Automatic stream detection and assignment algorithm
A Solid State Drive (SSD) is disclosed. The SSD may include flash memory to store data and may support a plurality of device streams. A SSD controller may manage reading and writing data to the flash memory, and may store a submission queue and a chunk-to-stream mapper. A flash translation layer may include a receiver to receive a write command, an LBA mapper to map an LBA to a chunk identifier (ID), stream selection logic to select a stream ID based on the chunk ID, a stream ID adder to add the stream ID to the write command, a queuer to place the chunk ID in the submission queue, and background logic to update the chunk-to-stream mapper after the chunk ID is removed from the submission queue.
1. A Solid State Drive (SSD), comprising:
flash memory to store data;
support for a plurality of device streams in the SSD;
a SSD controller to manage writing data to the flash memory responsive to a plurality of write commands, the SSD controller including storage for a submission queue and a chunk-to-stream mapper; and
a flash translation layer, including:
a receiver to receive a write command including a logical block address (LBA);
an LBA mapper to map the LBA to a chunk identifier (ID);
stream selection logic to select a stream ID based on the chunk ID using the chunk-to-stream mapper;
a stream ID adder to add the stream ID to the write command;
a queuer to place the chunk ID in the submission queue; and
background logic to remove the chunk ID from the submission queue and update the chunk-to-stream mapper.
2. A SSD according to claim 1 , wherein the chunk-to-stream mapper includes a Sequential, Frequency, Recency (SFR) table, the SFR table including the chunk ID and the stream ID for the chunk ID.
3. A SSD according to claim 2 , wherein the background logic includes sequentiality logic to select a previous stream if the LBA is sequential to a second LBA of a previous write command.
4. A SSD according to claim 2 , wherein the background logic includes:
recency logic to calculate a recency weight based on a current access time for the chunk ID, a previous access time for the chunk ID, and a decay period;
an access count adjuster to adjust an access count for the chunk ID based on the recency weight producing an adjusted access count; and
a stream ID adjuster to adjust the stream ID based on the adjusted access count for the chunk ID.
5. A SSD according to claim 1 , wherein the chunk-to-stream mapper includes a node entry, the node entry including the chunk ID and the stream ID for the chunk ID.
6. A SSD according to claim 5 , wherein the background logic includes:
promotion logic to determine when to promote the stream ID based on the chunk ID;
a second queuer to place the chunk ID in a first of a plurality of queues corresponding to a plurality of stream IDs, responsive to the stream ID for the chunk ID; and
demotion logic to determine when to demote the stream ID based on the chunk ID.
7. A SSD according to claim 6 , wherein:
the demotion logic includes:
a comparator to determine if an expiration time for the chunk ID has passed; and
if the expiration time for the chunk ID has passed, a decrementor to decrement the stream ID; and
the second queuer is operative to place the chunk ID in a second of the plurality of queues corresponding to the plurality of stream IDs, responsive to the decremented stream ID for the chunk ID.
8. A non-transitory storage medium, the non-transitory storage medium having stored thereon instructions that, when executed by a machine, result in:
a receiver to receive a write command for a Solid State Drive (SSD), the write command including a logical block address (LBA);
an LBA mapper to map the LBA to a chunk identifier (ID);
stream selection logic to select a stream ID based on the chunk ID using a chunk-to-stream mapper stored in a memory in a host computer system;
a stream ID adder to add the stream ID to the write command;
a queuer to place the chunk ID in a submission queue stored in the memory; and
background logic to remove the chunk ID from the submission queue and update the chunk-to-stream mapper.
9. The non-transitory storage medium according to claim 8 , wherein the chunk-to-stream mapper includes a Sequential, Frequency, Recency (SFR) table, the SFR table including the chunk ID and the stream ID for the chunk ID.
10. The non-transitory storage medium according to claim 9 , wherein the background logic includes sequentiality logic to select a previous stream if the LBA is sequential to a second LBA of a previous write command.
11. The non-transitory storage medium article according to claim 9 , wherein the background logic includes:
recency logic to calculate a recency weight based on a current access time for the chunk ID, a previous access time for the chunk ID, and a decay period;
an access count adjuster to adjust an access count for the chunk ID based on the recency weight producing an adjusted access count; and
a stream ID adjuster to adjust the stream ID based on the adjusted access count for the chunk ID.
12. The non-transitory storage medium according to claim 8 , wherein the chunk-to-stream mapper includes a node entry, the node entry including the chunk ID and the stream ID for the chunk ID.
13. The non-transitory storage medium according to claim 12 , wherein the background logic includes:
promotion logic to determine when to promote the stream ID based on the chunk ID;
a second queuer to place the chunk ID in a first of a plurality of queues corresponding to a plurality of stream IDs responsive to the stream ID for the chunk ID; and
demotion logic to determine when to demote the stream ID based on the chunk ID.
14. The non-transitory storage medium according to claim 13 , wherein:
the demotion logic includes:
a comparator to determine if an expiration time for the chunk ID has passed; and
if the expiration time for the chunk ID has passed, a decrementor to decrement the stream ID; and
the second queuer is operative to place the chunk ID in a second of the plurality of queues corresponding to the plurality of stream IDs responsive to the decremented stream ID for the chunk ID.
15. A method, comprising:
receiving a write command from a software source;
determining a logical block address (LBA) in the write command;
identifying a chunk identifier (ID) for a chunk on a Solid State Drive (SSD) including the LBA;
accessing a stream ID associated with the chunk ID;
assigning the stream ID to the write command;
processing the write command using the assigned stream ID on the SSD; and
performing a background update of the stream ID associated with the chunk ID.
16. A method according to claim 15 , further comprising:
determining whether the logical block address is sequential to a second LBA in a second write command; and
if the logical block address is sequential to the second LBA in the second write command:
determining the second stream ID assigned to the second write command; and
assigning the second stream ID to the write command.
17. A method according to claim 15 , wherein performing a background update of the stream ID associated with the chunk ID includes:
adding the chunk ID to a submission queue; and
removing the chunk ID from the submission queue when the chunk ID is at a head of the submission queue.
18. A method according to claim 15 , wherein performing a background update of the stream ID associated with the chunk ID includes:
increasing an access count for the chunk ID;
calculating a recency weight for the chunk ID responsive to a current access time and a previous access time for the chunk ID;
updating the access count for the chunk ID responsive to the recency weight; and
determining the stream ID for the chunk ID responsive to the updated access count.
19. A method according to claim 15 , wherein performing a background update of the stream ID associated with the chunk ID includes:
placing the chunk ID in a queue corresponding to the stream ID, where the queue corresponding to the stream ID is one a plurality of queues; and
determining whether to demote the chunk ID when the chunk ID reaches the head of the queue.
20. A method according to claim 19 , wherein:
performing a background update of the stream ID associated with the chunk ID further includes determining an expiration time for the chunk ID responsive to the access count and a device lifetime; and
determining whether to demote the chunk ID when the chunk ID reaches the head of the queue includes, if the expiration time for the chunk ID has passed:
removing the chunk ID from the queue corresponding to the stream ID;
decrementing the stream ID; and
placing the chunk ID in a second queue corresponding to the decremented stream ID.