Patent Grant
9600515
Field of the Invention
The present invention relates in general to computers, and more particularly to an efficient calculation of both similarity search values and boundaries of digest blocks in a data deduplication system in a computing environment.
Description of the Related Art
In today's society, computer systems are commonplace. Computer systems may be found in the workplace, at home, or at school. Computer systems may include data storage systems, or disk storage systems, to process and store data. Large amounts of data have to be processed daily and the current trend suggests that these amounts will continue being ever-increasing in the foreseeable future. An efficient way to alleviate the problem is by using deduplication. The idea underlying a deduplication system is to exploit the fact that large parts of the available data are copied again and again, by locating repeated data and storing only its first occurrence. Subsequent copies are replaced with pointers to the stored occurrence, which significantly reduces the storage requirements if the data is indeed repetitive.
In one embodiment, a method is provided for efficient calculation of both similarity search values and boundaries of digest blocks in a data deduplication system using a processor device in a computing environment. In one embodiment, by way of example only, input data is partitioned into data chunks, and for each chunk a set of rolling hash values is calculated. A single linear scan of the rolling hash values is used to produce both similarity search values and boundaries of the digest blocks of the chunk. The rolling hash values are used to contribute to the calculation of the similarity search values and to the calculation of the boundaries of the digest blocks.
In addition to the foregoing exemplary method embodiment, other exemplary system and computer product embodiments are provided and supply related advantages. The foregoing summary has been provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter. The claimed subject matter is not limited to implementations that solve any or all disadvantages noted in the background.
In order that the advantages of the invention will be readily understood, a more particular description of the invention briefly described above will be rendered by reference to specific embodiments that are illustrated in the appended drawings. Understanding that these drawings depict embodiments of the invention and are not therefore to be considered to be limiting of its scope, the invention will be described and explained with additional specificity and detail through the use of the accompanying drawings, in which:
Data deduplication is a highly important and vibrant field in computing storage systems. Data deduplication refers to the reduction and/or elimination of redundant data. In data deduplication, a data object, which may be a file, a data stream, or some other form of data, is broken down into one or more parts called chunks or blocks. In a data deduplication process, duplicate copies of data are reduced or eliminated, leaving a minimal amount of redundant copies, or a single copy of the data, respectively. The goal of a data deduplication system is to store a single copy of duplicated data, and the challenges in achieving this goal are efficiently finding the duplicate data patterns in a typically large repository, and storing the data patterns in a storage efficient deduplicated form. A significant challenge in deduplication storage systems is scaling to support very large repositories of data. Such large repositories can reach sizes of Petabytes (1 Petabyte=250 bytes) or more. Deduplication storage systems supporting such repository sizes, must provide efficient processing for finding duplicate data patterns within the repositories, where efficiency is measured in resource consumption for achieving deduplication (resources may be CPU cycles, RAM storage, persistent storage, networking, etc.). In one embodiment, a deduplication storage system may be based on maintaining a search optimized index of values known as fingerprints or digests, where a (small) fingerprint represents a (larger) block of data in the repository. The fingerprint values may be cryptographic hash values calculated based on the blocks' data. In one embodiment, secure hash algorithm (SHA), e.g. SHA-1 or SHA-256, which are a family of cryptographic hash functions, may be used. Identifying fingerprint matches, using index lookup, enables to store references to data that already exists in a repository.
To provide reasonable deduplication in this approach, the mean size of the data blocks based on which fingerprints are generated must be limited to smaller sizes and may not be too large. The reason being that a change of a bit within a data block will probabilistically change the data block's corresponding fingerprint, and thus having large data blocks makes the scheme more sensitive to updates in the data as compared to having small blocks. A typical data block size may range from 4 KB to 64 KB, depending on the type of application and workload. Thus, by way of example only, small data blocks may range in sizes of up to 64 KB, and large data blocks are those data blocks having a size larger than 64 KB.
