italy in late 2006



Expertise







•

Image Processing, GeoSpatial Data Fusion

•

Java, Java Enterprise, C++, Python

•

JPEG2000, JPIP, Advanced 2D visualization

Supporting/Developing FOSS4G projects 

GeoTools, GeoServer



GeoBatch, GeoNetwork

Clients 

Public Agencies



Private Companies

http://www.geo-solutions.it FOSS4G 2011, Denver 12th-16th September 2011

Preparing raster inputs


Raster Data CheckList 







Objectives  Fast extraction of a subset of the data  Fast extraction of overviews Check-list  Avoid having to open a large number of files per request  Avoid parsing of complex structures  Avoid on-the-fly reprojection (if possible) Get to know your bottlenecks  CPU vs Disk Access Time vs Memory Experiment with  Format, compression, different color models, tile size, overviews, configuration (in GeoServer of course) FOSS4G 2011, Denver 12th-16th September 2011

Problematic Formats 



PNG/JPEG direct serving  Bad formats (especially in Java)  No tiling (or rarely supported)  Chew a lot of memory and CPU for decompression  Mitigate with external overviews NetCDF/grib1 and similar formats  Complex formats (often with many subdatasets)  Often contains un-calibrated data  Must usually use multiple dimensions 



Use ImageMosaic

Must usually massage the data before serving 

e.g. transpose X,Y, FOSS4G 2011, Denver 12th-16th September 2011

Problematic Formats 

Ascii Grid, GTOPO30, IDRISI and similar formats are bad 





No internal tiling, no compression, no internal overviews

JPEG2000 (with Kakadu) 





ASCII formats are bad

Extensible and rich, not (always) fast Can be difficult to tune for performance (might require specific encoding options)

ECW and MrSID 

Why bother it’s proprietary? FOSS4G 2011, Denver 12th-16th September 2011

Choosing Formats and Layouts 







To remember: GeoTiff is a swiss knife  But you don’t want to cut a tree with it!  Tremendously flexible, good fir for most (not all) use cases  BigTiff pushes the GeoTiff limits farther Single File VS Mosaic VS Pyramids Use single GeoTiff when  Overviews and Tiling stay within 4GB  No additional dimensions Consider BigTiff for very large file (> 4 GB)  Support for tiling  Support for Overviews  Can be inefficient with very large files + small tiling FOSS4G 2011, Denver 12th-16th September 2011


Use ImageMosaic when: 



 





A single file gets too big (inefficient seeks, too much metadata to read, etc..) Multiple Dimensions (time, elevation, others..) Avoid mosaics made of many very small files Single granules can be large Use Tiling + Overviews + Compression on granules

Use ImagePyramid when: 

Tremendously large dataset 



Need to serve at all scales 



Too many files / too large files Especially low resolution

For single granules (< 2Gb) GeoTiff is generally a good fit FOSS4G 2011, Denver 12th-16th September 2011


Examples: 

Small dataset: single 2GB GeoTiff file



Medium dataset: single 40GB BigTiff





Large dataset: 400GB mosaic made of 10GB BigTiff files Extra large: 4TB of imagery, built as pyramid of mosaics of BigTiff/GeoTiff files to keep the file count low


GeoTiff preparation 

STEP 0: get to know your data



gdalinfo utility is your friend

 

CheckList Missing CRS  



Missing georeferencing  



Fix with gdal_translate

Missing Overviews 



Add a World File Fix with gdal_translate

Bad Tiling 



Add a .prj file Fix with gdal_translate

Use gdaladdo

Compression 


Use gdal_translate




STEP 1: fix and optimize with gdal_translate Inner Tiling 





CRS and GeoReferencing 



gdal_translate -co "TILED=YES" -co "BLOCKXSIZE=512" -co "BLOCKYSIZE=512" in.tif out.tif Check also GeoTiff driver creation options here

gdal_translate –a_srs “EPSG:32619” –a_ullr 285409.2 2014405.2 287536.8 2011947.6 in.tif out.tif

STEP 2: add overviews with gdal_addo 



 

