DISTRIBUTED FRAGMENTS FILE SYSTEM

Abstract

The present invention relates to a distributed storage scheme, the distributed storage scheme, every file is encrypted, interleaved and fragmented, and the various fragments are stored on different constituent physical file systems.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to a system and method for distributing and storing files.

2. Description of Related Art

Conventional distributed file systems are known for allowing different subsets of files to be on different physical file systems. A given file or a specific replicated instance is stored in its entirety on a single underlying disk or physical file system.

A conventional file system accessed from a personal computer or a workstation appears as a tree of directories, each directory containing additional directories or files. A given file is typically stored entirely on a single hard-disk drive or on a cluster of drives in a scheme called Redundant Array of Independent Disks (RAID). Both types of file systems are prone to data loss and security vulnerabilities due to the following reasons. File data that is not backed up onto a separate drive is lost when the underlying single-drive system fails. RAID systems have the ability to protect against hard-drive failures by virtue of storing data that is replicated and striped over multiple drives, or striped and stored along with error-correction parity bytes. While the approach is more robust than single-drive systems, it has been shown that multiple disk-drive failures tend to be more frequent (somewhat negating the advantage of RAID systems) than would be expected from statistical and probabilistic analysis. The reasons for the higher-than-expected correlation in failure is the identical aging profile and similar environmental factors (such as temperature, humidity vibration and power fluctuations) experienced by the underlying physical drives when clustered in the same facility or cabinet. Traditional RAID systems employ sets of identical disks that are highly synchronized in terms of access times.

When file data is located in a conventional single physical file system, it is more vulnerable to theft as a potential intruder only needs to discover the security vulnerabilities of that particular physical file system.

It is desirable to provide an improved distributed storage scheme having reduced storage requirements while simultaneously improving security and availability by mitigating the impact of accidental loss or theft of one or more disks .

SUMMARY OF THE INVENTION

The present invention relates to a distributed storage scheme, referred to as

Distributed Fragments File System (DFFS). The Distributed Fragments File System (DFFS) uses N (N>=1) conventional physical file systems, such as Network File System (NFS), Internet SCSI (iSCSI), Service Message Block (SMB), or Hypertext Transfer Protocol (HTTP) REST-based Cloud storage, to create a unified and, if desired, physically distributed file system. In one embodiment of the distributed storage scheme, every file is encrypted, interleaved and broken into N fragments, and the various fragments are stored on different constituent physical file systems. This is in contrast to conventional distributed file systems where different groups of files may be on different physical file systems, but a given file or its replicated instance is stored in its entirety on a single underlying disk or physical file system. The distribution of file fragments of the present invention—as opposed to distribution of entire files—provides the following advantages:

1. Intrinsic security: Distributed Fragments File System (DFFS) improves security as a given file is not found in its entirety on any specific physical file system, but is encrypted, interleaved, fragmented, and dispersed across a plurality (N) of file systems.

2. Fault tolerance: The distributed storage scheme supports redundancy in storage that tolerates one or more faults in the constituent physical storage systems. Redundancy is achieved without replicating file data. File access remains uninterrupted and potential failures in the underlying physical storage systems are transparent to the end-user while reducing storage requirements.

3. Reduced storage requirements: Instead of full replication of files that are distributed across N file systems (implying a storage expansion factor of N), Distributed Fragments File System (DFFS) employs a scheme with an N/(N−1) storage expansion factor, where N is the number of constituent physical file systems. Redundancy storage overhead necessary to tolerate, for example, loss of a single physical file system, decreases as the number of physical file systems increases. For example, when N is 3, the storage expansion factor is 1.5. When N is 4, the storage expansion factor is only 1.33. Both those cases compare very favorably with the respective expansion factors of 3 and 4 for conventional storage schemes.

4. Use of disparate file systems with no stringent timing synchronization requirements or equality of lower-level parameters such as block sizes.

