Thrashing reduction in demand accessing of a data base through an LRU paging buffer pool

Claim 2

A CPU implementable method for minimizing thrashing among concurrent processes demand page accessing of a data base through an LRU page organized buffer pool, characterized in that the steps include: allocating to each new process an independent LRU page frame stack of buffer elements; ascertaining the set of pages over which there is looping access behavior by each process; scheduling for execution only those processes for which there exists a frame stack large enough to hold the counterpart set of pages; and upon the occurrence of a page fault, exercising an LRU page replacement discipline upon the counterpart stack.

Claim 3

A machine implementable method for minimizing thrashing among concurrent processes demand page accessing of a relational data base through an LRU page organized buffer pool in which some pages may be pinned, characterized in that the method steps include: allocating to each new process an independent LRU page frame stack of buffer elements; ascertaining the set of pages over which there is looping access behavior; scheduling for execution only those processes for which there exists a frame buffer large enough to hold the counterpart set of pages; and upon the occurrence of page fault, either executing an LRU page replacement on the stack counterpart to the referencing process if no pages are pinned, or assigning a page frame from any other stack formed from unpinned pages and whose actual frame occupancy is less than the number of pages ascertained for the process associated with the stack.

Claim 4

A machine implementable method for minimizing thrashing among concurrent processes demand page accessing of a relational data base through an LRU page organized buffer pool in which some pages may be pinned, characterized in that the method steps include: allocating to each new process an independent LRU page frame stack of buffer elements; ascertaining the set of pages over which there is looping access behavior; scheduling for execution only those processes for which there exists a frame buffer large enough to hold the counterpart set of pages; and upon the occurrence of page fault, either executing an LRU page replacement on the stack counterpart to the referencing process if no pages are pinned, or assigning a page frame from any other stack formed from unpinned pages and whose actual frame occupancy is less than the number of pages ascertained for the process associated with the stack.

Claim 5

A method according to claim 2, characterized in that subsequent to the scheduling of the processes, the method comprises a further step of searching for a page in the buffer pool upon processor request and updating an associated usage counter if the page is found in the local LRU stack.

Claim 6

A method according to claim 3, characterized in that the step of updating the stack includes the step of either increasing the reference count by a constant value if the requesting process pins a page, decreasing the reference count by a constant value if the requesting process unpins a page, or avoiding updating activity in the event a process releases a page.

Claim 7

A method according to claim 2, characterized in that the method further comprises the step of either reallocating the LRU stack of page frames among deficient processes in the event a given process terminates or in the absence of deficient processes returning the frame to a common list.

Claim 8

A method according to claim 2, characterized in that the sum total of the set sizes of the requesting processes being serviced is less than or equal to the size of the buffer pool.

Claim 9

A CPU implementable method for reducing the number of page frames stolen by a first process having a paging rereferencing interval nested within the rereferencing interval of a second process, both processes demand page accessing of a data base through an LRU page organized buffer pool, characterized in that the method includes the steps of: allocating to each process an arbitrarily large LRU page frame stack (numref); ascertaining the set of pages over which there is looping accessing behavior; scheduling the execution of only those processes for which there exists a buffer large enough to hold the counterpart set of pages; and upon the occurrence of a page fault, exercising an LRU page replacement discipline upon the counterpart stack, limiting the maximum number of pages that the process maintains pinned simultaneously to a value less than the arbitrarily large LRU page frame stack, whereby a process may steal pages only from itself and not from other processes.

Claim 10

A CPU implementable method for reducing the number of page frames stolen by a first process having a paging rereferencing interval nested within the rereferencing interval of a second process, both processes demand page accessing of a data base through an LRU page organized buffer pool, characterized in that the method includes the steps of: allocating to each process an arbitrarily large LRU page frame stack (numref); ascertaining the set of pages over which there is looping accessing behavior; scheduling the execution of only those processes for which there exists a buffer large enough to hold the counterpart set of pages; and upon the occurrence of a page fault, exercising an LRU page replacement discipline upon the counterpart stack, limiting the maximum number of pages that the process maintains pinned simultaneously to a value less than the arbitrarily large LRU page frame stack, whereby a process may steal pages only from itself and not from other processes.

Claim 11

A method according to claim 7, characterized in that the LRU page replacement discipline step includes the step of either executing an LRU page replacement on the counterpart stack if no pages are pinned, or assigning a page frame from a stack containing only unpinned pages and whose actual frame occupancy is less than the number of frames requested by the process.

Claim 12

A method according to claim 7 characterized in that the upper bound of operability is constrained to the condition when all the pages and page frames in the buffer are pinned.