Efficient IO processing

Patent metadata

Field	Value
Publication number	US-12455790-B2
Application number	US-202418624619-A
Country	US
Kind code	B2
Filing date	Apr 2, 2024
Priority date	Jan 21, 2020
Publication date	Oct 28, 2025
Grant date	Oct 28, 2025

Abstract

A method for a storage system to process input and output operations. The method includes receiving writes over time to an address at a base virtual volume, storing each of the writes in a physical storage at a new location that is without existing data, tagging each stored write with a different generation number to distinguish between different versions of data written to the address at the base virtual volume, receiving a read of the address at the base virtual volume, and, in response to the read of the address at the base virtual volume, returning one of the stored writes that is tagged with a newer generation number than a remainder of the stored writes.

First claim

What is claimed is:

1. A method for operating a storage system, the method comprising:
receiving a series of write requests from one or more storage clients to a base virtual volume, the write requests specifying respective data to be written to respective offsets in the base virtual volume, at least two of the write requests requesting writing to a same first offset in the base virtual volume, the two write requests corresponding to two versions of a same respective data;
for individual write requests of the series of write requests, performing a write process comprising:
assigning to the write request a generation number from a series of generation numbers, the generation number uniquely identifying the write request;
storing the data from the write request in a physical storage; and
adding an entry to a database, the entry identifying the offset in the base virtual volume, the generation number of the write request, and a location in which the data from the write request is stored.

2. The method of claim 1 , further comprising:
receiving a first read request requesting the data at the first offset in the base virtual volume; and
in response to the first read request, performing a read process comprising:
identifying a set of entries in the database including the entries that identify the first offset in the base virtual volume; and
returning the data stored at the location identified by a newest entry of the set of entries, the newest entry determined by having a generation number that is newer than the generation numbers of a remainder of the set of entries.

3. The method of claim 2 , further comprising performing a garbage collection process that reclaims space in the physical storage, the garbage collection process comprising:
invalidating the data at the locations identified by the remainder of the entries in the set, thereby reclaiming space in the physical storage; and
deleting from the database the remainder of the entries in the set.

4. The method of claim 1 , further comprising:
capturing a snapshot virtual volume of the base virtual volume by creating a first metadata structure that identifies a creation generation number of the snapshot virtual volume without copying data in the physical storage.

5. The method of claim 4 , further comprising:
receiving a second read request requesting data from the first offset in the snapshot virtual volume; and
in response to the second read request, performing a second read process comprising identifying a second set of entries in the database including entries that identify the first offset in the base virtual volume, and returning the data stored in the location identified by the entry that is in the second set that, based on the creation generation number of the snapshot and the generation numbers of the entries in the second set, is the newest of the entries in the set that are older than of the snapshot virtual volume.

6. The method of claim 1 , further comprising performing a garbage collection process that reclaims space in the physical storage, the garbage collection process comprising, for a range between a creation generation number of the base virtual volume and the creation generation number of a static view, deleting all but a stored write having the most recent generation number in the range.

7. The method of claim 1 , further comprising:
capturing a first snapshot virtual volume of the base virtual volume by creating a first metadata structure that identifies a first creation generation number of a first static view;
capturing a second snapshot virtual volume of the base virtual volume by creating a second metadata structure that identifies a second creation generation number of a second static view; and
reclaiming space in the physical storage by:
identifying a first set of entries including all entries in the database that identify offsets in the base virtual volume and generation numbers between a creation generation number of the base virtual volume and the first creation generation number; and
for individual offsets of the offsets identified by entries in the first set, identifying which of the entries in the first set and identifying the offset has a most recent generation number and invalidating the data at the locations identified by a remainder of the entries that are in the first set and identify the offset, thereby reclaiming space in the physical storage; and
identifying a second set of entries including all entries in the database that identify offsets in the base virtual volume and generation numbers between the first creation generation number and the second creation generation number; and
for individual offsets of the offsets identified by entries in the second set, identifying which of the entries in the second set and identifying the offset has a most recent generation number and invalidating the data at the locations identified by a remainder of the entries that are in the second set and identify the offset, thereby reclaiming space in the physical storage.

8. The method of claim 1 , further comprising:
receiving a command to copy data from a source offset range at the base virtual volume to a target offset range at a second virtual volume;
creating a metadata structure that identifies the source offset range, the target offset range, a creation generation number of the base virtual volume, and a creation generation number of the metadata structure;
receiving a second read request requesting the data at a second offset that is within the target offset range at the second virtual volume; and
in response to the second read request, performing a read process comprising identifying a second set of entries in the database that includes the entries that identify the second offset in the second virtual volume and that includes entries that identify an offset in the source offset range, and returning the data stored in the location identified by a newest one of the entries in the second set.

9. The method of claim 8 , further comprising reclaiming space in the storage by:
identifying a third set of entries that identify generation numbers in range between the creation generation number of the base virtual volume the creation generation number of the command; and
for each of the offsets identified by entries in the third set, identifying which of the entries in the third set and identifying the offset has a most recent generation number and invalidating the data at the locations identified by a remainder of the entries that are in the third set and identify the offset, thereby reclaiming space in the physical storage.

10. The method of claim 1 , further comprising performing an unmap command by:
receiving a command to unmap data from an offset range at the base virtual volume;
creating a first query range that identifies the offset range, a creation generation number of the base virtual volume, and a creation generation number of the first query range;
attaching the first query range to the base virtual volume;
creating a second query range that identifies (1) the base virtual volume, (2) a range between the creation generation number of the first query range and a current generation number, and (3) a zero view for returning zero data; and
attaching the second query range to the virtual volume.

11. The method of claim 10 , further comprising:
receiving a read of an offset in the offset range at the base virtual volume;
in response to receiving the read, using the first query range to search the base virtual volume for a stored write to the offset and having a generation number in the range;
when using the first query range to search the base virtual volume succeeds, returning

Assignees

• Nvida Corp
• Nvidia Corp

Inventors

• Wang Jin
• Nazari Siamak

Classifications

• G06F3/0653

  Monitoring storage devices or systems · CPC title

• G06F2201/84

  Using snapshots, i.e. a logical point-in-time copy of the data · CPC title

• G06F3/0689 Primary

  Disk arrays, e.g. RAID, JBOD · CPC title

• G06F3/0665

  at area level, e.g. provisioning of virtual or logical volumes · CPC title

• G06F3/0659

  Command handling arrangements, e.g. command buffers, queues, command scheduling · CPC title