Leverages on tiff support for multipage files and reduced resolution pages gdaladdo -r cubic output.tif 2 4 8 16 32 64 128 Choose the resampling algorithm wisely Chose the tile size and compression wisely (use GDAL_TIFF_OVR_BLOCKSIZE)



Consider external overviews FOSS4G 2011, Denver 12th-16th September 2011

GeoTiff preparation







Compression 

Consider when disk speed/space is an issue



Control it with gdal_translate and creation options

GeoTiff tiles can be compressed 

LZW/Deflate are good for lossless compression



JPEG is good for visually lossless compression

From experience 



Use LZW/Deflate on geophysical data (DEM, acquisitions) USE JPEG visually lossless with Photometric Interpretation to YCbCr for RGB FOSS4G 2011, Denver 12th-16th September 2011

Time, Elevation and other dimensions 



Use Cases: 

MetOc data (support for time, elevation)



Data with additional indipendent dimensions

WorkFlow 

Split in multiple GeoTiff files



Optimize the files individually



Use ImageMosaic



Use a DBMS for indexing granules







Use File Name based property collectors to turn properties into DB rows attributes Filter by time, elevation and other attributes via OGC and CQL filters

Check back up slides for more info! FOSS4G 2011, Denver 12th-16th September 2011

Time, Elevation and other dimensions 

Indexing multiple dimensions with DB support (video here)

datastore.properties

timeregex.properties

stringregex.properties indexer.properties


Time, Elevation and other dimensions


Proper Mosaic Preparation 





ImageMosaic stitches single granules together with basic processing 

Filtered selection



Overviews/Decimation on read



Over/DownSampling in memory



ColorMask (optional)



Mosaic/Stitch



ColorMask again (optional)

Optimize files as if you were serving them individually Keep a balance between number and dimensions of granules FOSS4G 2011, Denver 12th-16th September 2011

Proper Mosaic Configuration 



STEP 0: Configure Coverage Access (see slide 22) STEP 1: Configure Mosaic Parameters  ALLOW_MULTITHREADING 





Use a proper Tile Size 





In-memory processing, must not be too large Disk tiling should larger

If memory is scarce:  



Load data from different granules in parallel Needs USE_JAI_IMAGE_READ set to false (Immediate Mode)

USE_JAI_IMAGREAD to true USE_MULTITHREADING to false*

Otherwise  

USE_JAI_IMAGREAD to false ALLOW_MULTITHREADING to true


Proper Mosaic Configuration 

Optional (Advanced): Configure Mosaic Parameters Directly  Caching  

 



Load the index in memory (using JTS SRTree) Super fast granule lookup, good for shapefiles Bad if you have additional dimension to filter on Based on Soft References, controlled via Java switch SoftRefLRUPolicyMSPerMB

ExpandToRGB 



Expand colormapped imagery to RGB in memory Trade performance for quality  



SuggestedSPI Default ImageIO Decoder class to use Don’t touch unless expert


Proper Pyramid Preparation 

Use gdal_retile for creating the pyramid



Prepare the list of tiles to be retiled



Create the pyramid with GDAL retile (grab a coffee!)









Chunks should not be too small (here 2048x2048)  Too many files is bad anyway Use internal Tiling for Larger chunks size If the input dataset is huge use the useDirForEachRow option  Too many files in a dir is bad practice Make sure the number of level is consistent  Too few  bad performance at high scale FOSS4G 2011, Denver 12th-16th September 2011

Proper Pyramid Configuration 



STEP 0: Configure Coverage Access (see slide 22) STEP 1: Configure Pyramid Parameters  ALLOW_MULTITHREADING 





Use a proper Tile Size 









ImagePyramid relies on ImageMosaic

In-memory processing, must not be too large Disk tiling should larger

If memory is scarce: 



Load data from different granules in parallel Needs USE_JAI_IMAGE_READ set to false (Immediate Mode)

USE_JAI_IMAGREAD to true USE_MULTITHREADING to false*

Otherwise  

USE_JAI_IMAGREAD to false ALLOW_MULTITHREADING to true


Proper Pyramid Configuration 

Optional (Advanced): Configure Mosaic Parameters Directly  Caching  

 



Load the index in memory (using JTS SRTree) Super fast granule lookup, good for shapefiles Bad if you have additional dimension to filter on Based on Soft References, controlled via Java switch SoftRefLRUPolicyMSPerMB

