DIFFERENCING ENGINE FOR MOVING PICTURES

Abstract

Techniques for efficiently storing multiple versions of a moving picture are disclosed. A first version of a moving picture is stored in a memory. For additional versions of the moving picture, in lieu of storing entire versions thereof, one or more files with data subsets are generated and stored, wherein the data subsets represent the identified differences between the first version and the additional versions.

Description

FIELD

The present disclosure relates generally to data deduplication and efficient data storage, and more particularly, to techniques for storing moving pictures.

BACKGROUND

Moving pictures can require substantial amounts of memory for storage in digital form. In the motion picture industry, for example, when a major motion picture is released, it is not uncommon for a movie studio to produce between 30-70 variations of the movie. Some variations can be minor in nature. These variations may include, for example, an original master and a number of modified versions of the original master. The different versions may be created for various reasons. Some versions may include adjusted introduction information or credits to account for different audiences or formats. Other versions may include additional or different scenes (such as for a director's cut and a theatre version, for instance). Still other versions may be the result of editing the movie to make it compatible with a specific rating (e.g., PG-13, R, GP, NC-17, and the like), or with the laws of a specific country in which the film is intended for distribution. The editing to produce such variations may involve editing, changing, and deleting scenes, and/or producing more than one version of a scene.

The end result is that the studio or production company typically stores a potentially large number of versions of a film, albeit, in many cases, each version including minor differences as determined by the overall percentage of similarities of the file(s) with the original. For full-length feature films, each such version can occupy 250-500 GB or more of storage. In addition, any given studio may produce or otherwise own hundreds or thousands of different movies altogether, each with their own set of variations. Current libraries used to store these types of files often include over 200,000 moving pictures. As a result, these production and studio companies often store hundreds of thousands of copies of the movies in multiple locations to operate. These requirements can translate to exabytes of data storage and exorbitant storage costs.

These and other limitations are addressed in the present disclosure.

SUMMARY

In an aspect of the disclosure, a deduplication method, a processing system, and a computer readable medium for storing moving picture files are provided.

In one aspect of the disclosure, a deduplication method includes storing, in a memory, first data representing a moving picture, determining a data subset comprising differences between second data representing another version of the moving picture and the first data, and storing the data subset in the memory, wherein the data subset and the first data are sufficient to reproduce the second data.

In another aspect of the disclosure, a processing system includes a memory configured to store first data representing a moving picture, and one or more processors configured to identify second data representing another version of the moving picture, determine a data subset comprising differences between the first data and the second data such that the second data is reproducible using at least the data subset and the first data and without the second data, and store the data subset in the memory.

In another aspect of the disclosure, a computer readable medium includes code that, when executed by one or more processors, causes the one or more processors to store, in a memory, first data representing a moving picture, identify second data representing another version of the moving picture, determine a data subset comprising differences between the first data and the second data such that the second data is reproducible using at least the data subset and the first data and without the second data, and store the data subset in the memory.

Additional advantages and novel features will be set forth in part in the description that follows, and in part will become more apparent to those skilled in the art upon examination of the following or upon learning by practice thereof.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a conceptual diagram illustrating a file transfer using delta differential compression based on static mapping.

FIG. 2 is a flow diagram illustrating a file transfer using delta differential compression based on static mapping.

FIG. 3 is a conceptual diagram illustrating a file transfer using delta differential compression based on on-the-fly mapping.

FIG. 4 is a flow diagram illustrating a file transfer using delta differential compression based on on-the-fly mapping.

FIG. 5 is a conceptual diagram illustrating a file transfer using delta differential compression based on cumulative mapping.

FIG. 6 is a flow diagram illustrating a file transfer using delta differential compression based on cumulative mapping.

FIG. 7 is a diagram illustrating an engine for effecting file transfers using delta differential compression.

FIG. 8 is a diagram illustrating an apparatus for file transfers using delta differential compression.

FIG. 9 is a conceptual diagram illustrating a technique for creating a fingerprint map of a file.

FIG. 10 is a flow diagram illustrating a technique for creating a fingerprint map of a file.

FIG. 11 is a flow diagram illustrating a technique for computing a delta of a file and a fingerprint map of another file.

FIG. 12 is a flow diagram illustrating a technique for performing delta differencing on a raw data stream.

FIG. 13 is a flow diagram illustrating a technique for storing multiple versions of moving picture files.

DETAILED DESCRIPTION

The detailed description set forth below in connection with the appended drawings is intended as a description of various configurations and is not intended to represent the configurations in which the concepts described herein may be practiced. The detailed description includes specific details for the purpose of providing a thorough understanding of various concepts. However, it will be apparent to those skilled in the art that these concepts may be practiced without these specific details. In some instances, well known structures and components are shown in block diagram form in order to avoid obscuring such concepts.

Several aspects of systems for data transfer will now be presented with reference to various apparatus and methods. These apparatus and methods will be described in the following detailed description and illustrated in the accompanying drawings by various blocks, modules, components, circuits, steps, processes, algorithms, etc. (collectively referred to as “elements”). These elements may be implemented using electronic hardware, computer software, or any combination thereof. Whether such elements are implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system.

By way of example, an element, or any portion of an element, or any combination of elements may be implemented with a “processing system” that includes one or more processors. Examples of processors include microprocessors, microcontrollers, digital signal processors (DSPs), field programmable gate arrays (FPGAs), programmable logic devices (PLDs), state machines, gated logic, discrete hardware circuits, and other suitable hardware configured to perform the various functionality described throughout this disclosure. One or more processors in the processing system may execute software. Software shall be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executables, threads of execution, procedures, functions, etc., whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise.

Accordingly, in one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software, the functions may be stored on or encoded as one or more instructions or code on a computer-readable medium. Computer-readable media includes computer storage media. Storage media may be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can comprise a random-access memory (RAM), a read-only memory (ROM), an electrically erasable programmable ROM (EEPROM), compact disk ROM (CD-ROM) or other optical disk storage, magnetic disk storage or other magnetic storage devices, combinations of the aforementioned types of computer-readable media, or any other medium that can be used to store computer executable code in the form of instructions or data structures that can be accessed by a computer.

