Content-based deduplicated storage having multilevel data cache
View Patent ↗Methods and apparatus for a content-based deduplicated storage system for generating an address to hash (A2H) value for a control module as data blocks are received and generating, for a data module, hash to physical (H2P) values corresponding to the A2H values. A first cache can be provided for the control module, where the first cache can comprise an address value, a hash value, and physical location information. A second cache can be provided for the data module, where the second cache can comprise a bucket value, a hash value, and a filter mechanism, where the filter mechanism is configured to determine whether the hash value is present.
1. A method, comprising:
generating address to hash (A2H) values for a control module as data blocks are received;
generating, for a data module, hash to physical (H2P) values corresponding to the A2H values;
providing a first cache for the control module, the first cache comprising an address value, a hash value, and physical location information; and
providing a second cache for the data module having a bucket value, a hash cache, and a filter mechanism, wherein the filter mechanism is configured to determine whether a hash value for a write operation exists.
2. The method according to claim 1 , wherein the first cache uses prefetching based on offset values.
3. The method according to claim 1 , wherein the filter mechanism comprises a bloom filter.
4. The method according to claim 3 , wherein the bloom filter enables write operations without accessing meta data on disk.
5. The method according to claim 3 , wherein the bloom filter corresponds to a bucket, where a bucket corresponds to a hash space.
6. The method according to claim 3 , further including rebuilding the bloom filter when reaching a threshold of dirty entries.
7. The method according to claim 1 , wherein the second cache includes an eviction mechanism for retaining entries having higher reference counts.
8. The method according to claim 7 , wherein a reference count corresponds to a number of volumes having a given hash value.
9. The method according to claim 7 , further including caching hash values having reference counts at or above a given value at an address to hash level to avoid double caching.
10. The method according to claim 1 , further including for an incoming write, determining whether a hash for the incoming write exists in the filter mechanism, and, if the hash for the incoming write does not exist in the filter mechanism, writing data for the incoming write to disk.
11. The method according to claim 10 , when the hash for the incoming write does exist in the filter mechanism, retrieving a full hash value for the hash that exists in the filter mechanism.
12. The method according to claim 1 , wherein the filter mechanism comprises a bloom filter, wherein the bloom filter corresponds to a superbucket, where a superbucket corresponds to a hash space, wherein a superbucket comprises a logical aggregation of buckets, and further including using the superbucket to detect hash collisions.
13. An article, comprising:
a non-transitory computer-readable medium having stored instructions that enable a machine to perform:
generating address to hash (A2H) values for a control module as data blocks are received;
generating, for a data module, hash to physical (H2P) values corresponding to the A2H values;
providing a first cache for the control module, the first cache comprising an address value, a hash value, and physical location information; and
providing a second cache for the data module having a bucket value, a hash cache, and a filter mechanism, wherein the filter mechanism is configured to determine whether a hash value for a write operation exists.
14. The article according to claim 13 , wherein the first cache uses prefetching based on offset values.
15. The article according to claim 13 , wherein the filter mechanism comprises a bloom filter, wherein the bloom filter corresponds to a bucket, wherein the bucket corresponds to a hash space.
16. The article according to claim 15 , further including rebuilding the bloom filter when reaching a threshold of dirty entries.
17. The article according to claim 13 , wherein the second cache includes an eviction mechanism for retaining entries having higher reference counts.
18. The article according to claim 13 , further including instructions for determining, for an incoming write, whether a hash for the incoming write exists in the filter mechanism, and, if the hash for the incoming write does not exist in the filter mechanism, writing data for the incoming write to disk.
19. The article according to claim 18 , when the hash for the incoming write does exist in the filter mechanism, retrieving a full hash value for the hash that exists in the filter mechanism.
20. A content-based storage system with deduplication, comprising:
a processor and memory configured to:
generate address to hash (A2H) values for a control module as data blocks are received;
generate, for a data module, hash to physical (H2P) values corresponding to the A2H values;
provide a first cache for the control module, the first cache comprising an address value, a hash value, and physical location information; and
provide a second cache for the data module having a bucket value, a hash cache, and a filter mechanism, wherein the filter mechanism is configured to determine whether a hash value for a write operation exists.