The Distributed Fragments File System (DFFS) of the present invention is broadly applicable, but is especially useful when storing data at facilities—such as a public cloud storage provider—that are not controlled or maintained by the creator of the data. Users of the Distributed Fragments File System (DFFS) can utilize multiple public cloud service providers to create a customized and virtualized cloud storage system with the following significant advantages:

No single public cloud storage provider (or physical file system) has the entire file. Only fragments derived by a sequential process involving encryption, interleaving and slicing of a file are stored at each public cloud provider or physical file system. The reverse process applies for recreating files from fragments.

The interleaving and fragmentation can be performed post-encryption. Security is enhanced as theft of a data fragment from any of the underlying file systems is inconsequential. Access to all other fragments residing at other cloud providers or constituent file systems - is essential to reconstruct the original file.

When used in conjunction with the special error correction techniques described later, DFFS does not require replication of data for tolerating failures of the underlying physical file systems or cloud storage providers. This translates into lower expenses compared to traditional replication-based storage systems.

The invention will be more fully described by reference to the following drawings.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a schematic diagram of a Distributed Fragmented File System (DFFS) in accordance with the teachings of the present invention.

FIG. 2 is a flow diagram of a process to create and store a file at a virtual file system controller (VFC).

FIG. 3 is a flow diagram of a process for reading the file from the virtual file system controller (VFC).

FIG. 4 is a schematic diagram of an interleaver.

FIG. 5. is a schematic diagram of another embodiment of a system utilizing DFFS.

DETAILED DESCRIPTION

Reference will now be made in greater detail to an embodiment of the invention, an example of which is illustrated in the accompanying drawings. Wherever possible, the same reference numerals will be used throughout the drawings and the description to refer to the same or like parts.

FIG. 1 is a schematic diagram of a Distributed Fragments File System (DFFS) 10. Distributed Fragments File System (DFFS) 10 aggregates multiple physical file systems (PFS) 14a-14n—some of which may be geographically dispersed—into a virtual file system (VFS) 12.

The following terminology is used:

1. N is the number of underlying physical file systems, each capable of hosting a conventional file system structure comprised of directories and files.

2. Virtual file system (VFS) aggregates constituent physical file systems denoted PFS₀, PFS₁, . . . PFS_N−1.

3. A file in VFS is denoted as F, and it is fragmented and stored as file fragments F0, F1, . . . FN−1 in the underlying physical file systems, PFS₀, PFS₁, . . . PFS_N−1, respectively. Under this scheme, fragment F0 is a valid file under physical file system PFS₀, fragment F1 is a valid file stored at PFS₁, and so on.

4. A virtual file system controller (VFC) is a master controller which maintains and controls the virtual file system (VFS).

5. End-user PCs and workstations communicate with the virtual file system controller (VFC) and mount virtual file system (VFS) using conventional schemes such as Network File System (NFS) and Service Message Block (SMB). The VFC in turn uses protocols specific to each of the physical file systems in a manner transparent to the end-user PCs and workstations. Physical File Systems PFS₀, PFS₁, . . . PFS_N−1 can employ a myriad of different file system protocols such as NFS, iSCSI, SMB, or can use HTTP/HTTPS REST-based cloud storage or other protocol and storage mechanisms known in the art in accordance with the teachings of the present invention.

6. The virtual file system controller (VFC) supports conventional file operations such as open( ), close( ), read( ), write( ), delete( )and stat( ).

7. The virtual file system controller (VFC) also supports directory operations such as mkdir( ), list( ) and rmdir( ).

8. The virtual file system controller (VFC) uses local dedicated storage for its own operations and, optionally, for caching of frequently used files. It relies primarily on physical files systems PFS₀, . . . PFS_N−1 for actual storage.

Physical file systems, PFS₀ 14a and PFS₁ 14b and virtual file controller (VFC) 16 are on Local Area Network (LAN) 13. Physical file systems PFS₂ 14c, . . . , PFS_N−1 14_n−1 are in other geographic areas connected via network 20. Network 20 can be the internet or a private IP network. PFS₀ 14a and PFS₁ 14b are used as examples to show flexibility of placement of physical file systems. It will be appreciated that PFS₀ 14a and PFS₁ 14b do not need to be local. Physical file systems, PFS₀ 14a and PFS₁ 14b can be realized using the following different types of file systems: network-attached file system (NAS), network file system (NFS), service message block (SMB) and Cloud-based Storage. Suitable physical file systems include Netgear NFS.