FIG. 1 is a conceptual diagram 100 illustrating a file transfer using delta differential compression based on static mapping. FIG. 1 represents one example of a static mapping mode. In an embodiment, as files are transferred, a map of file fingerprints is stored in a hidden file at the source node. As described herein, a static fingerprint map may consume on average less than 0.1% of the file size. Subsequently, when files are changed and designated for transfer, only the changed data is sent, significantly reducing bandwidth utilization and speeding the remote file synchronization. Concurrently, the locally stored static fingerprint map may be updated to reflect the now current file versions.

Block 102 represents a source node, such as the apparatus 800 for transferring files in FIG. 8. The source node 102 may represent a server or other computing device, or a collection of computing devices. Block 112 represents a destination node. Blocks 104, 106, 108, 110 and 111 represent events that may occur within source node 102 and these events may occur in the order of time from lower to higher reference numerals, as illustrated by the vertical arrow designated by “t”. Similarly, blocks 114, 116 and 117 represent events that may occur within destination node 112 and these events may also occur in the order of time from lower to higher numbers. Beginning at 104, a file F1 is received at source node. In other configurations, F1 may already be resident at the source node. F1 may be stored in a memory, such as a non-volatile memory, at source node 102. At 106 a fingerprint map MAP1 is generated using an appropriate algorithm, as discussed in greater detail below with reference to FIGS. 9 and 10. Thereupon, MAP1 is stored in memory MEM at source node 102.

As illustrated in FIG. 1 by the arrow designated A, F1 at that time may be transferred to the destination node 112 via network 120. The arrival of F1 at destination node 112 is illustrated by the corresponding arrow designated A and block 114. In other configurations, source node 102 transfers MAP1 to destination node 112 as illustrated by the dotted arrow designated B and block 116. As illustrated in block 116 at destination node 112, the transfer of MAP1 is optional and may be performed in lieu of or in addition to the transfer of F1 to destination node 112. In still other embodiments, F1 may be transferred to destination node 112 as indicated by block 114, but then the destination node 112, rather than the source node 102, may create the FM (MAP1) corresponding to F1 and then may transmit MAP1 back to the source node 112.

In the above illustration, there is an original version of F1 (104) and a copy of F1 (114) where the copy was either transferred directly from source node 102, was reproduced from MAP1 and other data or was made available by some other means.

The technique of FIG. 1 involves computing a fingerprint map of file F1, which is discussed with reference to FIGS. 9 and 10. Because creating the fingerprint map is generally one of the initially steps, one embodiment of a technique for creating the fingerprint map is now described. FIG. 9 is a conceptual diagram 900 illustrating a technique for creating a fingerprint map of a file. A file or a section of a file may be regarded as a pattern of bytes known as fingerprints. A fingerprint may include a well-defined relatively short length data packet sampled at an arbitrary position from a data stream. The fingerprints may represent sections of a file or data stream and may be selected on the basis of their relative “uniqueness” within the fingerprint map. An example of a short fingerprint is one that is much shorter than the length of the data stream itself but, in one configuration, not longer than 1024 bits. The fingerprinting of a stream map includes finding all the qualifying fingerprints throughout the stream. A fingerprint map of a stream includes the collection of the fingerprints and their positions within that stream. Finding a quality fingerprint is generally dependent upon the entropy (degree of randomness) of the stream.

Using an appropriate algorithm as illustrated below, large files may be represented as one or more maps of fingerprints where the map is comparatively smaller than the original file, where the fingerprint map may be used to identify similarities within files.

A file in general consists of any number of bytes that can be represented by values in the range {00h . . . FFh}. If a subset of byte patterns can be identified whose probability of random occurrence is very small and it is ensured that the patterns occur once, such patterns may be used as fingerprints of the file. A fingerprint is similar to a checksum in this regard, except that it is not computed but created from the file itself.

Four criteria for a statistically qualifying fingerprint include (i) minimal frequency of occurrence of the fingerprint within the data stream, (ii) minimal frequency of occurrence of the fingerprint with the fingerprint map, (iii) a high entropy (i.e., highly random bits), and (iv) low collision probability. In the context of the 128-bit fingerprinting algorithm discussed in greater detail below, the following general process may be used to gather the fingerprint map of a stream:

1. Divide the stream into fragments, which gives the sample window length.

2. Sample a 128 bit packet in window of data, which represents the 128 bit candidate fingerprint.

3. Check the 128 bit candidate fingerprint for minimum entropy.

4. Check the 128 bit candidate fingerprint for low frequency.

5. If the conditions are satisfied and the fingerprint meets a quality threshold, shift the window by 128 bits and repeat from step 2.

FIG. 9 shows a section of a file comprising a pattern of bytes 902. Byte pattern 902 includes a preceding section, a 256 byte section, and a remaining section. In one configuration, each adjacent 256 byte section is consecutively analyzed. An exemplary 128 bit fingerprint algorithm searches for and selects eight bytes each within a 256 byte boundary from a given pattern by using a statistical approach that identifies uniqueness within the pattern. For example, a portion 904 of the 256 byte section is shown in FIG. 9. In the portion 904 as well as the entire 256 byte section being analyzed, FE and E8 hex only occur once. Based on their minimal occurrence within the pattern, the following bytes 906 are selected including FE and E8, and their coordinates 908 identifying their relative position in the pattern are packed into another 8 byte array. The occurrence field 907 in this example indicates the number of occurrences of the respective bytes 906 in the pattern. As will be appreciated by one skilled in the art, a general objective of the fingerprint selection algorithm is to ensure that the selected fingerprints do not occur anywhere else in the pattern so as to satisfy the relative “uniqueness” of the fingerprint within the pattern.

