System and method for distributed erasure coding
A system and method for distributed erasure coding. A plurality of storage devices is directly connected to one or more host computers, without an intervening central controller distributing data to the storage devices and providing data protection. Parity codes are stored in one or more dedicated storage devices or distributed over a plurality of the storage devices. When a storage device receives a write command, it calculates a partial parity code, and, if the parity code for the data being written is on another storage device, sends the partial parity code to the other storage device, which updates the parity code using the partial parity code.
1. A method for storing data in an array of storage devices, the method comprising:
receiving, by a first storage device of the array of storage devices, a write command from a host computer, the write command including write data to be written to the first storage device at a logical block address in the first storage device;
retrieving by the first storage device, old data stored at the logical block address;
calculating, by the first storage device, from the old data and the write data, a partial parity code, the partial parity code being a number that:
when evaluated, together with an old parity code,
the old parity code being a data protection code for an old virtual stripe of data including the old data and extending across all of the array of storage devices,
results in a new parity code, the new parity code being a data protection code for a new virtual stripe of data extending across all of the array of storage devices and including the write data instead of the old data;
evaluating, by a second storage device of the array of storage devices, the old parity code together with the partial parity code, to generate the new parity code;
storing, by a third storage device of the array of storage devices, the new parity code; and
storing, by the first storage device, the write data.
2. The method of claim 1 , wherein the third storage device is a same storage device as the second storage device.
3. The method of claim 1 , further comprising:
calculating, by the first storage device, an identifier of the second storage device;
determining whether the identifier of the second storage device is different from an identifier of the first storage device; and
when the identifier of the second storage device is different from the identifier of the first storage device, sending, to the second storage device, a parity message comprising the partial parity code.
4. The method of claim 3 , wherein the write command further includes a namespace identifier, and the calculating of the identifier of the second storage device comprises:
obtaining a namespace offset from a namespace offset table based on the namespace identifier; and
calculating a virtual stripe identifier as the sum of the namespace offset and the logical block address.
5. The method of claim 4 , wherein the calculating of the identifier of the second storage device comprises:
retrieving, by the first storage device, from a first table in the first storage device, a stride size; and
calculating:
the virtual stripe identifier
modulo
the stride size.
6. The method of claim 5 , wherein the sending, to the second storage device, of the parity message comprises:
retrieving, by the first storage device, from a second table in the first storage device, an address of the second storage device; and
sending the parity message to the address of the second storage device.
7. The method of claim 6 , wherein:
the first storage device is an Ethernet storage device;
the second storage device is an Ethernet storage device; and
the address of the second storage device is an Ethernet address of the second storage device.
8. The method of claim 6 , further comprising storing, by a storage system management circuit, the address of the second storage device in the second table in the first storage device.
9. The method of claim 8 , further comprising storing, by the storage system management circuit, the stride size in the first table.
10. The method of claim 6 , wherein:
the first storage device is connected to a peripheral component interconnect express switch;
the second storage device is connected to the peripheral component interconnect express switch; and
the address of the second storage device is a peripheral component interconnect express address of the second storage device.
11. The method of claim 1 , wherein each of the old parity code and the new parity code is a data protection code utilized in a data protection system selected from the group consisting of RAID 4, RAID 5, RAID 6, and erasure code data protection systems.
12. The method of claim 1 , wherein the first storage device is connected to a storage interface switch, the method further comprising, upon failure or disconnection of the first storage device:
reconfiguring, by a storage system management circuit, the storage interface switch to route storage commands addressed to the first storage device to the storage system management circuit; and
sending, in response to a received storage command, by the storage system management circuit, the same command completion, through the storage interface switch, as the first storage device would have sent were it present and operating correctly.
13. The method of claim 12 , further comprising:
recovering, by the storage system management circuit, data previously stored on the first storage device; and
storing the recovered data on a fourth storage device.
14. The method of claim 1 , wherein the first storage device is connected to a storage interface switch, the method further comprising, upon failure or disconnection of the first storage device:
reconfiguring, by a storage system management circuit, the storage interface switch to route storage commands addressed to the first storage device to a fourth storage device; and
sending, in response to a received storage command, by the fourth storage device, the same command completion, through the storage interface switch, as the first storage device would have sent were it present and operating correctly.
