The Design And Implementation Of Multi-target Tracker For The Cluster Storage System

With the development of computer technology, applications and kernel modulesused in a variety of fields are being gradually developed, bringing great convenienceto people’s life and work. In order to help programmers and testers to understand therunning status of the product, and product data information for each event occurredduring run, so tracking tools produced. The tracker is not only a performance tuningtools, and but also a record of tools for analysis application load behavior. Throughcollection and analysis of data, we can get a lot of information to track the target,which can help the system designers to improve the overall design to achievesolving a problem or to optimizing the system’s goals.The subject is the test portion of the block device cache project which comes fromthe Tianjin zhongke blue whale, which is design as a tracking tool to performanceanalysis module for each existing module and working condition.The main content of this paper is focused on needs analysis and design ofclustered storage system tracker. Different from other past tracking test tools,detection target of the tracker is IO situation. The tracker is based on the tracepointmechanism which inserted the tracker external interfaces into the IO critical path oftracked objects, and the interface can collect the IO information of tracking object,which`s passed to the kernel module that is stored in the cache. The tracker consistsof two parts, the kernel core part and the user layer part. The user layel part includescommand line parsing module and tracking information analysis module, and thekernel core part includes the I/O contrl module, cache management module andexport interface.With the tracker, testers can quickly and accurately identify the bug’s cause. Inorder to provide data for clustered storage, the tracker need identify the I/Oprocessing time in the core module and the underlying device, and obtin the I/Oprocessing time on the IO path data information related performance, etc., so theseinformation can offer help to analysising and solving the problem.The test results show that the tracer can complete the scheduled tracking task andcollect enough data for the analysis work, and the tracking performance meets therequirements. Tracker will have little impact on the cluster storage system, and canmeet the requirements of the test.