The bytes 906 and coordinates 908 are placed into a candidate 128-bit fingerprint 910. The candidate fingerprint 910 is then matched against any other corresponding fingerprints selected to date, as shown in pattern 912. In one configuration, a determination of whether a match is present is made at each byte boundary of the original pattern. If a match is found, the candidate fingerprint 910 is discarded and the next 256 byte section of the pattern is analyzed.

Generally, it can be shown that assuming a “normal” distribution of bits in a sample of any size, the probability of occurrence of any given fingerprint is the same, and that probability depends only on the size of the sample. For example, the probability of a 128 bit fingerprint occurring once in a one Terabyte random sample is about 1 in a five hundred million. While the presence of a “normal” distribution of bits is rare in practice with everyday files, matching fingerprints may be eliminated from the map by checking the file for fingerprint repetitions as described above. Once the distribution of a 256 byte sample is analyzed, the byte spectrum may be taken as the basis to select the fingerprint that would have the lowest probability of occurrence if the file had the same distribution as the block sample.

Thus, in the embodiment of FIG. 9 where a preceding 256 byte section 914 is analyzed in the event of a match, the section shifts forward by 256 bytes and the next section 916 is analyzed, followed by the next section 918. The above steps are repeated until the end of the pattern is reached or a suitable fingerprint is found.

In sum, the 128-bit (16-byte) fingerprints are constructed by sampling 256-byte continuous blocks from the file. The coordinates 908 in FIG. 9 represent 8 bit offsets identifying the respective positions of the bytes 906 within the block, as noted above. The 8 bit offsets obtained in this manner also serve as a byte mask. More specifically, when the block is shifted within the file as described with reference to FIG. 9, the mask selects a new set of fingerprint bytes. The fingerprints—i.e., the eight bytes and their respective offsets—are selected by performing statistical analyses of the block to ensure that they meet a high entropy criteria or the lowest probability of random occurrence. Using the offset mask, a fingerprint can be matched with all other fingerprints in the file by byte shifting the block. By searching fingerprints at arbitrary offsets an array of fingerprints that maps the entire file may be obtained.

FIG. 10 is a flow diagram 1000 illustrating a technique for creating a fingerprint map of a file. In the embodiment shown, obtaining the fingerprint map may be effected by first deciding how many fingerprints are required to map the file (1002). The number of fingerprints is known as the granularity. To obtain a better resolution for obtaining a Δ of a file as discussed herein, a higher granularity can be set to obtain more fingerprints. Based on the granularity, the file being mapped is divided into sections referred to herein as links (1002). An example of a link may be the pattern 902 in FIG. 9. While the average size of a link may be arbitrary, its actual size may vary depending on the location where the fingerprints are found. Here, as in FIG. 9, a 256-byte section is sampled (1004) and eight bytes are selected on the basis of a small probability of random occurrence pursuant to the entropy equation being used in the fingerprinting algorithm (1006). After an 8-bit offset is selected for each byte representing a part of the candidate fingerprint (1008), the candidate fingerprint is matched against the rest of the masked fingerprints in the link (1012) and selected if it matches no other. Specifically, if no match is found, the candidate fingerprint is added along with its 20 bit hash value is added to the list of fingerprints in the map (1014). If a match is found, control may return to 1006 or it may proceed to 1016 depending on the algorithm.

The offset at which the fingerprint found marks the beginning of the link, and the next fingerprint will mark the links end (1016). In this manner, each link is assigned a new fingerprint and each new candidate fingerprint is ensured to not match against any previous one, which in turn ensures that each fingerprint occurs only once in the file. In addition to the fingerprint, each link size and offset is also stored along with its sha2 message digest in the fingerprint map (1018). Continuing this technique, an array of fingerprints of the file is obtained (1020). The number of fingerprints may be determined by the number of divisions of the file and the byte entropy of the file. Generally, this number is no more than the granularity selected. The granularity, in turn, may be the predetermined average number of divisions, which in one embodiment is at a minimum ten times the number of expected file changes. The total fingerprint map FM in one embodiment is the total number of fingerprints, array hashes, message digests and link offsets of the original file. The map is sufficient to compute the Δ between the file and any other file, as described further below.

In an embodiment, the fingerprint map is computed using an entropy equation characterized by an alphabet of N letters, whereby for each data chunk a probability is determined that a K length word has less than L number of repeating letters. For a fingerprint of 16 bytes, this problem is solved for L=1 to 16, with K=16 and N=256. The data chunk length may vary depending on the boundary of each identified fingerprint.

In one embodiment, the minimum entropy is defined as the frequency of occurrence of any octet at an 8 bit boundary in the 128 bit fingerprint (128 bit sequence). This frequency of occurrence may be designated as a number range between 1-16 such that the higher the number the lower the entropy and the lower the worth of a fingerprint, and vice versa. If the entropy of a fingerprint is greater than a minimum threshold, it is not a good candidate and another candidate is selected. Based on real world data in the experience of the inventors, the best entropy minimum appears to be around 4 in the 1-16 number range. The given criteria can be mathematically verified based on the minimum entropy and assuming a normal random distribution of bits in a stream. The probabilities of occurrence for the different minimum entropies are set forth in the following tables:

TABLE 1
PROBABILITES FOR FINGERPRINTS WITH A CERTAIN
(E)NTROPY
E = 1: 7.52E−37, <0.00E+00, >=1.00E+00
E = 2: 6.29E−30, <7.52E−37, >=1.00E+00
E = 3: 3.48E−25, <6.29E−30, >=1.00E+00
E = 4: 2.12E−21, <3.48E−25, >=1.00E+00
E = 5: 3.41E−18, <2.12E−21, >=1.00E+00
E = 6: 2.13E−15, <3.41E−18, >=1.00E+00
E = 7: 6.40E−FIG. 13
<2.14E−15, >=1.00E+00
E = 8: 1.04E−10, <6.42E−13, >=1.00E+00
E = 9: 9.88E−09, <1.05E−10, >=1.00E+00
E = 10: 5.76E−07, <9.99E−09, >=1.00E+00
E = 11: 2.12E−05, <5.86E−07, >=1.00E+00
E = 12: 4.94E−04, <2.18E−05, >=1.00E+00
E = 13: 7.24E−03, <5.16E−04, >=9.99E−01
E = 14: 6.40E−02, <7.76E−03, >=9.92E−01
E = 15: 3.09E−01, <7.17E−02, >=9.28E−01
E = 16: 6.20E−01, <3.80E−01, >=6.20E−01

