Method for managing a buffer pool referenced by batch and interactive processes

Claim 1

A method for operating a computing apparatus including a main processor, a plurality of main storage buffers, and at least one external storage device; the main processor being operated under control of a data base system which includes a buffer manager for allocating data set pages read from the storage device into the buffers for concurrent access by both sequential and random processes; the method being characterized by the steps of: (1) establishing a sequential steal threshold value of a queue as the product of the number of levels of out-of-line nesting most often expected per sequential processing instance times the maximum number of concurrent sequential processing instances most often expected; (2) maintaining a normal least recently used (LRU) buffer chain and a sequential LRU buffer chain with sequential buffers included in both chains and the random buffer included only in the normal LRU buffer chain; (3) placing buffers released from a sequential process that is not a prefetch process in both LRU buffer chains, and placing buffers released from a random process or from a sequential prefetch process only in the normal LRU buffer chain; and (4) responsive to a sequential access request for a buffer, stealing the head buffer from the sequential LRU buffer chain provided the number of buffers in that chain exceeds the queue value, otherwise, stealing the head buffer from the normal LRU chain, because the number of buffers on the sequential LRU chain is not greater than the queue value or the request is for access by a random process, thereby preventing recursive steals of buffers employed in nested sequential processes.

Claim 2

A method for operating a computing apparatus including a main processor, a plurality of main storage buffers, and at least one external storage device; the main processor being operated under control of a data base system which includes a buffer manager for allocating data set pages read from the storage device into the buffers for concurrent access by both sequential and random processes; the method being characterized by the steps of: (1) establishing a sequential steal threshold value of a queue as the product of the number of levels of out-of-line nesting most often expected per sequential processing instance times the maximum number of concurrent sequential processing instances most often expected; (2) maintaining a normal least recently used (LRU) buffer chain and a sequential LRU buffer chain with sequential buffers included in both chains and the random buffer included only in the normal LRU buffer chain; (3) placing buffers released from a sequential process that is not a prefetch process in both LRU buffer chains, and placing buffers released from a random process or from a sequential prefetch process only in the normal LRU buffer chain; and (4) responsive to a sequential access request for a buffer, stealing the head buffer from the sequential LRU buffer chain provided the number of buffers in that chain exceeds the queue value, otherwise, stealing the head buffer from the normal LRU chain, because the number of buffers on the sequential LRU chain is not greater than the queue value or the request is for access by a random process, thereby preventing recursive steals of buffers employed in nested sequential processes.

Claim 4

A method for preventing monopolization of a buffer pool by sequential processes sharing the buffer pool concurrently with random processes, comprising the steps of: establishing a sequential steal threshold value of a queue as the product of the number of levels of out-of-line nesting most often expected per sequential processing instance times the maximum number of concurrent sequential processing instances most often expected; maintaining a normal least recently used (LRU) buffer chain and a sequential LRU buffer chain with sequential buffers included in both chains and the random buffer included only in the normal LRU buffer chain; placing buffers released from a sequential process that is not a prefetch process in both LRU buffer chains, and placing buffers released from a random process or from a sequential prefetch process only in the normal LRU buffer chain; and responsive to a request for a buffer for a sequential process, stealing the head buffer from the sequential LRU buffer chain provided the number of buffers in that chain exceeds the queue value, otherwise, stealing the head buffer from the normal LRU buffer chain because the number of buffers on the sequential LRU chain is not greater than the queue value or the request is for access by a random process.

Claim 5

A method for preventing monopolization of a buffer pool by sequential processes sharing the buffer pool concurrently with random processes, comprising the steps of: establishing a sequential steal threshold value of a queue as the product of the number of levels of out-of-line nesting most often expected per sequential processing instance times the maximum number of concurrent sequential processing instances most often expected; maintaining a normal least recently used (LRU) buffer chain and a sequential LRU buffer chain with sequential buffers included in both chains and the random buffer included only in the normal LRU buffer chain; placing buffers released from a sequential process that is not a prefetch process in both LRU buffer chains, and placing buffers released from a random process or from a sequential prefetch process only in the normal LRU buffer chain; and responsive to a request for a buffer for a sequential process, stealing the head buffer from the sequential LRU buffer chain provided the number of buffers in that chain exceeds the queue value, otherwise, stealing the head buffer from the normal LRU buffer chain because the number of buffers on the sequential LRU chain is not greater than the queue value or the request is for access by a random process.