A controller receives a write request. If it generates a hash that is already recorded in the hash table, what happens next?

The Purity//FA operating environment uses an inline, five-stage data reduction process. When a write request is received, Purity generates a high-fidelity hash for the variable-sized data block. This hash is checked against an in-memory hash table. If a matching hash is found, it indicates a potential duplicate. To guarantee data integrity and prevent data loss from the rare possibility of a hash collision (where two different blocks produce the same hash), Purity performs a definitive byte-for-byte comparison between the new block and the existing block already stored on flash. Only after this comparison confirms the blocks are identical is the new block discarded and the metadata updated.

A. The next incoming block is then hashed to see if it can be deduplicated.

This is incorrect. The system must complete the verification process for the current block before proceeding to the next one in the I/O stream.

B. Deep level compression is then applied to the newly hashed block.

This is incorrect. Compression is applied to unique data blocks after the deduplication check and before they are written to flash, not to a block that is a potential duplicate.

D. Purity//FA will expand the block to see if it can deduplicate a larger dataset.

This is incorrect. "Expanding the block" is not a function of the Purity//FA deduplication engine; the system works to reduce, not expand, data.

1. Pure Storage FlashArray Architecture White Paper. In the section describing the Purity//FA write path and data reduction, the process is detailed. It explicitly states that after a hash match, a direct comparison of the data blocks is performed to verify a true duplicate before discarding the incoming data. This step is crucial for ensuring data integrity. (Typically found in sections titled "Purity Data Reduction" or "Write I/O Path").

2. Pure Storage Technical Report, "An Inside Look at the Purity FlashArray Operating Environment". This document outlines the stages of data reduction. Stage 2 is "High-performance inline deduplication." The description of this stage clarifies that a hash hit triggers a read of the original data block for a bit-for-bit comparison to confirm the duplicate, thereby preventing hash collisions from causing data corruption. (Refer to the section on "The Five Stages of Data Reduction").