| 摘要 |
A system and apparatus intermediates between a centralized remote file service and several distributed virtual machine data center servers. When several file blocks are committed by applications running in virtual machines, they are aggregated together into a fewer number of file write operations to exploit the parallelism of multi-disk arrays. An application within a virtual machine may be released without risk of data loss as soon as the commitment has been acknowledged. Requests for frequently retrieved file blocks are redirected to the location of the freshest retention. Each data center has a non-transitory file block store locally attached which provides images of virtual machines, applications, and data. Each file block librarian provides a file system view to each virtual machine. Several Virtual Machine Data Centers are coordinated by a tracker which locates, redirects, and retrieves any file block within its name space. |