TABLE 2
BYTES TO SAMPLE UNTIL THERE BECOMES 50%
CHANCE OF FINDING FP-128 LESS THEN E
E < 2: 12786308680337326080.00 EBytes
E < 3: 1530270711808.00 EBytes
E < 4: 27641568.00 EBytes
E < 5: 4541.08 EBytes
E < 6: 2.82 EBytes
E < 7: 4.67 PBytes
E < 8: 15.71 TBytes
E < 9: 98.74 GBytes
E < 10: 1.03 GBytes
E < 11: 18.05 MBytes
E < 12: 497.89 KBytes
E < 13: 20.99 KBytes
E < 14: 1.39 KBytes
E < 15: 149 Bytes
E < 16: 23 Bytes

In some situations, acceptable approximations may be made. Finding quality fingerprints cannot always be guaranteed as it is highly dependent upon the byte entropy of the file. For example, there may be large sections of a file consisting of only a few different bytes, and long sequences of repeating bytes. In many embodiments, the 128 bit fingerprint is strong down to a minimum of 4 different bytes in the maximum separation range of 256 bytes. At this or greater separation range, however, the data may become extremely compressible. Thus, while fingerprinting becomes progressively more difficult to find in the file, the file's compressibility may increase by magnitudes. In the event the fingerprinting algorithm identifies this situation, in one configuration, the algorithm may simply skip analysis of the entire link at issue because the link does not meet the minimum entropy requirement. Thereupon, the link may be marked as “compressible” and its byte statistics may be stored in the FM (MAP1 in FIG. 1) instead of its checksums. In an embodiment, no actual compression takes place at the FM analysis stage. Instead, the compression is performed subsequently when the Δ is computed as described below.

It is assumed for the purposes of this illustration that a number of changes or updates are made to F1, whether at source node 102 or otherwise, to produce file F2 as reflected at block 108. As is often the case, updated file F2 may contain large amount of redundancies. The redundancies may also consume a comparatively high amount of the overall file space. Examples of redundancies may include images, color graphics, and blocks of identical text. The greater the file size and the greater the number of redundancies, the greater the savings of bandwidth that can be achieved. Referring back to FIG. 1, at block 110, a file (here, Δ1) is generated which is a computation of the difference between F2 and MAP1. In some embodiments F1 may be used in lieu of MAP1 to compute the difference.

In one embodiment, Δ1 is obtained by first checking F2 at each byte offset of the file for matches with the fingerprints of MAP1. FIG. 11 is a flow diagram 1100 illustrating a technique for computing a delta (Δ) of a file and a fingerprint map of another file. The computation of Δ may commence by checking F2 at every byte offset for matches with the fingerprints from MAP1 (1102). This search for matches may be expedited by using the 20 bit fingerprint hashes stored earlier in connection with the creation of MAP1. If a fingerprint match is found (1103), it is next determined whether the linked fingerprint also matches. This determination may be accomplished by using the stored offset of the adjacent fingerprint. If the stored offset also matches, it is likely that an unmodified link has been found. This is true because the probability that the fingerprint occurred as a result of random chance is extremely small, and the probability of two matching fingerprints being present at the given distance apart is even smaller. Nevertheless, to ensure that the link is identical to the original link in F1, the sha2 digest of the link may be computed and compared with the corresponding sha2 digest stored in connection with the creation of MAP1 (1104). If a match exists it may be concluded that the portion of the file (link) is identical to the original (1108).

While the technique described in this embodiment does not guarantee this conclusion to a certainty (such a guarantee would generally require a byte to byte comparison), the conclusion is likely in view of the very small collision chance of the sha2 digest. Furthermore the chance that an sha2 collision occurs while the fingerprints themselves match is even smaller.

In the event that the sha2 digests do not match, the search for matching fingerprints continues, and all bytes that are processed are marked as new data (1110). Also, if an adjacent fingerprint does not match, all bytes processed continue to be marked as new data. After processing the remainder of F2 in this manner, the changes between F1 and F2 are identified and the unmodified links are discovered. The sum of the new data and the unmodified links is Δ, which can be applied to MAP1 to obtain F2.

In an embodiment, during the delta processing where certain portions of F2 are found to be compressible, they may be encoded with a run-length-encoded (RLE) compression formula, which can achieve an extremely high compression ratio. The degree of compressibility may be high in view of the fact that the byte entropy may be required to get very low before the fingerprinting fails for that portion. For example, a repeating pattern of 1 Kbyte may be compressed to 100:1 by RLE, and the compressed data is than added to Δ1. The described algorithms are easily tunable and scale well for multiple CPU cores. The fingerprint mapping may be computed simultaneously with the delta or from the backup copy of the file. How and when the delta is applied depends on the situation and may be implemented separately depending on the applications needs. The minimum link size and the fingerprint range are tunable to accommodate smaller files, but the algorithm is well suited for files greater than 1 Mb. In one configuration, the default and maximum fingerprint range is 256 bytes, which is the maximum distance between two fingerprint bytes. Compression is optional but for sparse files it may add significant additional benefits for accelerating file transfer operations.

Referring again back to FIG. 1, now that Al has been computed, Δ1 may be transferred as a data stream or a file (or part of another file or file set) over network 120 to the destination node 112, as indicated by the arrows marked as C. Assuming the destination nodes are not preconfigured to correctly interpret the file containing Δ1, then in some embodiments the source node 102 may transmit an indication to the destination node 112 that destination node 112 may reproduce F2 using the file containing Δ1 and MAP1. This indication, if not predetermined, may result from information contained in the Δ1 file, or in a separate file or data transmission.

At this point in time, destination node 112 has the Δ1 and F1 (or MAP1). Thereupon, destination node 112 is able to rapidly compute F2 based on a standard application of Δ1 to F1 (block 118).

