How is data locality utilized for filesystems other than HDFS in Hadoop?
As you mentioned, filesystems like S3 do not need the namenonde to run. In this case the data locality optimization is not available.
Reference: http://wiki.apache.org/hadoop/AmazonS3