Integrating a Flexible File Abstraction into the Linux Kernel

Performance of data systems are critical in today’s world due to the large amount of data collection being done. Database management systems (DBMS) employ structured files and large amounts of high cost in-place insertions and removals. Traditionally, applications use extra layers of indirection to offset these high cost operations but this only adds more complexity and higher access costs. In the face of this, we utilize a flexible address space located in user space called FlexFile. FlexFile utilizes in-place updates for arbitrary-sized data, which creates a more efficient way for insertions and removal operations to be performed. This research re-implements FlexFile in the Linux kernel and evaluates performance comparisons between both. First, the implementation overview is presented to efficiently realize the user space needs of a traditional file system interface. Next, the current and future implementation possibilities are discussed to perfect and expand this development. Last, both implementation results are compared and contrasted to show how specific locations within a system can be leveraged to produce different results within the FlexFile implementation.