In this paper, we explicitly study the online vertex cover problem, which is a natural generalization of the well-studied ski-rental problem. In the online vertex cover problem, we are required to maintain a monotone vertex cover in a graph whose vertices arrive online. When a vertex arrives, all its incident edges to previously arrived vertices are revealed to the algorithm. For bipartite graphs with the left vertices offline (i.e. all of the left vertices arrive first before any right vertex), there are algorithms achieving the optimal competitive ratio of $\frac{1}{1-1/e}\approx 1.582$. Our first result is a new optimal water-filling algorithm for this case. One major ingredient of our result is a new charging-based analysis, which can be generalized to attack the online fractional vertex cover problem in general graphs. The main contribution of this paper is a 1.901-competitive algorithm for this problem. When the underlying graph is bipartite, our fractional solution can be rounded to an integral solution. In other words, we can obtain a vertex cover with expected size at most 1.901 of the optimal vertex cover in bipartite graphs. The next major result is a primal-dual analysis of our algorithm for the online fractional vertex cover problem in general graphs, which implies the dual result of a 0.526-competitive algorithm for online fractional matching in general graphs. Notice that both problems admit a well-known 2-competitive greedy algorithm. Our result in this paper is the first successful attempt to beat the greedy algorithm for these two problems. On the hardness side, we show that no randomized online algorithm can achieve a competitive ratio better than 1.753 and 0.625 for the online fractional vertex cover problem and the online fractional matching problem respectively, even for bipartite graphs.