A Novel Cost Model of XML Serialization

详细信息    查看全文

• 作者：G. Imre ; M. Kaszó ; T. Levendovszky ; H. Charaf
• 关键词：Software performance ; XML serialization ; performance modeling
• 刊名：Electronic Notes in Theoretical Computer Science
• 出版年：2010
• 出版时间：22 February 2010
• 年：2010
• 卷：261
• 期：Complete
• 页码：147-162
• 全文大小：938 K

文摘

Using XML as a data representation format is a common choice when integrating software systems on different platforms. The serialization of in-memory object instances of a class into corresponding XML documents heavily influences the performance of the XML-based communication, even if we send the XML over HTTP as in the case of SOAP-based XML Web Services, or with asynchronous messaging such as Java Message Service (JMS), or simply saving it into a file. Several studies have been published analyzing the performance impact of XML serialization on different platforms. No models or measurement methodologies have been proposed however, to establish a relationship between the serialization cost of primitive types (e.g. int, double, string), and the serialization cost of composite types. Such a model can be very useful when the type of the XML messages exchanged during the communication are known a priori, recorded in an interface definition, similarly to the Web Services Description Language (WSDL) in case of XML Web Services. This paper introduces a model that is validated with measurements on .NET and Java platform. The opposite direction, deserialization is covered as well. The main mathematical tool used is linear regression, but cases are also shown and explained where linearity is compromised.