ExpandToRGB 



Expand colormapped imagery to RGB in memory Trade performance for quality  



SuggestedSPI Default ImageIO Decoder class to use Don’t touch unless expert


Proper GDAL Formats Configuration 

 

Fix Missing/Improper CRS with PRJ or coverage config Fix Missing GeoReferencing with World File Make sure GDAL_DATA is properly configured  Use a proper Tile Size 







If memory is scarce: 





In-memory processing, must not be too large Fundamental for striped data! JNI overhead Disk tiling should larger USE_JAI_IMAGREAD to true USE_MULTITHREADING to true*

Otherwise 



USE_JAI_IMAGREAD to false USE_MULTITHREADING is ignored


Proper JPEG2000 Kakadu Configuration 

 

Fix Missing/Improper CRS with PRJ or coverage config Fix Missing GeoReferencing with World File Make sure Kakadu dll/so is properly loaded  Use a proper Tile Size  





If memory is scarce:  



In-memory processing Must not be too large Disk tiling should larger USE_JAI_IMAGREAD to true USE_MULTITHREADING to true*

Otherwise 



USE_JAI_IMAGREAD to false USE_MULTITHREADING is ignored


Proper GeoServer Coverage Options Configuration 















Make sure native JAI and Image is installed

Enable ImageIO native acceleration Enable JAI Mosaicking native acceleration Give JAI enough memory Don’t raise JAI memory Threshold too high Rule of thumb: use 2 X #Core Tile Threads (check next slide)

Enable Tile Recycling only on trunk Enable Tile Recycling if memory is not a problem FOSS4G 2011, Denver 12th-16th September 2011

Proper GeoServer Coverage Options Configuration 











Multithreaded Granule Loading Allows to fine tuning multithreading for ImageMosaic Orthogonal to JAI Tile Threads Rule of Thumb: use 2 X #Core Tile Threads Perform testing to fine tune depending on layer configuration as well as on typical requests

ImageIO Cache threshold 

decide when we switch to disk cache (very large WCS requests)


Reprojection Performance Vs Quality 









GeoServer 2.1.x reprojects raster data using a piecewiselinear algorithm The area is divided in rectangular blocks, each having its own affine transform

The transformation between the full trigonometric expressions and the linear ones is driven by a tolerance, default value is 0.333 Larger value will make reprojection faster, but lower the quality -Dorg.geotools.referencing.resampleTolerance=0.5


Preparing vector inputs


Vector data checklikst 

What do we want from vector data: 

Binary data



No complex parsing of data structures



Fast extraction of a geographic subset



Fast filtering on the most commonly used attributes


Choosing a format 

Slow formats 

WFS



GML



DXF



Good formats, local and indexable 

Shapefile



Directory of shapefiles



SDE



Spatial databases: PostGIS, Oracle Spatial, DB2, MySQL*, SQL server*


Shapefiles vs DBMS 

Speed comparison vs spatial extent depicted:  



Shapefile 



Shapefile very fast when rendering the full dataset Database faster when extracting a small subset of a very large data set

no attribute indexing, avoid if filtering on attribute is important (filtering == reading less data, not applying symbols)

Database  



Rich support for complex native filters Use connection pooling (preferably via JNDI) Validate connections (with proper pooling)


Shapefile preparation 



Remove .qix file if present, let GeoServer 2.1.x rebuild it (more efficient) If there are large DBF attributes that are not in use, get rid of them using ogr2ogr, e.g.: ogr2ogr -select FULLNAME,MTFCC arealm.shp tl_2010_08013_arealm.shp



If on Linux, enable memory mapping, faster, more scalable (but will kill Windows):


Shapefile filtering 



Stuck with shapefiles and have scale dependent rules like the following? 

