Controller and method for virtual LUN assignment for improved memory bank mapping
A controller communicates with a plurality of multi-chip memory packages. Each multi-chip memory package comprises a plurality of memory dies, each having a respective plurality of memory blocks, some of which are good and some of which are bad. The controller determines a number of good blocks in each memory die. Based on the determined number of good blocks in each memory die, the controller selects a memory die from each of the multi-chip memory packages to access in parallel, wherein the selected memory dies are not necessarily all in the same relative position in each multi-chip package. The controller then creates a metablock from a set of good blocks from each of the selected memory dies, wherein a maximum number of metablocks that can be created across the selected memory dies is determined by a lowest number of good blocks in the selected memory dies.
1. A controller comprising:
one or more interfaces through which to communicate with a plurality of multi-chip memory packages, wherein each multi-chip memory package comprises a plurality of memory dies, and wherein each memory die has a respective plurality of memory blocks, some of which are good blocks and some of which are bad blocks; and
a processor in communication with the one or more interfaces, wherein the processor is configured to:
determine a number of good blocks in each memory die in each of the multi-chip memory packages;
based on the determined number of good blocks in each memory die, select a memory die from each of the multi-chip memory packages to access in parallel, wherein the selected memory dies are not necessarily all in the same relative position in each multi-chip package; and
create a metablock from a set of good blocks from each of the selected memory dies, wherein a maximum number of metablocks that can be created across the selected memory dies is determined by a lowest number of good blocks in the selected memory dies.
2. The controller of claim 1 , wherein the selecting and creating acts are performed for N instances of storage management software running on the controller, wherein N is the number of memory dies on each multi-chip package, and wherein different memory dies are selected in each instance.
3. The controller of claim 2 , wherein the memory dies are selected to uniformly spread good blocks across all the instances.
4. The controller of claim 2 , wherein the memory dies are selected to allocated the greatest number of good blocks together in a single instance.
5. The controller of claim 2 , wherein the processor is further configured to prioritize traffic to the instances according to a service level.
6. The controller of claim 2 , wherein the processor is further configured to prioritize traffic to the instances according to a traffic type.
7. The controller of claim 1 , wherein the controller is part of a solid-state drive.
8. The controller of claim 1 , wherein the controller is in communication with each multi-chip memory package using a different Open NAND Flash Interface (ONFI) channel.
9. The controller of claim 8 , wherein each ONFI channel supports a plurality of multi-chip memory packages.
10. The controller of claim 1 , wherein good blocks that are not allocated to a metablock are reserved as spare blocks for bad block remapping.
11. A method for creating a metablock across multi-chip memory packages, the method comprising:
performing in a controller comprising one or more interfaces through which to communicate with a plurality of multi-chip memory packages, wherein each multi-chip memory package comprises a plurality of memory dies, and wherein each memory die has a respective plurality of memory blocks, some of which are good blocks and some of which are bad blocks:
determine a number of good blocks in each memory die in each of the multi-chip memory packages;
based on the determined number of good blocks in each memory die, select a memory die from each of the multi-chip memory packages to access in parallel, wherein the selected memory dies are not necessarily all in the same relative position in each multi-chip package; and
create a metablock from a set of good blocks from each of the selected memory dies, wherein a maximum number of metablocks that can be created across the selected memory dies is determined by a lowest number of good blocks in the selected memory dies.
12. The method of claim 11 , wherein the selecting and creating acts are performed for N instances of storage management software running on the controller, wherein N is the number of memory dies on each multi-chip package, and wherein different memory dies are selected in each instance.
13. The method of claim 12 , wherein the memory dies are selected to uniformly spread good blocks across all the instances.
14. The method of claim 12 , wherein the memory dies are selected to allocated the greatest number of good blocks together in a single instance.
15. The method of claim 12 further comprising prioritizing traffic to the instances according to a service level.
16. The method of claim 12 further comprising prioritizing traffic to the instances according to a traffic type.
17. The method of claim 11 , wherein the controller is part of a solid-state drive.
18. The method of claim 11 , wherein the controller is in communication with each multi-chip memory package using a different Open NAND Flash Interface (ONFI) channel.
19. The method of claim 18 , wherein each ONFI channel supports a plurality of multi-chip memory packages.
20. The method of claim 11 , wherein good blocks that are not allocated to a metablock are reserved as spare blocks for bad block remapping.