15. The method of claim 1 , wherein the first storage device is connected to a storage interface switch, the method further comprising, upon occurrence of a discrete failure in the first storage device and receipt of a read command:
recovering data affected by the discrete failure, by the first storage device; and
sending, in response to the read command, by the first storage device, the same command completion, through the storage interface switch, as the first storage device would have sent had the discrete failure not occurred.
16. The method of claim 1 , wherein the first storage device is connected to a storage interface switch, the method further comprising, upon occurrence of a discrete failure in the first storage device and receipt of a read command:
recovering data affected by the discrete failure, by a storage system management circuit; and
sending, in response to the read command, by the storage system management circuit, the same command completion, through the storage interface switch, as the first storage device would have sent had the discrete failure not occurred.
17. A system for storing data, comprising an array of storage devices, the system being configured to:
receive, by a first storage device of the array of storage devices, a write command from a host computer, the write command including write data to be written to the first storage device at a logical block address in the first storage device;
retrieve by the first storage device, old data stored at the logical block address;
calculate, by the first storage device, from the old data and the write data, a partial parity code, the partial parity code being a number that:
when evaluated, together with an old parity code,
the old parity code being a data protection code for an old virtual stripe of data including the old data and extending across all of the array of storage devices,
results in a new parity code, the new parity code being a data protection code for a new virtual stripe of data extending across all of the array of storage devices and including the write data instead of the old data;
evaluate, by a second storage device of the array of storage devices, the old parity code together with the partial parity code, to generate the new parity code;
store, by a third storage device of the array of storage devices, the new parity code; and
store, by the first storage device, the write data.
18. The system of claim 17 , wherein the third storage device is a same storage device as the second storage device.
19. The system of claim 17 , further configured to:
calculate, by the first storage device, an identifier of the second storage device;
determine whether the identifier of the second storage device is different from an identifier of the first storage device; and
when the identifier of the second storage device is different from the identifier of the first storage device, sending, to the second storage device, a parity message comprising the partial parity code.
20. The system of claim 19 , wherein the calculating of the identifier of the second storage device comprises:
retrieving, by the first storage device, from a first table in the first storage device, a stride size; and
calculating:
the logical block address
modulo
the stride size.
21. The system of claim 19 , wherein the sending, to the second storage device, of the parity message comprises:
retrieving, by the first storage device, from a second table in the first storage device, an address of the second storage device; and
sending the parity message to the address of the second storage device.
22. The system of claim 21 , wherein:
the first storage device is an Ethernet storage device;
the second storage device is an Ethernet storage device; and
the address of the second storage device is an Ethernet address of the second storage device.
23. The system of claim 21 , further comprising a storage system management circuit configured to store the address of the second storage device in the second table in the first storage device.
24. The system of claim 21 , wherein:
the first storage device is connected to a peripheral component interconnect express switch;
the second storage device is connected to the peripheral component interconnect express switch; and
the address of the second storage device is a peripheral component interconnect express address of the second storage device.
25. The system of claim 17 , wherein each of the old parity code and the new parity code is a data protection code utilized in a data protection system selected from the group consisting of RAID 4, RAID 5, RAID 6, and erasure code data protection systems.
26. A system for storing data, comprising:
an array of storage devices; and
storage system management means,
the system being configured to:
receive, by a first storage device of the array of storage devices, a write command from a host computer, the write command including write data to be written to the first storage device at a logical block address in the first storage device;
retrieve by the first storage device, old data stored at the logical block address;
calculate, by the first storage device, from the old data and the write data, a partial parity code, the partial parity code being a number that:
when evaluated, together with an old parity code,
the old parity code being a data protection code for an old virtual stripe of data including the old data and extending across all of the array of storage devices,
results in a new parity code, the new parity code being a data protection code for a new virtual stripe of data extending across all of the array of storage devices and including the write data instead of the old data;
evaluate, by a second storage device of the array of storage devices, the old parity code together with the partial parity code, to generate the new parity code;
store, by a third storage device of the array of storage devices, the new parity code; and
store, by the first storage device, the write data.
27. The system of claim 26 , wherein the storage system management means is configured to store, in each of the storage devices, an address of another storage device.