The Optimization Research Of Java Card Performance

Java Card is now the accepted standard for smart card technology because its advantages like platform independence, multi-application and high security. However, as a kind of low-end embedded platform, Java Card has low execution speed for card application by the limitation of processor ability, memory capacity and bytecode interpretation, which had affected user experience even its prospect. In this paper, we hope to improve the Java Card performance during downloading,installing and executing applet through analyzing and improving the strategy of memory management as well as object management.In order to improve the execution speed of Java Card applets, the focus of our paper is to reduce the number of EEPROM writing caused by EEPROM writing mechanism, EEPROM allocating algorithm and transaction mechanism of Java Card. Firstly, a RAM-based Caching-Buffer are designed to improve the Page-Buffer mechanism, which is based on the spatial locality of persistent object. The Cache Strategy can reduce the number of writing to the same EEPROM page. Secondly, a mixed transaction model is designed for persistent object, which combines the advantages both Old-Value-Log and New-Value-Log transaction mechanism. Finally, a new transaction-based EEPROM Heap allocating algorithm is described when we analyzed the EEPROM Heap management strategy between downloading and installing applets. In new algorithm, EEPROM allocations take place in a reserved EEPROM segment which can avoid modifying EEPROM segment table frequently.A performance test is implemented for the optimization plans we design whose result proves that our research improves the performance of Java Card,though it causes some endurable negative effect. The Caching-Buffer mechanism and mixed-model transaction reduce the number of EEPROM writing during updating of persistent objects. Meanwhile, the transaction-based EEPROM Heap allocating algorithm reduces the number of EEPROM writing during modifying EEPROM segment table. At last, the result of this paper and some suggestions for further improvement of our plan as well as other fields of Java Card optimization are present.