A Web repository is a large special-purpose collection of Web pages and associated indexes. Many useful queries and computations over such repositories involve traversal and navigation of the Web graph. However, efficient traversal of huge Web graphs containing several hundred million vertices and a few billion edges is a challenging problem. An additional complication is the lack of a schema to describe the structure of Web graphs. As a result, naive graph representation schemes can significantly increase query execution time and limit the usefulness of Web repositories. In this paper, we propose a novel representation for Web graphs, called an S-Node representation. We demonstrate that S-Node representations are highly space-efficient, enabling in-memory processing of very large Web graphs. In addition, we present detailed experiments that show that S-Node representations can significantly reduce query execution times when compared with other schemes for representing Web graphs.