FIG. 2 is a flow diagram 200 illustrating a file transfer using delta differential compression based on static mapping. In FIG. 2, the delta differential file engine may operate on an arbitrary number of n files. It is assumed for purposes of simplicity that an (n=1) file represents an updated or modified version of an n_th file, or otherwise represents a separate file with possible similarities so as to justify the use of delta compression as describe herein. An nth file is received at a source node such as source node 102 of FIG. 1 (202). Source node 102 thereupon creates an nth FM corresponding to the n_th file (204). In alternative embodiments, the n_th FM may be created by the destination node 112 and transmitted back to source node 102. The n_th FM is written to memory for use in subsequent operations (206). The n_th file (and/or the n_th FM) is transferred to a destination node, such as destination node 112 of FIG. 1 (208). Subsequently, an (n+1)_th file is received or otherwise made available at source node 102 (210). The source node 102 generates, using the (n+1)_th file and the n_th FM, an n_th Δ or data representing a difference between the n_th file and the (n+1)_th file (212). The source node 102 transfers the data over a network to destination node 112 along with an indication to destination node 112 to generate the (n+1)_th file using the n_th Δ (i.e., the data) and the n_th file (214). Thereafter, the source node 102 creates an (n+1)_th FM corresponding to the (n+1)_th file (216) and saves the (n+1)_th FM to memory (218) for use in a subsequent delta operation. The process may be repeated (230) or varied as additional files are received that qualify for delta compression.

A number of alternative embodiments may be contemplated in view of FIGS. 1 and 2 and may vary depending on, for example, the manner in which files at the source node are determined to be updated or modified versions of other files, or otherwise determined to be candidates for computing deltas.

In FIG. 2, the fields 220, 222, 224, 226, and 228 represent examples of various events. Field 220 represents an event where a request has been made, such as by a destination node (or in some embodiments by the source node or another external node), to transfer the n_th file to the destination node. Field 222 represents an event where the source node has received the n_th file or the n_th file is otherwise made available at the source node. Field 224 represents an event where a request has been made, such as by the destination node (or in some embodiments by the source node or another external node), to transfer the (n+1)_th file to the destination node. Field 226 represents an event where the source node has received the (n+1)_th file or the (n+1)_th file is otherwise made available at the source node. Field 228 represents an event where the file has been received at the destination node, or is otherwise available at the destination node (for example, an FM of the file may have been received or generated at the destination node). The small circles in FIG. 2 and the corresponding arrows constitute a matrix representing that in certain illustrative embodiments, the identified steps may be performed in response to the identified events. By way of example, in some situations a FM may be created (204) in response to the request for the file (220), the receipt at the source node of the file (222), a request for a modified version of the file (224), the receipt at the source node of the modified version of the file (226), or the availability of the file at the destination node (228), or some combination thereof The illustrated events and resulting steps are exemplary in nature, and other implementations may be equally suitable depending on the application.

Whether any given file is a candidate for delta compression as described herein may be determined using any of a variety of methods at the source node. For example, certain fields or metadata corresponding to a document may be retrieved at the source node to determine whether the file has been identified as an updated or modified version of an existing file. In another configuration, the document title may be provided in an identifiable format. Alternatively, a basic comparison of the content of a candidate file may be made with one or more existing files to determine its suitability for delta compression.

In another embodiment where a file is subsequently updated a number of times, the source node may only maintain and save a FM of the original file, and may reuse the FM and recomputed respective deltas for each subsequent modified version of the file. Alternatively, the source node may create a new FM of each iteration of the file.

By using the techniques disclosed herein, substantial bandwidth savings can be achieved since only the changes to documents need to be transmitted over the network to a destination. The FM may be maintained in memory at the source node for use in computing deltas corresponding to future modifications.

FIG. 3 is a conceptual diagram 300 illustrating a file transfer using delta differential compression based on on-the-fly mapping. In short, unlike the static mapping as described in FIGS. 1 and 2 in which the FM is created and stored in memory for use in accelerating transfers of updated files, the approach in FIG. 3 involves creation of the FM “on-the-fly” when either the modified file is made available at the source node or a source file is received which is determined to be different from a destination file at the destination node. In this latter embodiment, the source and destination files may be compared to determine whether the files are different so as to initiate a delta operation. A metadata check may determine if the source and destination files contain differences. If a file change is detected, the FM may be created at the destination and transmitted to the source service. The delta is computed at the source node based on the source file and the FM transmitted to the source. As above, only the identified changed data is transferred to the destination node, and the file is updated at the destination node.

Further, unlike the embodiment of FIGS. 1 and 2, the FM need not be saved in memory but may immediately be used to calculate the delta. At source node 302, file F1 is received (block 304). At that time, F1 may be written to memory or transferred immediately over network 320 to destination node 312 (block 314) over the arrows designated A. Subsequently, file F2 is received or made available at source node 302 (block 306). It is determined by any means that F2 is a modified version of F1 or is otherwise a possible candidate for delta compression (e.g., there may be substantial similarities between F1 and F2). Alternatively, upon receiving F2 at the source node, a comparison between F1 at the destination node and F2 at the source node may be made to ascertain whether differences are present in the file. This determination may be made by, among other means, a metadata check as described above. MAP1 is thereupon generated at source node 302 (block 308), and MAP1 may be transferred to destination node 312 via network 320 in addition to F1 (block 316) as represented by the arrow designated B. F2 may be reproduced using the computed Δ and F1 (or MAP1) (block 318).

In alternative embodiments, MAP1 may be generated on the fly at the destination node and transferred to the source node, such as, for example, in response to determining that F2 at the source node and F2 at the destination node contain differences. Thereupon, Δ is calculated (block 310) using MAP1 and F2 (block 310). The Δ is then transmitted via network 320 to destination node 312, where it is used along with F1 to reproduce F2. Subsequently, where a third file is received that is a modified version of the second, the source node 302 or destination node 312 may create MAP2 on the fly corresponding to that file for subsequent delta operations (block 311).

