dfs.replication is by default 3 and you can configure it in hdfs-site.xml file (as you have seen it). This is because in Hadoop we use commodity hardware which are more failure prone. This is the reason why we maintain multiple copies of data in HDFS. If a datanode fails, yet you can recover the data from other data nodes. So, for backup we have replica of data blocks stored in HDFS.
For more information you can refer this HDFS blog: