Here - Agents Lab - University of Nottingham

Java serialization and Jadex Binary still show an advantage in compactness withJadex Binary maintaining a slim margin over Java serialization.60005000509840003609Time (ms)300020001831200610008732010Jadex XMLJadex XML(compressed)Jadex BinaryJadex Binary(compressed)JavaSerializationJavaSerialization(compressed)Fig. 12. An additional compression pass increases total encoding timeSince the compression pass substantially reduces the size of messages, especiallyfor verbose formats, it may suggest that starting out with a compactformat gives only a marginal advantage. However, data compression is not free interms of computation time. While compression helps compactness, this issue hasto be weighed against the performance message feature. Despite the DEFLATEalgorithm being a comparably fast compression algorithm, Figure 12 shows thatit adds a substantial amount to the total encoding time of the content. In fact,the additional time required seems to grow with the number of bytes in theuncompressed content, which is reasonable considering the algorithm must evaluateevery byte of the uncompressed data at least once to produce a reversibleoutput.As a result, data compression does not appear to be generally beneficial whenboth performance and compactness are important; however, it is another usefultool to adjust the balance between the two language features. In the next sectionwe will discuss further improvements, future work and provide a conclusion onthe performance of Jadex Binary.6 Future Work and ConclusionThe evaluation of Jadex Binary in Section 5 appears to provide sufficient evidencethat Jadex Binary already has significant advantages in both compactness andperformance. However, the performance results of the Java serialization showsthat further performance improvements may be possible. One way of further114