Efficient computation of sketches
Determining a summary feature set is disclosed. A plurality of subsegments of a first segment are selected. For each subsegment, a plurality of values by applying a set of functions to each subsegment are computed. From all the values computed for all the subsegments, a first subset of values is selected.
1. A system for resemblance compression comprising:
a processor configured to:
divide a data stream into a plurality of segments;
for a segment of the plurality of segments:
compute a summary feature set for the segment;
determine whether the segment resembles a stored segment using the summary feature set; and
in the event that the segment resembles the stored segment, store the segment as a composite of the stored segment and one or more deltas, wherein the stored segment corresponds to a base segment and the one or more deltas at least correspond to a difference between two different segments.
2. A system as in claim 1 , wherein determining whether the segment resembles a stored segment comprises selecting a plurality of subsegments of the segment.
3. A system as in claim 2 , wherein determining whether the segment resembles a stored segment comprises computing a plurality of values by applying a set of functions to each subsegment.
4. A system as in claim 3 , wherein the set of functions comprise a set of linear functions.
5. A system as in claim 3 , wherein the set of functions comprise m k x k +a k wherein m k and a k are predetermined values for k=1 to n, and wherein x k comprises a function of the subsegment values for k=1 to n.
6. A system as in claim 3 , wherein the set of functions comprise m k x+a k wherein m k and a k are predetermined values for k=1 to n and wherein x comprises a function of the subsegment values.
7. A system as in claim 6 , wherein the function of the subsegment values comprises one or more of the following: SHA-1, MD5, RIPEMD-160, a Rabin hash, and a hash function.
8. A system as in claim 3 , wherein from all the values computed for all the subsegments, selecting a first subset of values.
9. A system as in claim 8 , wherein selecting the first subset of values comprises selecting one or more of the following: a plurality of extrema values of the function results, a plurality of minima values of the function results, a plurality of maxima values of the function results, and a plurality of minima and maxima values of the function results.
10. A system as in claim 8 , wherein the first subset of values are used in determining resemblance between the first segment and a second segment.
11. A system as in claim 1 , further comprising hashing a plurality of underlying segments corresponding to the first subset of values to a second subset of values.
12. A system as in claim 11 , wherein the second subset of values is used in determining resemblance between the first segment and a second segment.
13. A system as in claim 1 , further comprising the step of composing a third subset of values based on the first subset of values.
14. A system as in claim 13 , wherein the third subset of values is used in determining resemblance between the first segment and a second segment.
15. A system as in claim 13 , wherein the order of composing the third subset of values is pre-determined and independent of the values in the first subset.
16. A system as in claim 13 , further comprising composing a fourth subset of values based on the second subset of values.
17. A system as in claim 16 , wherein the fourth subset of values is used in determining resemblance between the first segment and a second segment.
18. A system as in claim 16 , wherein the order of composing the fourth subset of values is pre-determined and independent of the values in the second subset.
19. A computer program product for resemblance compression, the computer program product being embodied in a non-transitory computer readable medium and comprising computer instructions for:
dividing a data stream into a plurality of segments;
for a segment of the plurality of segments:
computing a summary feature set for the segment;
determining whether the segment resembles a stored segment using the summary feature set; and
in the event that the segment resembles the stored segment, storing the segment as a composite of the stored segment and one or more deltas, wherein the stored segment corresponds to a base segment and the one or more deltas at least correspond to a difference between two different segments.
20. A method for resemblance compression comprising:
dividing, by a processor, a data stream into a plurality of segments;
for a segment of the plurality of segments:
computing, by the processor, a summary feature set for the segment;
determining whether the segment resembles a stored segment using the summary feature set; and
in the event that the segment resembles the stored segment, storing in a memory coupled to the processor the segment as a composite of the stored segment and one or more deltas, wherein the stored segment corresponds to a base segment and the one or more deltas at least correspond to a difference between two different segments.