To support very large repositories scaling to Petabytes (e.g., repositories scaling to at least one Petabyte), the number of fingerprints to store coupled with the size of a fingerprint (ranging between 16 bytes and 64 bytes), becomes prohibitive. For example, for 1 Petabyte of deduplicated data, with a 4 KB mean data block size, and 32 bytes fingerprint size (e.g. of SHA-256), the storage required to store the fingerprints is 8 Terabytes. Maintaining a search optimized data structure for such volumes of fingerprints is difficult, and requires optimization techniques. However existing optimization techniques do not scale to these sizes while maintaining performance. For this reason, to provide reasonable performance, the supported repositories have to be relatively small (on the order of tens of TB). Even for such smaller sizes, considerable challenges and run-time costs arise due to the large scale of the fingerprint indexes that create a bottle-neck in deduplication processing.
To solve this problem, in one embodiment, a deduplication system may be based on a two step approach for searching data patterns during deduplication. In the first step, a large chunk of incoming data (e.g. a few megabytes) is searched in the repository for similar (rather than identical) data chunks of existing data, and the incoming data chunk is partitioned accordingly into intervals and paired with corresponding (similar) repository intervals. In the second step, a byte-wise matching algorithm is applied on pairs of similar intervals, to identify identical sub-intervals, which are already stored in a repository of data. The matching algorithm of the second step relies on reading all the relevant similar data in the repository in order to compare it byte-wise to the input data.
Yet, a problem stemming from a byte-wise comparison of data underlying the matching algorithm of the second step, is that data of roughly the same size and rate as the incoming data should be read from the repository, for comparison purposes. For example, a system processing 1 GB of incoming data per second, should read about 1 GB of data per second from the repository for byte-wise comparison. This requires substantially high capacities of I/O per second of the storage devices storing the repository data, which in turn increases their cost.
Additional trends in information technology coinciding with the above problem are the following: (1) Improvements in the computing ability by increasing CPU speeds and the number of CPU cores. (2) Increase in disk density, while disk throughput remains relatively constant or improving only modestly. This means that there are fewer spindles relative to the data capacity, thus practically reducing the overall throughput. Due to the problem specified above, there is a need to design an alternative solution, to be integrated in a two step deduplication system embodiment specified above, that does not require reading from the repository in high rates/volumes.
Therefore, in one embodiment, by way of example only, additional embodiments address these problem, as well as shifts resource consumption from disks to the CPUs, to benefit from the above trends. The embodiments described herein are integrated within the two step and scalable deduplication embodiments embodiment described above, and uses a similarity search to focus lookup of digests during deduplication. In one embodiment, a global similarity search is used as a basis for focusing the similarity search for digests of repository data that is most likely to match input data.
The embodiments described herein significantly reduce the capacity of I/O per second required of underlying disks, benefit from the increases in computing ability and in disk density, and considerably reduce the costs of processing, as well as maintenance costs and environmental overhead (e.g. power consumption).
In one embodiment, input data is segmented into small segments (e.g. 4 KB) and a digest (a cryptographic hash value, e.g. SHA1) is calculated for each such segment. First, a similarity search algorithm, as described above, is applied on an input chunk of data (e.g. 16 MB), and the positions of the most similar reference data in the repository are located and found. These positions are then used to lookup the digests of the similar reference data. The digests of all the data contained in the repository are stored and retrieved in a form that corresponds to their occurrence in the data. Given a position of a section of data contained in the repository, the digests associated with the section of data are efficiently located in the repository and retrieved. Next, these reference digests are loaded into memory, and instead of comparing data to find matches, the input digests and the loaded reference digests are matched.
The described embodiments provide a new fundamental approach for architecting a data deduplication system, which integrates a scalable two step approach of similarity search followed by a search of identical matching segments, with an efficient and cost effective digest/fingerprint based matching algorithm (instead of byte-wise data comparison). The digest/fingerprint based matching algorithm enables to read only a small fraction (1%) of the volume of data required by byte-wise data comparison. The present invention proposed herein, a deduplication system can provide high scalability to very large data repositories, in addition to high efficiency and performance, and reduced costs of processing and hardware.
