Abstract:

Objectives: Mimic voter is a crucial component under the dynamic heterogeneous redundancy architecture of mimic defense technology,but the existing mimic voting method needs to collect and process the complete output data of heterogeneous redundant executives. In the application scenario where the connection continuously transmits data in chunked transfer encoding,there are problems that the mimic voting efficiency is too low and the memory resource overhead of mimic voting is too significant.This paper designs and implements an adaptive mimic voter oriented to the scenario of the continuous output of chunked transfer encoded data in a persistent connection to reduce the memory resource overhead of the mimic voter and improve voting efficiency.

Methods: The proposed mimic voter adaptively divides the chunked-transfer-encoded data arriving at the voter successively from the heterogeneous redundant executives,dynamically votes, and then outputs the data in the form of a sliding window during the continuous data transmission process.Gradually releasing the data of the voted blocks can reduce the memory consumption of the mimic voter and lower the voting processing time while maintaining the continuity of data transmission of the persistent connection.On the one hand,a voting algorithm selection strategy set is constructed to keep the voting accuracy by analyzing the data characteristics in the sliding window.On the other hand,an inventory model of the data voting process of the adaptive mimic voter is established,and an adaptive voting window control strategy is proposed based on the cost optimization of the inventory model to provide the best adaptive segmentation scheme for the data to be voted.

Results:A series of comparative experiments between the prototype system of the adaptive mimic voter and the traditional mimic voter is conducted as follows. (1) The evaluation of memory resource occupancy shows that the peak physical memory consumption and the total time of consuming physical memory when the adaptive mimic voter transmits 20MB web resources in chunked transfer encoding are significantly lower than those of the traditional mimic voter. (2) The evaluation of transmission delay shows that the response time of the adaptive mimic voter in the voting processing of 10MB to 320MB chunked transfer-encoded webpage resources is relatively low, indicating that its voting speed has been significantly improved. (3) The concurrency performance evaluation shows that the average of response time of the system applying the adaptive mimic voter under the request concurrency of 1000 to 5000 is lower than that of the traditional mimic voter,and the voting processing throughput is higher than that of the traditional mimic voter.(4)The evaluation of voting accuracy shows that the adaptive mimic voter based on the voting algorithm selects the strategy set is slightly weaker than the semantic feature algorithm and the AHP algorithm while far superior to the character similarity algorithm in the traditional mimic voter, which reveals that the adaptive mimic voter has an acceptable voting accuracy.

Conclusions: The design of the adaptive mimic voter effectively solves the problem of service performance degradation caused by the excessive occupation of memory resources in voting chunked transfer encoding data of persistent connection. The memory occupancy experiment shows the improvement effect of the adaptive mimic voter on this problem, and the voting accuracy evaluation experiment shows that the adaptive mimic voter can improve voting efficiency while maintaining acceptable voting accuracy.The experiments under different service pressures give the feasibility analysis of the adaptive mimic voter in general application scenarios with micro-benchmarks. Therefore, the adaptive mimic voter reduces resource overhead and improves voting efficiency with acceptable voting accuracy, which can effectively support the mimic transformation of applications that transmit data in persistent connections.

Key words: mimic defense, mimic voting, HTTP, persistent connection, chunked transfer encoding