摘要
The overhead in iSCSI subsystems is analyzed through the model of iSCSI reading and writing requests. An analytic model of iSCSI subsystem overhead is developed. According to the analytic model, the overhead of iSCSI subsystems is measured, which reveals that the main overhead is caused by protocol processing as well as kernel functions for fair allocation of system resources. Several methods have been proposed to optimize iSCSI subsystems, such as small I/O requests adherent that can be merged into a large I/O request. Checksum is found to be a time consuming work and may not be always necessary for applications.
The overhead in iSCSI subsystems is analyzed through the model of iSCSI reading and writing requests. An analytic model of iSCSI subsystem overhead is developed. According to the analytic model, the overhead of iSCSI subsystems is measured, which reveals that the main overhead is caused by protocol processing as well as kernel functions for fair allocation of system resources. Several methods have been proposed to optimize iSCSI subsystems, such as small I/O requests adherent that can be merged into a large I/O request. Checksum is found to be a time consuming work and may not be always necessary for applications.
基金
Funded by National Network and Security Continuous Developing Plan (2004 Research 1-917-021)
