Abstract: A virtual machine template is created. The template includes a file system containing files to be deduplicated across multiple virtual machines. For each file to deduplicate, a hash of the content is generated and stored in association with the file. The content of the file is moved from the virtual machine template to a file store. The entry for the file in the store is indexed according to the hash. Multiple virtual machines are created by cloning the template, each containing a copy of its file system and the hashes stored locally in association with the corresponding deduplicated files. File access operations are monitored on each one of the multiple virtual machines, and attempts to access deduplicated file are detected. In response, the corresponding locally stored hash is used to retrieve the content of the file from the central file store, and provide it to the virtual machine.