tag:blogger.com,1999:blog-12168966.post8804336130630195524..comments2011-12-13T17:59:06.390+08:00Mohanraj C Periyasamyhttp://www.blogger.com/profile/13625981004007296580noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-12168966.post-71532407387042479432011-12-13T17:59:06.390+08:002011-12-13T17:59:06.390+08:00Hi Nick,<br /><br />I am not sure. But my guess would be that there is no clustering of hardware servers here as they follow the shared nothing principle. But the data is stored across multiple clusters and even replicated across multiple clusters for failover.<br /><br />May be there&#39;s a primary node which knows how to apply the reduce-map operations based on the requests coming in.Mohanraj C Periyasamyhttp://www.blogger.com/profile/13625981004007296580noreply@blogger.comtag:blogger.com,1999:blog-12168966.post-32589312618905957522011-12-13T06:01:48.061+08:002011-12-13T06:01:48.061+08:00Nice article. How is the clustering between servers organized? Does it have special principle?Nicknoreply@blogger.com