A large number of online health communities exist today, helping millions of people with social support during difficult phases of their lives when they suffer from serious diseases. Interactions between members in these communities contain discussions on practical problems faced by people during their illness such as depression, side-effects of medications, etc and answers to those problems provided by other members. Analyzing these interactions can be helpful in getting crucial information about the community such as dominant health issues, identifying sentimental effects of interactions on individual members and identifying influential members. In this paper, we analyze user messages of an online cancer support community, Cancer Survivors Network (CSN), to identity the two types of social support present in them: emotional support and informational support. We model the task as a binary classification problem. We use several generic and novel domain-specific features. Experimental results show that we achieve high classification performance. We, then, use the classifier to predict the type of support in CSN messages and analyze the posting behaviors of regular members and influential members in CSN in terms of the type of support they provide in their messages. We find that influential members generally provide more emotional support as compared to regular members in CSN.