Show highways first



Show all streets when zoomed in

Use ogr2ogr to build two shapefiles, one with just the highways, one with everything, and build two layers, e.g.: ogr2ogr -sql "SELECT * FROM tl_2010_08013_roads WHERE MTFCC in ('S1100', 'S1200')" primaryRoads.shp tl_2010_08013_roads.shp


PostGIS specific hints 







PostgreSQL out of the box configured for very small hardware: http://wiki.postgresql.org/wiki/Performance_Optimization

Make sure to run ANALYZE after data imports (updates optimizer stats) As usual, avoid large joins in SQL views, consider materialized views If the dataset is massive, CLUSTER on the spatial index: 



http://postgis.refractions.net/documentation/manual1.3/ch05.html

Careful with prepared statements (bad performance) FOSS4G 2011, Denver 12th-16th September 2011

Optimize styling


Use scale dependencies 

Never show too much data 

the map should be readable, not a graphic blob. Rule of thumb: 1000 features max in the display


Labeling 





Labeling conflict resolution is expensive, limit to the most inner zooms Halo is important for readability, but adds significant overhead Careful with maxDisplacement, makes for various label location attempts


FeatureTypeStyle 



GeoServer uses SLD FeatureTypeStyle objects as Z layers for painting Each one allocates its own rendering surface (which can use a lot of memory), use as few as possible


Use translucency sparingly 

Translucent display is expensive, use it sparingly


Scale dependent rules 

Too often forgotten or little used, yet very important: 







Hide layers when too zoomed in (raster/vector example) Progressively show details Add more expensive rendering when there are less features

Key to any high performance / good looking map


Example


Hide as you zoom in 



Add a MinScaleDenominator to the rule This will make the layer disappear at 1:75000 (towards 1:1)


Alternative rendering 



Simple rendering at low scale (up to 1:2000) More complex rendering when zoomed in (1:1999 and above)


Alternative rendering


Point symbols

• 600 loc for 6 different points types • Painful… FOSS4G 2011, Denver 12th-16th September 2011

Prepare data 







alter table pointlm add column image varchar;

update pointlm set image = 'shop_supermarket.p.16.png' where MTFCC = 'C3081' and (FULLNAME like '%Shopping%' or FULLNAME like '%Mall%'); update pointlm set image = 'peak.png' where MTFCC = 'C3022'

update pointlm set image = 'amenity_prison.p.20.png' where MTFCC = 'K1236';



update pointlm set image = 'museum.p.16.png' where MTFCC = 'K2165';



update pointlm set image = 'airport.p.16.png' where MTFCC = 'K2451';



update pointlm set image = 'school.png' where MTFCC = 'K2543';





update pointlm set image = 'christian3.p.14.png' where MTFCC = 'K2582';

update pointlm set image = 'gate2.png' where MTFCC = 'K3066';


Dynamic symbolizers


Output tuning


WMS output formats JPEG

PNG 8bit

PNG 24bit

23.8KB

66KB

169.4KB

27KB

27KB

64KB

Compression artifacts

Color reduction FOSS4G 2011, Denver 12th-16th September 2011

Large size

WFS output formats 35 30 25 20 15 10 5 0

Dimension MB 

HTTP GZip compression is transparent in GeoServer, make sure proxies keep it (or pay 10x price) FOSS4G 2011, Denver 12th-16th September 2011

Tile caching


Tile caching with GeoWebCache 

Tile oriented maps, fixed zoom levels and fixed grid



Useful for stable layers, backgrounds



Protocols: WMTS, TMS, WMS-C, Google Maps/Earth, VE





Speedup compared to dynamic WMS: 10 to 100 times, assuming tiles are already cached (whole layer preseeded) Suitable for: 



Mostly static layer No (or few) dynamic parameters (CQL filters, SLD params, SQL query params, time/elevation, format options) FOSS4G 2011, Denver 12th-16th September 2011

Embedded GWC advantage 

No double encoding when using meta-tiling, faster seeding


Space considerations 

Seeding Colorado, assuming 8 cores, one layer, 0.1 sec 756x756 metatile, 15KB for each tile



Do yours: http://tinyurl.com/3apkpss



Not enough disk space? Set a disk quota Zoom level 13 14 15 16 17 18 19 20

Tile count

Size (MB)

58,377 232,870 929,475 3,713,893 14,855,572 59,396,070 237,584,280 950,273,037

1 4 14 57 227 906 3,625 14,500

Time to seed Time to seed (hours) (days) 0 0 0 0 0 0 1 0 6 0 23 1 92 4 367 15


Resource control


WMS request limits 





Max memory per request: avoid large requests, allows to size the server memory (max concurrent request * max memory)

