Implement the partition operator based on s3qs in spike cloud queries, and use s3qs for data shuffling to achieve better scalability.