In addition to the advantages associated with the embodiment of FIG. 1, by creating the FM on the fly and processing the FM prior to or in lieu of storing it to memory, the latencies associated with an extra file write may be eliminated, which can accelerate the overall file transfer.

FIG. 4 is a flow diagram 400 illustrating a file transfer using delta differential compression based on on-the-fly mapping. The diagram shown relates to updates or successive iterations of an n_th file. By way of example, a treatise may be periodically updated in electronic form. An n_th file is received at source node 302 (402). Subsequently an (n+1)_th file is received at source node 302 (404). Following a determination that the files are different, an n_th FM is created corresponding to the n_th file, in response to receiving the (n+1)_th file (406). The n_th file (and in some embodiments, also the n_th FM) is transferred to destination node 312 (408). The Δ representing the difference between the (n+1)_th file and the n_th file is generated using the (n+1)_th file and the n_th FM (412). The data may then be transferred to the destination node (414). Subsequently an (n+2)_th file is received (416), and an (n+1)_th FM is created in response to receiving the (n+2)_th file (417). Then the source node 302 generates a new Δ representing a difference between the (n+2)_th file and the (n+1)_th file using the (n+2)_th file and the (n+1)_th FM (418). The steps may be repeated to accommodate the arrival of new files (430).

In FIG. 4, as shown by the fields 420, 422, 424, 426 and 428 and associated arrow matrix, the various identified steps may be performed in response, for example, to one or more of the events corresponding to these fields, as more specifically described above with reference to FIG. 2 above.

FIG. 5 is a conceptual diagram 500 illustrating a file transfer using delta differential compression based on cumulative mapping. In an embodiment of cumulative mode, an initial set of files is mapped during a transfer. A collective FM of all the files in the set is created. This collective FM may contain reference to all chunks of data and all file paths in the set of files. Subsequently, some subset of the original files may change, as determined by any one of a number of methods. The cumulative Δ of the changed files is obtained. This Δ may contain references to any known chunks across the initial file set. The destination file set is thereupon updated, taking advantage of possible common chunks found in any file of the set. The collective FM is also updated. This technique may have significant benefits if the files in the set contain common chunks.

Referring to FIG. 5, a directory of files F1, F2, F3 is received at the source node 502 (block 504) and transferred via network 520 to destination node 512 (block 514) as represented by the arrows designated A. A collective or cumulative MAP1 is created corresponding to the files in the directory (block 506), and MAP1 may optionally be provided to the destination node 512 (block 516) via network 520 in addition to the directory of files, as represented by the arrows designated B. Thereupon, at source node 502 a file F4 is received which contains F1′ and F2′ (block 508). F1′ and F2′ contains one or more chunks of F1, and F2′ contains one or more chunks of F2. In other embodiments, file F′ may be received which contains chunks from two or more of files F1, F2, F3, and other, unrelated files. The corresponding Δ is generated using MAP1 and F4, in a manner described herein (block 510). Δ is transferred to destination node 512 via network 520, as represented by the arrows designated C. The destination node next generates F4 using the Δ and the three files F1, F2, F3 (block 518). MAP1 may be updated to MAP2 concurrently at the source node 502 (block 511) to account for the addition of F4.

In alternative embodiments to those shown in FIGS. 1-5, the FM may be computed at the destination node and transmitted back to the source node.

FIG. 6 is a flow diagram 600 illustrating a file transfer using delta differential compression based on cumulative mapping. A set of m files is written to a directory where F=F1+F2+F3+Fm, with m equal to some integer greater than or equal to four (602). An n_th cumulative FM is created corresponding to the initial set of m files (604). In embodiments where F is transferred to the destination node 512, the destination node 512 may create the map corresponding to F and transmit it back to the source node. The n_th FM is written to memory (606), and F (and in some cases the n_th FM) is transferred to the destination node 512 (608). Subsequently the source node 502 receives F′, which in this example is a subset of F1′+F2′+Fm′ (610). Each of F1′, F2′ and Fm′ contain chunks of data from, respectively, F1+F2+Fm. The source node 502 generates, using F′ and the n_th FM, Δ representing a difference between F′ and F (612). The Δ is then transferred over the network to destination node 512 along with an indication to the destination node that the F′ file set can be reproduced using the Δ and F (614). The FM is updated to an (n+1)_th FM corresponding to F′ (616), and the updated FM is written to memory (618). The process may then continue for subsequent file sets (630).

In FIG. 6, as shown by the fields 620, 622, 624, 626 and 628 and associated arrow matrix, the various identified steps may be performed in response to one or more of the events corresponding to these fields, as more specifically described above with reference to FIG. 2 above.

FIG. 7 is a diagram illustrating an engine 700 for effecting file transfers using delta differential operations. A source node 705 may include source host 702, user interface (UI) 701, and storage 705. A destination node 719 may include similar or substantially the same resources as source host 702. An example of such a configuration may include a set of distributed file server nodes across a network. Destination node 719 may include destination host 716 and may be coupled to other peripheral devices for receiving the files, such as mobile station (MS) 740 and personal computing device (PC) 720. Incorporated separately or as part of destination node 719 is backup drive or storage array 703. In one embodiment, the source node 705 and the destination node 719 transfer data over one or more links 708 across a network 710. A1 in this embodiment constitutes a file transfer application employing delta differencing operations as described herein. In one embodiment, file transfer application A1 fragments the data in a set of files to be transferred into n data streams 730, processes and packetizes the data streams using buffer 704 and the processor system of the source host 702, and transmits the resulting packets over links 708 to the destination node 719. At the destination host 716, the results are de-packetized using buffer 734 into n data streams 736 and reconsolidated and reassembled into the original files, represented by 2. The files 2 may then be backed up in a suitable storage array 703.

A2 represents one or more separate applications from which the data constituting the files are obtained. A2 may reside on the same machine or a different machine to that of source host 702. The “P/S” indication shows that the data may be sent via a pipe or socket connection to application A1, or another connection type depending on the physical configuration employed.

