| In the context of the “Internet+”, the Internet continues to develop traditional industries. A sharp increase in data on the Internet. These data also includes image files and other multimedia files. Image files will be given a lot of new properties. The traditional image storage systems will only consider the size of image files,while ignoring the new property to the people entrusted to them.After the user's access habits and behavior analysis, we found the same or similar properties pictures tend to be accessed together. Users for the purpose of comparison the same or similar attributes of the image, thereby to make a choice. Because HDFS obvious disadvantage in small files stored,so when storing small files, usually small files into one large file storage. By this analysis, the same properties combined in one large image file can effectively save access time.According to the characteristics of multi-attribute files and storage characteristics of HDFS,designed and implemented multi-attribute image storage system based on HDFS. First, we need to set the properties of the picture file, and assigned to each property. Then generate directory for these attributes and attribute values. Then upload the same or similar properties pictures stored in the same directory, and wait for the file and upload it to the right size combined HDFS. So they put a image of the same or similar attributes stored in the same block.Users accessing the same or similar properties pictures, these pictures are stored in the same block, so long as a block to open the file, and then file a similar picture in the block according to the file offset accessed files, thus saving interview time.Test results showed that the access time to the same or similar properties on the same image stored in a data node will be significantly lower than these pictures discrete data stored on different nodes. According to the design of test cases and test the entire system functions can be implemented properly. |
PDF Full Text Request