Implementing an S3 native filesystem in Hadoop