Virtual file system (VFS) 12 in this embodiment is shown hosting a directory structure comprised of the root directory (/) 18, sub directories D019a and D119b, and files F015a . . . F315d. The example directory structure is not intended to imply limitations in the number of files, directory structure depth, or file sizes. Distributed Fragmented File System (DFFS) 10 scales to support extremely large file systems with directory depth levels and the number of files only limited by the amount of storage available across the constituent physical file systems.

Personal computer (PC) 21 mounts virtual file system (VFS) 12 from virtual file system controller (VFC) 16 using a drive mapping (Y:, for e.g.) 24 and protocol 25. For example, virtual file system controller (VFC) 16 can use one of the standard protocols, such as network file system (NFS) or service message block (SMB) for the drive mapping. Workstation 22 mounts virtual file system (VFS) 12 from virtual file system controller (VFC) using protocol 25. For example, workstation 22 can be a UNIX workstation. Personal computer (PC) 21 and workstation 22 mount root directory structure 18.

Virtual file system controller (VFC) includes local cache 17. Local cache 17 can utilize a Least-Recently-Used (LRU) scheme for management of the overall cache. It will be appreciated that other cache management schemes known in the art can be used in accordance with the teachings of the present invention. Subdirectories 19a,b and files 15a-d are available to both personal computer (PC) and workstation 22. Subdirectories 19a,b and files 15a-d shown for this embodiment are /D0/F0, /D0/F1, /D1/F2 and /D1/F3.

FIG. 2 is a flow diagram of processing steps undertaken by virtual file system controller (VFC) 16 to create and store a file, such as file F015a. In block 50, virtual file system controller (VFC) 16 receives a write request from personal computer 21 or workstation 22 to create file F015a and its associated data within the file. In block 51, virtual file system controller (VFC) 16 stores file F015a in local cache 17 for further processing. In block 52, if configured for encryption, virtual file system controller VFC 16 uses an encryption key, K0, and applies an encryption scheme, such as AES or 3DES, to the contents of file F015a. For example, K0 can be prepared through any number of conventional schemes, such as for example the use of random number generators. Example encryption schemes include advanced encryption standard (AES) and triple data encryption algorithm (3DES).

In block 53, encrypted bytes of file F015a are byte-interleaved and, optionally, subject to Forward Error Correction (FEC) coding as described below to create N equal-sized fragments denoted F0′0, F0′1, . . . F0′N−1. In block 54, each file fragment is optionally encrypted with a separate key.

In block 55, each file fragment is transmitted to its corresponding physical file system (PFS) 14a-14n for storage using its supported file system protocol. File fragment F0′k is stored on PFS_k in the directory specified by the full pathname of the file. In block 56, virtual file system controller VFC 16, optionally, also stores an encrypted version of the original file in local cache 17 for fast response to future read requests from other clients. In block 57, file fragment meta-information about file F015a is stored in local cache 17 of virtual file system controller (VFC) 16. For example, file fragment meta-information can include information about the physical file system (PFS) instances that store the fragments of the file; the order of distribution of fragments across physical file systems; whether FEC was used; access permissions; creation/modification times of the original file, and creation times of the respective fragments. File fragment meta-information can also include the size of the file and the key K0 used for the encryption of the overall file. Storage of other types of information in the meta headers is not precluded.

FIG. 3 is a flow diagram of a process for reading a file 15a-15n from virtual file system controller (VFC) 16. In block 60, virtual file system controller (VFC) 16 receives a read request from one of its client nodes for file F015a. If file F015a is found in local cache 17, virtual file system controller (VFC) 16 responds with a decrypted version of file F015a with a decryption key obtained from a metafile for F0 in block 62. If file F015a is not found in local cache 17, virtual file system controller VFC 16 reads local meta information about file F015a in order to determine the location of file fragments in block 63. In block 64, virtual file system controller (VFC) 16 initiates read requests from physical file systems PFS₀ 14a, . . . PFS_N−1 14n to retrieve file fragments F0′0, F0′1, . . . F0′N−1. If one of the physical file systems PFS₀ 14a, . . . PFS_N−1 14n does not respond then the corresponding fragment file fragments F0′0, F0′1, . . . F0′N−1 is assumed filled with 0s. For example, physical file systems PFS₀ 14a, . . . PFS_N−1 14n may not respond because of loss of network connectivity, failures in its constituent disk drive systems and other reasons.