FIG. 8 is a diagram illustrating an apparatus 800 for file transfers using delta differential operations. The apparatus includes a processor, which may be implemented as the processing system described above. The apparatus includes memory 804 (which in some configurations may be part of the processing system). Memory 804, in turn, includes a buffer such as ring buffer 806 with separate buffer locations 814, and a main memory such as random access memory 808. The apparatus further includes non-volatile memory/storage 810 (which in some configurations may be part of the processing system), such as a hard drive or storage array. Coupled to the processor 802, memory 804 and non-volatile memory/storage 810 is transceiver 812, which typically contains the electronics and protocol for transmitting and receiving files in the form of data packets over the network 818 and associated links. A separate PC 816 may allow a user to download files received on apparatus 800. The steps described in any of the conceptual diagrams or flow diagrams, and further steps as described in the disclosure herein, may be implemented as one or more software module run on processor 802, one or more firmware modules, or one or more dedicated hardware modules.

In another aspect of the disclosure, a technique for raw data differencing is disclosed, which can accelerate the transfer of raw data across a network. One use of the technique described in this implementation is an exemplary file transfer engine such as that described in connection with FIG. 7. Source host 702 and destination host 719 may be involved in the transfer of raw data across the network, such that source host 702 has no concept of files or directories and just transfers the unstructured data. In such an embodiment, an on-the-fly technique may be used. Alternatively, a streaming technique may be used as described with reference to FIG. 12.

In FIG. 12, a stream of raw data D is received from an external source at a source host (e.g., source host 702). A predetermined length of the raw data is assembled as a file D1 and stored in a first portion 1202 of a history cache. Concurrently or soon thereafter, FM module 1206 generates a fingerprint map FM1 corresponding to file D1 and/or to packets in file D1, and FM1 is stored in a second portion 1204 of the history cache. The file D1 is sent across the network 1210 over transmission medium A and written to memory 1212 at a destination. Subsequently, from the continued stream D, an additional file D2 may be created and stored in the first portion of the history cache 1202, and FM2 may be created and stored in the second portion 1204 of the history cache. Additional files D3-D5 may be generated and transmitted over network 1210, and so forth. The history cache 1204 may be updated with the storage of additional fingerprint maps F3-F5 until the cache needs to be overwritten with new data.

As new files (e.g. D6, D7, D8) from stream D are generated and stored in the history cache, host 702 compares, via pipe 1208 or a similar interface, the new files to the fingerprint maps (e.g., FM1-FM5) stored in the history cache. If a match is detected, such as if packets in the file are detected to be identical to those represented in one of the stored fingerprint maps, the host 702 transmits an indication to the destination that the file (or particular sections thereof) is already present at the destination and also may send a pointer identifying the file or sections.

As an illustration involving a transfer of a data and a roughly 1 TB history cache 1204, as a first 100 MB of data arrives at the host 702, the history cache 1204 may be empty. After 1 TB of data is received, the host may create a 1 TB file constituting the data and a fingerprint map of the file. The file and fingerprint map may be stored in the history cache at the host. As additional data is received beyond the 1 TB, the data may be identified to correspond with one of the stored maps, whereupon the source 702 sends an appropriate indication over the network. It is understood that the specific order or hierarchy of blocks in the processes/flow charts disclosed is an illustration of exemplary approaches. Based upon design preferences, it is understood that the specific order or hierarchy of blocks in the processes/flow charts may be rearranged. Further, some blocks may be combined or omitted. The accompanying method claims present elements of the various blocks in a sample order, and are not meant to be limited to the specific order or hierarchy presented.

In other aspects of the disclosure, deduplication techniques for storage of moving picture data are presented. For the purposes of this disclosure, a “moving picture” broadly includes all forms of video and media files, motion pictures, film, television, video clips, shorts, interactive graphics, moving images, animation, and the like. While embodiments in this disclosure are described for clarity in the illustrative case of multiple versions of a motion picture, the concepts disclosed and claimed herein broadly encompass applications relating to all forms of moving pictures as defined above. Data deduplication refers broadly to data compression techniques for eliminating duplicate copies of repeating data.

With reference to the problem of storage inefficiencies arising from the need to store multiple versions of the same motion picture, approaches are presented herein for substantially reducing storage requirements. As noted above, many such versions involve minor differences including, for example, deletion of a promiscuous scene to change a film's rating (e.g., for distribution to an airline), changing of title or credit information, and the like. In lieu of storing multiple versions, a file can be generated to include only a data subset representing a difference between an original moving picture and another version thereof. The other version is represented as a difference in a small file. Only the original moving picture and the difference file(s) are stored. To reproduce the other version, the file is “differenced” with the original, using any of a plurality of techniques including without limitation those described in this disclosure.

Using these techniques, a studio or distribution company need only store one master of the film and one or more difference files representing different versions of the file. Each version of the movie may be reproduced using the difference file(s) and the master. The inventors have estimated that content distribution companies like Netflix® can save on the order of $100 million a year in storage costs. Similar companies in the content distribution space and the satellite and cable business stand to achieve comparable savings.

FIG. 13 is a flow diagram illustrating a technique for storing multiple versions of moving pictures. An original moving picture 1302 is provided at I₁. The original moving picture may be in the form of a file or a plurality of files, a collection of streaming data, or another known data structure. The moving picture may also be in the form of a fingerprint map. In one embodiment, the steps discussed below may occur in the context of a processing system as described herein. The processing system may include one or more processors on a single machine, or the processing system may be distributed or networked across multiple machines in one or more locations. The processing system may include other components such as memory, as illustrated in the exemplary embodiment of FIG. 8. The memory, such as memory 1304, may similarly be situated in one location or distributed/networked across multiple locations.

