Shortest distance computation is one of the widely researched areas in theoretical computer science and graph databases. Distance labeling are well-known for improving the performance of shortest distance queries. One of the best distance labeling approaches is Pruned Landmark Labeling (PLL). PLL is a 2-hop distance labeling which prunes a lot of unnecessary labels while doing breadth-first-search. Another well-known 2-hop labeling is Pruned Highway Labeling (PHL) which is designed for undirected road networks. Both PLL and PHL suffer from the problem of large index size. In this paper, we propose two approaches to address the problem, one is to compress the PLL index as well as the graph for directed graphs; the other is to compress undirected road networks using linear sets, which are essentially maximal-length non-branching paths. Our aim is to reduce the index size and index construction time without significantly compromising query performance. Extensive experiments with real world datasets confirm the effectiveness of our approaches.