In block 65, each file fragment F0′0, F0′1, . . . F0′N−1 is separately decrypted, if encrypted during the storage process. In block 66, each file fragment F0′0, F0′1, . . . F0′N−1 is processed through a de-interleaver and, optionally, decoded using FEC techniques as described below. The result is file F0′. In block 67, file F0′ is decrypted using the inverse of the encryption function used during the storage process. The decryption key K0 is obtained from the meta-file associated with file F0′. The virtual file system controller (VFC) responds to the requesting client with the decrypted contents of file F0.

An example interleaver is shown in FIG. 4. Byte interleaver array 70 of R rows 72 and C columns 74. Wherein R is a fixed number of the form 2^m−1, where m is typically [4-7] but higher values are not precluded. C is max (N, (S+R−1)/R) where S is the size of the file. If FEC support is desired for fault tolerance against failures of one or more constituent file systems, then the last K rows of the interleaver array contain encoded bytes that are computed using block FEC encoders such as Reed Solomon (RS), BCH or LDPC. Otherwise the last K rows contain normal file data. The FEC encoding and decoding process along with calculation of K are described below

During interleaving, the bytes of the file are stored column-first in byte interleaver array 70. Bytes are stored in column 0 starting at row 0 and all rows except the last K rows (if FEC is enabled) are filled. If FEC is not used, last K rows contain regular data. Once column 0 is filled, the process is repeated for the remaining C-1 columns. Zero padding is used to fill out a partially filled array.

It will be appreciated that other interleaving and de-interleaving processes known in the art can be used in accordance with the teachings of the present invention. Rows of interleaver array 70 are assigned to N different file fragments in a round-robin fashion. Row q is assigned to fragment number (q mod N). Thus row 0 maps to fragment 0, row N−1 to fragment N−1, row N to fragment 0, row N+1 to fragment 1, and so on. All fragments that map to the same PFS instance are stored in a single file on that instance.

For the de-interleaving process, file bytes received in file fragments are stored row-wise and read out column-wise. A file fragment received from a physical file system (PFS_i) is unpacked and consecutive sets of C (the number of columns in the de-interleaver array) bytes are stored at row indices i, i+N, i+2*N . . . respectively. It will be appreciated that other fragmenting and unpacking of file fragment processes known in the art can be used in accordance with the teachings of the present invention.

In an alternate embodiment, file interleaving can be performed when error correction support is enabled for tolerating failures or loss of access to one of the underlying constituent physical file systems (PFS). When storing the original bytes of the file in interleaver array 70, only K rows of each column are filled. The remainder R-K rows of each column are filled with the parity bytes computed using a suitable erasure code, such as for example Reed Solomon.

K is defined as (N−1)/N*R resulting in R/N parity bytes.

Erasure encoding is done column-wise. The first K rows of every column contain actual file data and provide input of information bytes to the erasure encoding process, which generates R-K parity bytes for every column.

All R rows of interleaver array act 70 act as the source of data put into fragments stored on a plurality of physical file systems (PFS). It will be appreciated that other erasure processes known in the art can be used in accordance with the teachings of the present invention.

File de-interleaving can be performed when error correction support is enabled. File fragments received from various physical file systems (PFS) 14a-14n are unpacked and bytes written to de-interleaver array 70 as described above. In addition, all R rows corresponding to a file fragment that is not received from a physical file system (PFS) 14a-14n are marked for erasure. Once all R rows are filled, erasure decoding is done column-wise using the converse of the erasure encoding process described above. A row marked for erasure results in a single erasure byte in every column. Use of sufficient parity bytes (R-K) per column ensures missing bytes, such as from a fragment not received from a physical file system, are recreated. Only the first K rows of de-interleaver array 70 are used for recreating the original file, i.e., the parity bytes are discarded once the erasure decoding process is run on each column.