Referring back to FIG. 13, the processing system stores the original moving picture 1302 as file F_O in memory 1304, which may in some embodiments include one or more dedicated hard drives, tape drives, or any suitable type of memory on which data or files may be stored. Thereupon, three versions MP1-MP31306-10 of the moving picture are made available. Using any known suitable technique such as herein-described, the processing system generates a data subset of differencing files D₁-D₃ (1312, 1314, 1316) (which may each include a single file or multiple files, depending on the embodiment) that respectively correspond to versions MP1-MP3. In other embodiments, the processing system may compute a single differencing file (or file set thereof) D_T (1318) that includes the differencing data relative to all three versions MP1-MP3.

Differencing files D₁-D₃ and D_T include data that represents the differences in, additions to, and or deletions from, the original moving picture 1302. In most instances, such files represent a very small percentage of the respective versions MP1-MP3. Having computed and generated the differencing files, the processing system stores D₁-D₃ (or alternatively, D_T) in the memory 1304 for “long term” storage. The original moving picture files MP1-MP3 need not be preserved. In some embodiments, only a fingerprint map(s) of the original moving picture 1302 is stored in memory 1304. In some embodiments, fingerprint maps of one or more of the original moving picture files 1302 and/or files MP1-MP3 (1312, 1314, 1316) may be created, stored in memory 1304, and/or used to facilitate generation of the differencing file(s), as contemplated previously in this disclosure.

In other embodiments, the processing system may physically constitute two or more separate systems implementing different functions relevant to the above-described steps. For example, one part of the processing system may handle the generation of the differencing data, while another part of the processing system may handle the receiving of files and data and the storage of the files and data in memory, as well as the retrieval therefrom.

Subsequently, to reproduce the versions MP1-MP31306, 1308, 1310 based on the content of memory 1304, the processing system uses the original file F_O and the differencing file D₁ (1320) (using the techniques described herein, for example) to reproduce version MP1 (1326). The processing system uses the original file F_O and the differencing file D₂ (1322) to reproduce version MP2 (1328). The processing system uses the original file F_O and the differencing file D₃ (1322) to reproduce version MP3 (1330).

The described techniques may be implemented one or more standalone application suites in some embodiments that have user interfaces for manipulation by a user, or they may be implemented behind the file system relative to the perspective of a user. In one implementation of the interface for end users, the engine for differencing and storing different versions of an original moving picture can be hidden inside a user space file system. The process in this embodiment is transparent to the end user and integrates seamlessly with other capabilities of the processing system. In this embodiment, the user will not need to do anything new in order use the benefits of this technique other than mount to and read from another device (file system). In another embodiment, the differencing method is non-block aligned.

Other embodiments also may not rely on a single original master and may instead store more than one file, each of which may be used to recreate other versions. For the purposes of this disclosure, a “version” broadly refers to a file that includes at least in part the substance of another file.

The previous description is provided to enable any person skilled in the art to practice the various aspects described herein. Various modifications to these aspects will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other aspects. Thus, the claims are not intended to be limited to the aspects shown herein, but is to be accorded the full scope consistent with the language claims, wherein reference to an element in the singular is not intended to mean “one and only one” unless specifically so stated, but rather “one or more.” The word “exemplary” is used herein to mean “serving as an example, instance, or illustration.” Any aspect described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects. Unless specifically stated otherwise, the term “some” refers to one or more. Moreover, nothing disclosed herein is intended to be dedicated to the public regardless of whether such disclosure is explicitly recited in the claims. No claim element is to be construed as a means plus function unless the element is expressly recited using the phrase “means for.”

Claims

1. A deduplication method, comprising: storing, in a memory, first data representing a moving picture;generating a data subset comprising differences between second data representing another version of the moving picture and the first data; andstoring the data subset in the memory,wherein the data subset and the first data are sufficient to reproduce the second data.

2. The method of claim 1, further comprising: reproducing the second data using the first data and the data subset; andstoring the second data in the memory.

3. The method of claim 1, wherein the data subset is generated using at least the second data.

4. The method of claim 1, wherein the moving picture comprises at least one moving picture file.

5. The method of claim 4, further comprising: generating a first fingerprint map of the at least one moving picture file,wherein the first data comprises the first fingerprint map.

6. The method of claim 5, wherein the data subset is generated using at least the second data and the first data.

7. The method of claim 1, further comprising transferring a file comprising the data subset to a destination node.

8. A processing system; comprising: a memory configured to store first data representing a moving picture; andone or more processors configured to: identify second data representing another version of the moving picture;determine a data subset comprising differences between the first data and the second data such that the second data is reproducible using at least the data subset and the first data and without the second data; andstore the data subset in the memory.

9. The processing system of claim 7, wherein the one or more processors are further configured to: reproduce the second data using at least the first data and the data subset and without the second data; andstore the second data in the memory.

10. The processing system of claim 7, wherein the data subset is determined using at least the second data.

11. The processing system of claim 7, wherein the moving picture comprises at least one moving picture file.

12. The processing system of claim 10, wherein the one or more processors are further configured to: generate a first fingerprint map of the at least one moving picture file,wherein the first data comprises the first fingerprint map.

13. The processing system of claim 11, wherein the data subset is determined using at least the second data and the first data.

14. The processing system of claim 10, further comprising: a transceiver configured to transmit a file comprising the data subset to a destination node.

15. A non-transitory computer readable medium comprising code that, when executed by one or more processors, causes the one or more processors to: store, in a memory, first data representing a moving picture;identify second data representing another version of the moving picture;determine a data subset comprising differences between the first data and the second data such that the second data is reproducible using at least the data subset and the first data and without the second data; andstore the data subset in the memory.

16. The computer readable medium of claim 14, wherein the code causes the one or more processors to: reproduce the second data using at least the first data and the data subset and without the second data; andstore the second data in the memory.

17. The computer readable medium of claim 14, wherein the data subset is determined using at least the second data.

18. The computer readable medium of claim 14, wherein the moving picture comprises at least one moving picture file.

19. The computer readable medium of claim 16, wherein the code causes the one or more processors to: generate a first fingerprint map of the at least one moving picture file,wherein the first data comprises the first fingerprint map.

20. The computer readable medium of claim 17, wherein the data subset is determined using at least the second data and the first data.