230 likes | 464 Views
An Evaluation of Contemporary Commercial SOAP Implementations. Alex Ng Department of Computing, Macquarie University. Shiping Chen CSIRO ICT Centre. Paul Greenfield CSIRO ICT Centre. Presented by: Alex Ng. AWSA 2004, Melbourne, 13-14 April 2004. Outline. Background
E N D
An Evaluation of Contemporary Commercial SOAP Implementations Alex Ng Department of Computing, Macquarie University Shiping Chen CSIRO ICT Centre Paul Greenfield CSIRO ICT Centre Presented by: Alex Ng AWSA 2004, Melbourne, 13-14 April 2004
Outline • Background • Experimental design • Result discussion • Related work • Future work
Web Services Definition (W3C) A Web service is a software system designed to support interoperable machine-to-machineinteraction over a network. It has an interface described in a machine-processable format (specifically WSDL). Other systems interact with the Web service in a manner prescribed by its description using SOAP-messages, typically conveyed using HTTP with an XML serialisation in conjunction with other Web-related standards. W3C Web Services Architecture Working Draft (8 August 2003)
W3C Web Services Conceptual Stack Source: Ferguson D. Secure, Reliable, Transacted Web Services: Architecture and Composition. MSDN technical article, Sept. 2003.
SOAP (Simple Object Access Protocol) Request POST /StockQuote HTTP/1.1 Host: www.stockquoteserver.com Content-Type: text/xml Content-Length: nnnn SOAPAction: “Some-URI” <SOAP:Envelope xmlns:SOAP="urn:schemas.xmlsoap.org:soap.v1"> <SOAP:Header> <t:Transaction xmlns:t=”URI” mustUnderstand=”1”>5</t:Transaction> </SOAP:Header> <SOAP:Body> <m:GetLastTradePrice xmlns:m="URI"> <symbol>DIS</symbol> </m:GetLastTradePrice> </SOAP:Body> </SOAP:Envelope> • Simple • Extensible • Based on XML • De facto standard for XML Messaging in supporting Web Services • V1.1(May/00) • V1.2 (June/03)
Other Competitors of SOAP Web Services • Use different transport mechanism • Microsoft DIME (Direct Internet Message Encapsulation) • Work with or without SOAP & HTTP • send attachments of various types • BEEP (Blocks Extensible Exchange Protocol) • Connection oriented, min 257 channels, asynchronous message exchange • Work with or without HTTP • Use different Web Architecture • REST (Representational State Transfer) • An architectural style (Roy Fielding) • Use XML/HTML, no need for SOAP
Performance Issues of SOAP Web Services • Serialisation and de-serialisation • Message size ( 6-8 times larger than binary) • Introduce latency in transmission, storing or retrieving XML data • XML Processing • Parsing (encode/decode XML elements) • Validation (conformation to a pre-defined schema) • Transformation (XML <-> other format)
Network settings Issues • HTTP 1.1 chunking (avoids content-length calculation) • HTTP 1.1 Keep-Alive (reduces connection set up cost) • HTTP 1.1 Pipelining (handles multiple requests/connection) • TCP-Delayed-ACK (defers receiver acknowledgement) • Nagle Algorithm (defers the sending of small packets)
Evaluation of contemporary commercial SOAP implementations (1) What level of performance is shown by major commercial SOAP Web Services implementations? (2) How does the performance of SOAP compare to that offered by conventional non-SOAP technologies? Document SOAP Body RPC Literal SOAP Parameter Encoded
Implementation Doc/ Lit Doc/ Enc RPC/ Lit RPC/ Enc Product A Product B Simple Message Single customer inquiry/update Product C Medium Message 20 customers Inquiry Complex Message Invoice: One Customer, 50 product lines Experimental Design Property.xml 4x1.6Ghz, 3.6Gb Test Driver Result Log Platform Specific Runner 100Mbps Ethernet Target Application Server Target Web Service Test Document 4x1.6Ghz, 3.6Gb
Encoding Style Simple Message Medium Message Complex Message Product A Doc/Lit 873 7581 21392 Doc/Enc 1593 13973 38842 RPC/Enc 1542 13969 38838 Product B Doc/ Lit 870 7513 21566 RPC/Enc 1597 14035 38999 Product C RPC/Enc 1664 14566 40273 Result: Message Size No.of Char.