In one embodiment, by way of example only, the term “similar data” may be referred to as: for any given input data, data which is similar to the input data is defined as data which is mostly the same (i.e. not entirely but at least 50% similar) as the input data. From looking at the data in a binary view (perspective), this means that similar data is data where most (i.e. not entirely but at least 50% similar) of the bytes are the same as the input data.
In one embodiment, by way of example only, the term “similar search” may be referred to as the process of searching for data which is similar to input data in a repository of data. In one embodiment, this process may be performed using a search structure of similarity elements, which is maintained and searched within.
In one embodiment, by way of example only, the term “similarity elements” may be calculated based on the data and facilitate a global search for data which is similar to input data in a repository of data. In general, one or more similarity elements are calculated, and represent, a large (e.g. at least 16 MB) chunk of data.
Thus, the various embodiments described herein provide various solutions for digest retrieval based on a similarity search in deduplication processing in a data deduplication system using a processor device in a computing environment. In one embodiment, by way of example only, input data is partitioned into fixed sized data chunks. Similarity elements digest block boundaries and digest values are calculated for each of the fixed sized data chunks. Matching similarity elements are searched for in a search structure (i.e. index) containing the similarity elements for each of the fixed sized data chunks in a repository of data. Positions of similar data are located in a repository. The positions of the similar data are used to locate and load into the memory stored digest values and corresponding stored digest block boundaries of the similar data in the repository. It should be noted that in one embodiment the positions may be either physical or logical (i.e. virtual). The positions are of data inside a repository of data. The important property of a ‘position’ is that given a position (physical or logical) in the repository's data, the data in that position can be efficiently located and accessed. The digest values and the corresponding digest block boundaries are matched with the stored digest values and the corresponding stored digest block boundaries to find data matches.
Thus, the various embodiments described herein provide various solutions for digest retrieval based on a similarity search in deduplication processing in a data deduplication system using a processor device in a computing environment. In one embodiment, by way of example only, input data is partitioned into fixed sized data chunks. Similarity elements, digest block boundaries and digest values are calculated for each of the fixed sized data chunks. Matching similarity elements are searched for in a search structure (i.e. index) containing the similarity elements for each of the fixed sized data chunks in a repository of data. Positions of similar data are located in a repository. The positions of the similar data are used to locate and load into the memory stored digest values and corresponding stored digest block boundaries of the similar data in the repository. The digest values and the corresponding digest block boundaries are matched with the stored digest values and the corresponding stored digest block boundaries to find data matches.
In one embodiment, the present invention provides a solution for utilizing a similarity search to load into memory the relevant digests from the repository, for efficient deduplication processing. In a data deduplication system, deduplication is performed by partitioning the data into large fixed sized chunks, and for each chunk calculating (2 things−similarity elements and digest blocks/digest values) hash values (digest block/digest value) for similarity search and digest values. The data deduplication system searches for matching similarity values of the chunks in a search structure of similarity values, and finds the positions of similar data in the repository. The data deduplication system uses these positions of similar data to locate and load into memory stored digests of the similar repository data, and matching input and repository digest values to find data matches.
In one embodiment, the present invention provides for efficient calculation of both similarity search values and segmentation (i.e. boundaries) of digest blocks using a single linear calculation of rolling hash values. In a data deduplication system, the input data is partitioned into chunks, and for each chunk a set of rolling hash values is calculated. A single linear scan of the rolling hash values produces both similarity search values and boundaries of the digest blocks of the chunk. Each rolling hash value corresponds to a consecutive window of bytes in byte offsets. The similarity search values are used to search for similar data in the repository. The digest blocks segmentation is used to calculate digest block boundaries and corresponding digest values of the chunk, for digests matching. Each rolling hash value contributes to the calculation of the similarity values and to the calculation of the digest blocks segmentations. Each rolling hash value may be discarded after contributing to the calculations. The described embodiment provides significant processing efficiency and reduction of CPU consumption, as well as considerable performance improvement.