In another embodiment shown in FIG. 5, no VFC is required and meta information is stored in small-sized blocks with each of the file fragments. In this embodiment, computer 100 (a general-purpose stored-program computer) is shown hosting a directory structure comprised of the root directory (/) 18, sub directories DO 19a and D119b, and files F015a . . . F315d. Processing blocks of 50-57 of FIGS. 2 and 60-67 of FIG. 3 and interleaver array 70 can be implemented in computer 100.

In accordance with an embodiment of the present invention, a computer-readable medium (e.g., CD-ROM, DVD, computer disk or any other suitable memory device) can be encoded with computer-executable instructions for performing the functions of blocks of 50-57 of FIGS. 2 and 60-67 of FIG. 3 and interleaver array 70.

It is to be understood that the above-described embodiments are illustrative of only a few of the many possible specific embodiments, which can represent applications of the principles of the invention. Numerous and varied other arrangements can be readily devised in accordance with these principles by those skilled in the art without departing from the spirit and scope of the invention.

Claims

1. A method for distributed file storage comprising the steps of: a. fragmenting a file into a plurality of file fragments;b. distributing the plurality of file fragments to one or more physical file systems; andc. storing the distributed plurality of file fragments at the respective one or more physical file systems, each of the physical file systems has one file containing all the respective file fragments generated by step a above.

2. The method of claim 1 further comprising the step of: reading the stored distributed plurality of file fragments; andrecreating said file from the read plurality of file fragments.

3. The method of claim 1 wherein in step a., said file fragments are determined by interleaving and, optionally, use of FEC coding.

4. The method of claim 3 wherein said file fragments are byte interleaved.

5. The method of claim 4 further comprising the step of: reading the stored distributed plurality of file fragments by de-interleaving said plurality of file fragments and recreating said file from the read plurality of file fragments.

6. The method of claim 3 further comprising the step of erasure coding of said file fragments.

7. The method of claim 6 further comprising the steps of: reading the stored distributed plurality of file fragments by de-interleaving said plurality of file fragments;erasure decoding of said de-interleaved plurality of file fragments; andrecreating said file from the decoded plurality of file fragments.

8. The method of claim 1 wherein before step a., further comprising the step of: encrypting said file by applying an encryption scheme to data of said file.

9. The method of claim 8 further comprising the step of: storing the encrypted file.

10. The method of claim 9 wherein in step a. further comprising encrypting each of the plurality of file fragments.

11. The method of claim 10 wherein the file fragments are determined by interleaving and further comprising the step of generating file fragment meta-information comprising: One or more of location information of the physical file system which stored the respective file fragments, an order of said file fragments, a size of the file and a key used in said step of encrypting each of the fragments and storing said fragment meta-information.

12. The method of claim 11 further comprising the step of receiving a file read request; if the encrypted file is found, decrypting the encrypted file with the key of the file fragment metafile-information orif the encrypted file is not found, using the location information and the order of said file fragments of the file fragment metafile-information for retrieving the stored distributed plurality of file fragments from the respective one or more physical file systems, decrypting the retrieved plurality of file fragments with the key of the file fragment metafile-information and recreating said file with said decrypted plurality of file fragments.

13. The method of claim 12 wherein the plurality of file fragments are decrypted using an inverse of an encryption used in the step of encrypting each of the plurality of file fragments.

14. The method of claim 11 further comprising the step of erasure coding of said file fragments.

15. The method of claim 14 further comprising the steps of: reading the stored distributed plurality of file fragments by de-interleaving said plurality of file fragments;erasure decoding of said de-interleaved plurality of file fragments; andrecreating said file from the decoded plurality of file fragments.

16. The method of claim 12 wherein an interleaving array is used in said interleaving step and bytes of said file fragments are stored row-wise and read out column wise.

17. The method of claim 15 wherein an interleaving array is used in said interleaving step and bytes of said file fragments wherein erasure encoding is performed column wise only K rows of each column are filled and the remainder R-K rows of each column are filled with parity bytes computed using an erasure code.

18. The method of claim 1 wherein step a. is performed at a virtual file system controller.

19. The method of claim 1 wherein step a. is performed at a computer.