Max time per request: avoid requests taking too much time (e.g., using a custom style provided with dynamic SLD in the request) Max errors: best effort renderer, but handling errors takes time


WFS request limits 

Max feature returned, configured as a global limit



Return feature bbox: reduce amount of generated GML



Per layer max feature count


WCS request limits


Control flow 



Control how many requests are executed in parallel, queue others: 

Increase throughput



Control memory usage



Enforce fairness

More info here FOSS4G 2011, Denver 12th-16th September 2011

Control flow

17%

$GEOSERVER_DATA_DIR/controlflow.properties # don't allow more than 16 GetMap requests in parallel ows.wms.getmap=16 FOSS4G 2011, Denver 12th-16th September 2011

Auditing 

Log each and every request



Log contents driven by customizable template



Summarize and analyze requests with offline tools



More info here


JVM and deploy configuration


Premise 





The options discussed here are not going to help visibly if you did not prepare the data and the styles They are finishing touches that can get performance up once the major data bottlenecks have been dealt with

Check “Running in production” instructions here


JVM settings 









--server: enables the server JIT compiler --Xms2048m -Xmx2048m: sets the JVM use two gigabytes of memory --XX:+UseParallelOldGC -XX:+UserParallelGC: enables multi-threaded garbage collections, useful if you have more than two cores --XX:NewRatio=2: informs the JVM there will be a high number of short lived objects --XX:+AggressiveOpt: enable experimental optimizations that will be defaults in future versions of the JVM


Native JAI and JDK 



Install native JAI and use a recent Sun JDK! Benchmark over a small data set (the effect is not as visible on larger ones)


Setup a local cluster 

Java2D locks when drawing antialiased vectors 



Limits scalability severely

Use Apache mod_proxy_balance and setup a GeoServer each 2/4 cores mod_proxy_balance

GeoServer

GeoServer

GeoServer


Clustering advantage 



FOSS4G 2010 vector benchmarks (roads/buildings/isolines and so on, over the entire Spain) GeoServer was benchmarked without local clustering

66%


Benchmarking


Using JMeter 



Good benchmarking tool Allows to setup multiple thread groups, different parallelelism and request count, to ramp up the load



Can use CSV files to generate semi-randomized requests



Reports results in a simple table

http://jakarta.apache.org/jmeter/


Using JMeter 









Thread group: how many threads Loop: how many requests

HTTP sampler: the request CSV: read request params from CSV Summary table


Generating the CSV 



Simple randomized generation tool built during WMS shootouts, wms_request.py Generate csv with the bbox and width/height to be used in JMeter scripts: ./wms_request.py -count 1200 -region -180 -90 180 90 -minres 0.002 -maxres 0.1 -minsize 256 256 -maxsize 1024 1024



Get it here along with a corresponding JMeter script: http://demo1.geo-solutions.it/share/jmeter_2011.zip


Checking results 

Results table



Run the benchmarks 2-3 times, let the results stabilize



Save the results, check other optimizations, compare the results


Real world deploy


Deploy configuration


Raster data 







Whole Italy at 50cm per pixel Over 4TB, updated fully every 3 years (old data still available for historical access) Custom pyramid 

100 m per pixel: one image



20m per pixel: mosaic of 20 tiles



4m per pixel: mosaic of few hundred tiles



0.5m per pixel: 9000 tiles

Each tile is 10000x10000, with overviews


Vector data 









Cadastral data for the whole Italy, with full history (interval of validity for each parcel) 100 million polygons A query extracts a subset relative to a certain time interval and area the user is allowed to see

No data from this table is ever shown below 1:50000 (SLD scale dependencies) Physical table level partitioning (Oracle style) of the table based on geographic area to parallelize and cluster data loading, plus spatial indexing and indexes on commonly filtered upon attributes


The End

Questions? [email protected]

[email protected] FOSS4G 2011, Denver 12th-16th September 2011

GeoServer on steroids - GeoSolutions Demo Server

GeoServer on steroids - GeoSolutions Demo Server

Suggest Documents