Thus, as described above, the deduplication approach of the present invention uses a two-step process for searching data patterns during deduplication. In the first step, a large chunk of incoming data (e.g. 2 megabytes “MB”) is searched in the repository for similar (rather than identical) chunks of existing data, and the incoming chunk is partitioned accordingly into intervals, and paired with corresponding (similar) repository intervals. The similarity index used in the first step is compact and simple to maintain and search within, because the elements used for a similarity search are very compact relative to the data they represent (e.g. 16 bytes representing 4 megabytes). Further included in the first step, in addition to a calculation of similarity elements, is a calculation of digest segments and respective digest values for the input chunk of data. All these calculations are based on a single calculation of rolling hash values. In the second step, reference digests of the similar repository intervals are retrieved, and then the input digests are matched with the reference digests, to identify data matches.
In one embodiment, in the similarity based deduplication approach as described herein, a stream of input data is partitioned into chunks (e.g. at least 16 MB), and each chunk is processed in two main steps. In the first step a similarity search process is applied, and positions of the most similar reference data in the repository are found. Within this step both similarity search elements and digest segments boundaries are calculated for the input chunk, based on a single linear calculation of rolling hash values. Digest values are calculated for the input chunk based on the produced segmentation, and stored in memory in the sequence of their occurrence in the input data. The positions of similar data are then used to lookup the digests of the similar reference data and load these digests into memory, also in a sequential form. Then, the input digests are matched with the reference digests to form data matches.
When deduplication of an input chunk of data is complete, the input chunk of data's associated digests are stored in the repository, to serve as reference digests for subsequent input data. The digests are stored in a linear form, which is independent of the deduplicated form by which the data these digests describe is stored, and in the sequence of their occurrence in the data. This method of storage enables efficient retrieval of sections of digests, independent of fragmentation characterizing deduplicated storage forms, and thus low on IO and computational resource consumption.
Turning now to
To facilitate a clearer understanding of the methods described herein, storage controller 240 is shown in
In some embodiments, the devices included in storage 230 may be connected in a loop architecture. Storage controller 240 manages storage 230 and facilitates the processing of write and read requests intended for storage 230. The system memory 243 of storage controller 240 stores program instructions and data, which the processor 242 may access for executing functions and method steps of the present invention for executing and managing storage 230 as described herein. In one embodiment, system memory 243 includes, is in association with, or is in communication with the operation software 250 for performing methods and operations described herein. As shown in
In some embodiments, cache 245 is implemented with a volatile memory and non-volatile memory and coupled to microprocessor 242 via a local bus (not shown in
Storage 230 may be physically comprised of one or more storage devices, such as storage arrays. A storage array is a logical grouping of individual storage devices, such as a hard disk. In certain embodiments, storage 230 is comprised of a JBOD (Just a Bunch of Disks) array or a RAID (Redundant Array of Independent Disks) array. A collection of physical storage arrays may be further combined to form a rank, which dissociates the physical storage from the logical configuration. The storage space in a rank may be allocated into logical volumes, which define the storage location specified in a write/read request.
In one embodiment, by way of example only, the storage system as shown in
The storage controller 240 may include a data duplication module 255, a similarity index module 257 (e.g., a similarity search structure), and a similarity search module 259. The data duplication module 255, the similarity index module 257, and the similarity search module 259 may work in conjunction with each and every component of the storage controller 240, the hosts 210, 220, 225, and storage devices 230. The data duplication module 255, the similarity index module 257, and the similarity search module 259 may be structurally one complete module or may be associated and/or included with other individual modules. The data duplication module 255, the similarity index module 257, and the similarity search module 259 may also be located in the cache 245 or other components.
