Adaptive read prefetch to reduce host latency and increase bandwidth for sequential read streams

A system for prefetching data for a sequential read stream, comprising:

• a main memory;
• and
• processing circuitry configured to execute program instructions out of the main memory to:
• perform prefetch operations for a sequential read stream in accordance with a prefetch distance and a prefetch size;
• keep track of occurrences of failure scenarios in the prefetch operations while servicing the sequential read stream;
• dynamically adjust one of the prefetch distance and the prefetch size based on the occurrences of failure scenarios in the prefetch operations;
• receive a plurality of read requests for data stored on a storage object;
• determine that the plurality of read requests are part of the sequential read stream;
• and
• determine an initial value of the prefetch size based on an average input/output (IO) size of the plurality of read requests,
• wherein the initial value of the prefetch size is greater than a predetermined minimum size value.

A computer program product including a set of non-transitory, computer-readable media having instructions that, when executed by processing circuitry of a computerized apparatus, cause the processing circuitry to perform a method of prefetching data for a sequential read stream, the method comprising:

• performing prefetch operations for a sequential read stream in accordance with a prefetch distance and a prefetch size;
• keeping track of occurrences of failure scenarios in the prefetch operations while servicing the sequential read stream;
• dynamically adjusting one of the prefetch distance and the prefetch size based on the occurrences of failure scenarios in the prefetch operations,
• wherein performing prefetch operations for a sequential read stream includes storing prefetched data elements in a cache memory, and
• wherein each prefetched data element stored in the cache memory has a plurality of associated flags including a PREFETCHED flag and a REPROMOTED flag;
• and
• for each prefetched data element stored in the cache memory, setting its associated PREFETCHED flag, clearing its associated REPROMOTED flag, and placing an entry for the prefetched data element in a least-recently-used (LRU) queue.