The discussion process plays an important social task in Computer-Supported Collaborative Learning (CSCL) where participants can discuss about the activity being performed, collaborate with each other through the exchange of ideas that may arise, propose new resolution mechanisms, as well as justify and refine their own contributions and thus acquire new knowledge. Indeed, learning by discussion when applied to collaborative learning scenarios can provide significant benefits for students in collaborative learning, and in education in general. However, the discussion process in the context of distance education presents high dropout in comparison to traditional programs due chiefly to a sense of isolation of participants who do not have knowledge about others nor they can compare their own progress and performance to the group. To alleviate this problem, the provision of appropriate knowledge from the analysis of on-line interaction is rapidly gaining popularity due to its great impact on the discussion performance and outcomes. This implies a need to capture and structure all types of information generated by group activity and then to extract the relevant knowledge in order to provide participants with efficient awareness and feedback as regards group performance and collaboration. As a result, it is necessary to process and analyze complex event log files from group activity in a constant manner, and thus it may require computational capacity beyond that of a single computer. To this end, in this paper we show how a Grid approach can considerably increase the overall efficiency of processing group activity log files and thus allow discussion participants to receive effective knowledge even in real time.