The storage controller 240 includes a control switch 241 for controlling the fiber channel protocol to the host computers 210, 220, 225, a microprocessor 242 for controlling all the storage controller 240, a nonvolatile control memory 243 for storing a microprogram (operation software) 250 for controlling the operation of storage controller 240, data for control, cache 245 for temporarily storing (buffering) data, and buffers 244 for assisting the cache 245 to read and write data, a control switch 241 for controlling a protocol to control data transfer to or from the storage devices 230, the data duplication module 255, the similarity index module 257, and the similarity search module 259, in which information may be set. Multiple buffers 244 may be implemented with the present invention to assist with the operations as described herein. In one embodiment, the cluster hosts/nodes, 210, 220, 225 and the storage controller 240 are connected through a network adaptor (this could be a fibre channel) 260 as an interface i.e., via at least one switch called “fabric.”
In one embodiment, the host computers or one or more physical or virtual devices, 210, 220, 225 and the storage controller 240 are connected through a network (this could be a fibre channel) 260 as an interface i.e., via at least one switch called “fabric.” In one embodiment, the operation of the system shown in
As mentioned above, in one embodiment, the input data is partitioned into large fixed size chunks (e.g. 16 MB), and a similarity search procedure is applied for each input chunk. A similarity search procedure calculates compact similarity elements, which may also be referred to as distinguishing characteristics (DCs), based on the input chunk of data, and searches for matching similarity elements stored in a compact search structure (i.e. index) in the repository. The size of the similarity elements stored per each chunk of data is typically 32 bytes (where the chunk size is a few megabytes), thus making the search structure storing the similarity elements very compact and simple to maintain and search within.
The similarity elements are calculated by calculating rolling hash values on the chunk's data, namely producing a rolling hash value for each consecutive window of bytes in a byte offset, and then selecting specific hash values and associated positions (not necessarily the exact positions of these hash values) to be the similarity elements of the chunk.
One important aspect and novelty provided by the present invention is that a single linear calculation of rolling hash values, which is a computationally expensive operation, serves as basis for calculating both the similarity elements of a chunk (for a similarity search) and the segmentation of the chunk's data into digest blocks (for finding exact matches). Each rolling hash value is added to the calculation of the similarity elements as well as to the calculation of the digest blocks segmentation. After being added to the two calculations, a rolling hash value can be discarded, as the need to store the rolling hash values is minimized or eliminated. This algorithmic element provides significant efficiency and reduction of CPU consumption, as well as considerable performance improvement.
In one embodiment, the similarity search procedure of the present invention produces two types of output. The first type of output is a set of positions of the most similar reference data in the repository. The second type of output is the digests of the input chunk, comprising of the segmentation to digest blocks and the digest values corresponding to the digest blocks, where the digest values are calculated based on the data of the digest blocks.
In one embodiment, the digests are stored in the repository in a form that corresponds to the digests occurrence in the data. Given a position in the repository and size of a section of data, the location in the repository of the digests corresponding to that interval of data is efficiently determined. The positions produced by the similarity search procedure are then used to lookup the stored digests of the similar reference data, and to load these reference digests into memory. Then, rather than comparing data, the input digests and the loaded reference digests are matched. The matching process is performed by loading the reference digests into a compact search structure of digests in memory, and then for each input digest, querying the search structure of digests for existence of that digest value. Search in the search structure of digests is performed based on the digest values. If a match is found, then the input data associated with that digest is determined to be found in the repository and the position of the input data in the repository is determined based on the reference digest's position in the repository. In this case, the identity between the input data covered by the input digest, and the repository data covered by the matching reference digest, is recorded. If a match is not found then the input data associated with that digest is determined to be not found in the repository, and is recorded as new data. In one embodiment, the similarity search structure is a global search structure of similarity elements, and a memory search structure of digests' is a local search structure of digests in memory. The search in the memory search structure of digests is performed by the digest values.
As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as a system, method or computer program product. Accordingly, aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a “circuit,” “module” or “system.” Furthermore, aspects of the present invention may take the form of a computer program product embodied in one or more computer readable medium(s) having computer readable program code embodied thereon.
