DataFed Service Chain Example

From Earth Science Information Partners (ESIP)
WCS python demos by Hoijarvi
For questions regarding this page e-mail Hoijarvi


Service Chaining Example: Outline

This example uses python to

  • get point data
  • render it as an image
  • store the image on local computer

See WCS page SOAPifying_WCS

Getting Demo Script

Download and save it with extension .py

type from command line python and watch it run.

There are no parameters, the script is hard coded for demo purposes.

SOAP problems

There are two problems with SOAP and large amounts of data.

  • First, data has to be passed as xml, which results to enormous amounts of data for large tables, causing performance problems.
  • Second, SOAP defines 4 MB size limit for messages, so large data transfers have to partitioned, making things more complicated.
  • Third problem is WCS binary data, although binary encoding with MIME attachments is possoble, I do not know any system that would use it.

Datafed services have solved the issue by using common two phase pattern. The SOAP services return an envelope with small amount of metadata describing the result, and an uri pointer to the cached result.

Datafed SOAP services use this pattern also while passing data into services. For example in this chase, the table url is passed from WCS to RenderMapPoint directly. Since the services are located in the same machine, they can access the table directly, without turning it into xml ever.

Service Chain Example: Python Script

Function execute_chain

This is the main program for the service chain.

  • Query data with WCS,
  • Render it,
  • save locally

Since the output envelope does not contain data, but a reference to data, get_table_url function returns the reference url to the data table, which is then passed into the rendering service. It also returns a reference to the image, which is still in the server.

So the services don't pass data to each others, they pass a little metadata and a url to the data. If the services are in the same server, this bypasses reading and writing huge amounts of xml data.

def execute_chain():
    print "querying table"
    table_url = get_table_url()

    # table_url ==
    # rendering service can get this directly from the cache with the ID GetCoverage_91

    print "rendering table "   table_url
    image_url = get_image_url(table_url)

    # image_url ==
    # image processing services could again get the big image directly from
    # the server cache with ID RenderMapPoint_92
    # Now we get the result from the server and save it on the local drive.

    print "fetching image "   image_url
    image_stream = urllib2.urlopen(image_url)
        print "writing image to file"
        dump_stream(image_stream, "soapchain.png")

Query Template for WCS

get_coverage_query = """
<soap:Envelope xmlns:soap="">
        <wcs:GetCoverage version="1.0.0" service="WCS"
                xmlns:gml="" xmlns:wcs="">
            <!-- datafed dataset_abbr.param_abbr defines the coverage name -->
                        This element queries the USA.
                        The dataset has no elevation, so only lat and lon
                        limits are needed. 
                    <gml:Envelope srsName="WGS84(DD)">
                        <gml:pos>[[lon_min]] [[lat_min]]</gml:pos>
                        <gml:pos>[[lon_max]] [[lat_max]]</gml:pos>
                    <gml:Grid dimension="2">
                                grid size. This is a point dataset, so these numbers have no meaning.
                                <gml:low>0 0</gml:low>
                                <gml:high>99 99</gml:high>
                        query data for one time only.
                .NET dataset is a good format for point datasets

Query Template for Render Call

render_point_query = """
<soap:Envelope xmlns:soap="">
        <rmp:Render xmlns:rmp="">
            <Table xmlns="">
            <rmp:Settings xmlns:mi=""

Other Utility Functions

  • query_datafed: A generic soap call to datafed server.
  • replace_parameters: If you want to change most common parameters.
  • look_for_ns_name: xml utility
  • dump_stream: write image stream to file
  • get_table_url, get_image_url: soap calls