Result: Serialisation/De-serialisation 13% 23% 10% 30%
Evaluation Summary • SOAP Performance is affected by: • Encoding style (Document/Literal encoding is the choice) • Product of implementation (a sizable gap between two products using the same document/literal encoding) • Performance delivered by the majority SOAP implementations: • Simple Message – able to deliver reasonably good performance(4 to 6ms vs 1ms non-SOAP) • Medium Message – (10 to 22ms vs 7ms non-SOAP) • Complex Message – (20 to 63ms vs 15ms non-SOAP) • De-serialisation overhead is higher than that of serialisation.
Related Work (1) • The Extreme Lab, Indiana University • Govindaraju, et. al. • sending different size of linked list of doubles using different implementations: SOAPRMI, SUNRMI, and NEXUSRMI, and TCP. • TCP is two orders of magnitude better than SOAPRMI • Serialisation and de-serialisation are the bottlenecks. • Chiu, et.al. • sending different arrays of mathematical doubles. • Suggested Optimization: Pull Parser, schema-specific XML parser, trie and HTTP 1.1. • Kohlhoff and Steele: Compares SOAP against FIX and CDR • SOAP latency is 2 to 3 times worse than CDR • FIX is a text based encoding and outperforms CDR and CDR
Related Work (2) • Dhawan: Compares the performance of .NET Remoting against ASP.NET • ASP.NET XML serialisation is more efficient than .NET Remoting SOAP serialisation • Davis and Parashar • Different SOAP Implementations: SOAPRMI/Java1.1, Microsoft SOAP Toolkit SP2 in VB, Perl SOAP::Lite, Apache SOAP 2.2, and Apache Axis Alpha 3. • inappropriate TCP options had caused 200ms delays in some implementations (MS SOAP, Apache SOAP and SOAP:Lite ) • Conclusion: SOAP is in the orders of magnitude slower than JavaRMI and CORBA
Related Work (3) • Benkner, et. al.: Compares the performance of JAXM implementations against JAXRPC in different package sizes (1x1024 -> 8x128) • Different Implementations: Apache Axis 1.1 RC2, Sun JWSDP 1.1, Mind Electric GLUE4.0, Systinet WASP 4.5, Novell jBroker Web 2.0, and X-Treme XSOAP 1.2.20. • Conclusion: • JAXM & JAXRPC produce similar results • Best performance result is achieved when using large packages of data, i.e. splitting 1MB of source data into four packages of 256KB each for JAXM in Systinet WASP implementation and 2x 512 for JAXRPC in Apache implementation.
Future Work • Extend Evaluation to other SOAP Implementations, such as Apache Axis • Identify other factors when SOAP over HTTP is used across wide-area networks
Reference • Chiu K., et. al. (2002) Investigating the Limits of SOAP Performance for Scientific Computing In Proceedings of 11th. IEEE International Symposium on High Performance Distributed Computing HPDC-11 2002 (HPDC'02) • Benkner, S., Brandic, I., Dimitrov, A., et al. Performance of Java Web Services Implementations. In Proceedings of ICWS'03. Las Vegas, 2003 • Davis, D. and Parashar, M. Latency Performance of SOAP Implementations. In Proceedings of IEEE Cluster Computing and the GRID 2002 (CCGRID'02) • Dhawan, P. Performance Comparison: Exposing Existing Code as a Web Service, October 2001 (on-line) Accessed Date: 10 October 2002 http://msdn.microsoft.com/library/default.asp?url=/library/en-us/dnbda/html/bdadotnetarch11.asp • Kohlhoff, C. and Steele, R. (2003) Evaluating SOAP for High Performance Business Applications: Real–Time Trading Systems. In Proceedings of WWW2003. Budapest, Hungary. • Govindaraju, M., A. Slominski, et al. Requirements for and Evaluation of RMI Protocols for Scientific Computing. Supercomputing. IEEE, 2000