Any combination of one or more computer readable medium(s) may be utilized. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that may contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical fiber cable, RF, etc., or any suitable combination of the foregoing. Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).
Aspects of the present invention have been described above with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer readable medium that may direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the computer readable medium produce an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks. The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.
The flowchart and block diagrams in the above figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, may be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
This application is a Continuation of U.S. patent application Ser. No. 13/840,094, filed on Mar. 15, 2013, now U.S. Pat. No. 9,244,937.
| Number | Name | Date | Kind |
|---|---|---|---|
| 7406603 | MacKay et al. | Jul 2008 | B1 |
| 8326824 | Agrawal et al. | Dec 2012 | B2 |
| 8805796 | Hu et al. | Aug 2014 | B1 |
| 8825626 | Wallace et al. | Sep 2014 | B1 |
| 8914338 | Wallace et al. | Dec 2014 | B1 |
| 20090083563 | Murase | Mar 2009 | A1 |
| 20090307251 | Heller et al. | Dec 2009 | A1 |
| 20100094817 | Ben-Shaul et al. | Apr 2010 | A1 |
| 20100106691 | Preslan et al. | Apr 2010 | A1 |
| 20100205163 | Eshghi et al. | Aug 2010 | A1 |
| 20100250480 | Cherkasova et al. | Sep 2010 | A1 |
| 20100318759 | Hamilton et al. | Dec 2010 | A1 |
| 20110060876 | Liu | Mar 2011 | A1 |
| 20110099351 | Condict | Apr 2011 | A1 |
| 20110225385 | Tofano | Sep 2011 | A1 |
| 20110246741 | Raymond et al. | Oct 2011 | A1 |
| 20110307659 | Hans et al. | Dec 2011 | A1 |
| 20120023070 | Prahlad et al. | Jan 2012 | A1 |
| 20120023112 | Levow et al. | Jan 2012 | A1 |
| 20120143715 | Eshghi et al. | Jun 2012 | A1 |
| 20120143835 | Aronovich et al. | Jun 2012 | A1 |
| 20120166401 | Li et al. | Jun 2012 | A1 |
| 20120166448 | Li et al. | Jun 2012 | A1 |
| 20120185612 | Zhang et al. | Jul 2012 | A1 |
| 20120233135 | Tofano | Sep 2012 | A1 |
| 20120290537 | Smith et al. | Nov 2012 | A1 |
| 20130036277 | Szczepkowski et al. | Feb 2013 | A1 |
| 20130060739 | Kalach et al. | Mar 2013 | A1 |
| 20130073529 | Aronovich et al. | Mar 2013 | A1 |
| 20130086009 | Li et al. | Apr 2013 | A1 |
| 20130138620 | Yakushev et al. | May 2013 | A1 |
| 20140115182 | Sabaa et al. | Apr 2014 | A1 |
| 20140188828 | Hirsch et al. | Jul 2014 | A1 |
| 20140222770 | Lad et al. | Aug 2014 | A1 |
| 20140279951 | Akirav et al. | Sep 2014 | A1 |
| 20140279953 | Aronovich | Sep 2014 | A1 |
| 20140279954 | Aronovich | Sep 2014 | A1 |
| Number | Date | Country |
|---|---|---|
| 1962209 | Aug 2008 | EP |
| Entry |
|---|
| Bhagwat et al., “Extreme Binning: Scalable, Parallel De-duplication for Chunk-based File Backup,” Proceedings of the 17th IEEE/ACM International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS '2009), London, UK, Sep. 2009 (9 pages). |
| Anonymous, “Framework for Stream De-duplication using Biased Reservoir-Sampling,” http://priorartdatabase.com/IPCOM/000216344, Mar. 31, 2012 (7 pages). |
| Number | Date | Country | |
|---|---|---|---|
| 20160103868 A1 | Apr 2016 | US |
| Number | Date | Country | |
|---|---|---|---|
| Parent | 13840094 | Mar 2013 | US |
| Child | 14970740 | US |
