1. Efficient XML Aaron Braeckel Briefing to FAA Program Management 28 Dec 2008 National Center for Atmospheric Research Boulder, CO Copyright © 2008-2009 University Corporation for Atmospheric Research

2. 2 NextGen Network Enabled Weather (NNEW) Background and Motivation Extensible Flexible Human-readable Validatable Can be transformed Namespaced … XML Benefits: Copyright © 2008-2009 University Corporation for Atmospheric Research

3. 3 NextGen Network Enabled Weather (NNEW) Background and Motivation XML Issues: Copyright © 2008-2009 University Corporation for Atmospheric Research One of the unique (and difficult) weather problems is data volume Processing efficiency Encoding time Decoding time Message throughput Battery life (handheld devices, etc) Compactness Storage Archival Bandwidth/transmission Has a (non-trivial) recurring cost in some cases Next-generation aircraft Handheld devices

4. 4 NextGen Network Enabled Weather (NNEW) Background and Motivation XML Assessments: Copyright © 2008-2009 University Corporation for Atmospheric Research DoD Comparison (binary to XML) 1 10x, 17x, 22x, and 121x larger with XML W3C XML Binary Characterization Working Group Collected requirements W3C Efficient XML Interchange Working Group Performed measurements Made recommendations NCAR Preliminary Efficient XML Assessment Compared Fast Infoset vs normal XML representations Fast Infoset 75% of XML size Fast Infoset 33% of parsing time NCAR Efficient XML Assessment Measurements and recommendation Will likely use the EXI measurement framework WXXM 1 Efficient XML – Taking Net-Centric Operations to the Edge. John Schneider

5. 5 NextGen Network Enabled Weather (NNEW) Background and Motivation W3C Binary Characterization: Copyright © 2008-2009 University Corporation for Atmospheric Research Collected use cases Collected desirable features of different approaches Listed a minimum set of features for a recommended solution “Binary XML is needed” 2 2 http://www.w3.org/TR/exi-measurements/#conclusions http://www.w3.org/TR/exi-measurements/#conclusions 3 http://www.w3.org/TR/xbc-characterization/#N107D4 http://www.w3.org/TR/xbc-characterization/#N107D4 W3C Efficient XML Interchange: Created a measurement framework Collected a varied set of XML files for different use cases Made measurements of different solutions Recommended a solution Produced a specification (EXI) “the results indicate it is possible to achieve substantial gains over XML… in a wide variety of use cases” 3

6. 6 NextGen Network Enabled Weather (NNEW) Solutions Basic Approaches: Copyright © 2008-2009 University Corporation for Atmospheric Research Compression (GZIP, ZIP, TAR, etc) Worse processing efficiency, does not help with the characters -> value processing issue Must be decompressed before being operated upon (XSLT) XML-wrapped binary chunks XML for metadata or other selected portions Binary portions are “opaque”, no XML compatibility Efficient/binary XML formats Encode XML documents in a different form

7. 7 NextGen Network Enabled Weather (NNEW) Solutions Recommendation: Copyright © 2008-2009 University Corporation for Atmospheric Research EXI format Last Call Working Draft Likely to be the general industry standard for efficient XML transmission* Open source libraries (but still evolving) OpenEXI* Exificient … Advantages: XML flexibility XML tooling support (when developed) Increased efficiency Lossless conversion to/from XML Disadvantages: Not human readable (without conversion) Use SAX or StAX-based parsing techniques DOM has a non-trivial overhead in some cases (such as memory usage)

8. 8 NextGen Network Enabled Weather (NNEW) Exificient/GZIP experiment Copyright © 2008-2009 University Corporation for Atmospheric Research GZIP: 0.07 of original file size EXI: 0.13 of original file size

9. 9 NextGen Network Enabled Weather (NNEW) Fast Infoset to XML comparison Copyright © 2008-2009 University Corporation for Atmospheric Research File sizeParsing time AIRSIGMETs XML (5 records) 7kb18ms AIRSIGMETs FI3kb (0.43)13ms (0.72) METARs XML (1481 records) 1167kb84ms METARs FI373kb (0.32)56ms (0.667) PIREPs XML (158 records) 155kb29ms PIREPs FI51kb (0.33)29ms (1.0) TAFs XML (177 records)471kb57ms TAFs FI98kb (0.208)39ms (0.684)