DATA INTEGRITY FOR DATA STORAGE DEVICES SHARED BY MULTIPLE HOSTS VIA A NETWORK
Access by multiple hosts, such as computers, to a data storage device by way of a network while maintaining data integrity. In one embodiment, a method for accessing the storage device includes acquiring a resource “lock” that provides exclusive access to one of the hosts at a time. In another embodiment, the file systems of a first and second host provide file system attributes stored in a storage device to provide mutually exclusive access for each host to free blocks of the device. In another embodiment, a networked system contains a first host having exclusive direct access to a storage device over a digital network. A second host requiring access to the storage device communicates with the first host by way of the digital network. File access requests generated by the second host are transferred by a redirection filter driver within the second host to the first host.
Latest XiMeta Technology, Inc. Patents:
This application is a divisional patent application of U.S. patent application Ser. No. 10/951,474, filed Sep. 27, 2004 and entitled “Data Integrity for Data Storage Devices Shared by Multiple Hosts Via a Network;” which claims the benefit under 35 U.S.C. § 119(e) to U.S. Provisional Application No. 60/506,829, filed Sep. 26, 2003 and entitled “Device-Level Data Integrity Scheme for Data Devices Shared by Multiple Hosts through LAN;” and which claims the benefit under 35 U.S.C. § 119(e) to U.S. Provisional Application No. 60/590,722, filed Jul. 22, 2004 and entitled “Low-Level Communication Layers and Device Employing Same;” and which claims the benefit under 35 U.S.C. § 119(e) to U.S. Provisional Application No. 60/581,691, filed Jun. 21, 2004 and entitled “File System Features That Enable a Storage Device to Be Shared Directly by Multiple Hosts at Device Level;” the disclosures of which are hereby incorporated herein by reference in their entireties.
BACKGROUND OF THE INVENTIONa. Field of the Invention
The invention relates generally to a data storage device shared by multiple hosts by way of a network. More specifically, the invention relates to efficient access by multiple hosts of a data storage device over a network while maintaining the data integrity of the storage device.
b. Background of the Invention
Generally, hosts, as referred to herein, are electronic devices that employ data storage devices to store digital data for later retrieval and processing by the host. Hosts include, but are not limited to, computers (including desktop personal computers, laptop personal computers and workstations), personal digital assistants (PDAs), digital audio systems, digital television sets, television set-top boxes, digital game devices, smart phones, hand-held computers and other digital data processing devices. Data storage devices include, but are not limited to, hard disk drives, tape drives, flash memory units, and compact disc (CD) and digital versatile disc (DVD) drives. Further, the data written to or read from a storage device may take any of a variety of forms, including, for example, numerical, textual, audio and video data.
Often, data storage devices are not connected directly with a host, but instead communicate with the host via an intermediate electronic device called a device controller, which couples a data storage device with a central processing unit (CPU) and/or logic memory of the host, thus providing a mechanism for transferring data therebetween. Generally, the host also employs its operating system (i.e., the software resident on a host that controls the overall operation of the host) to facilitate communication between the CPU and/or logic memory and the device controller. For example,
Generally, in situations in which a data storage device is to be shared between two or more hosts, the storage device is not shared among the hosts directly at device level. Instead, the storage device often resides within one of the hosts involved, while the remaining hosts communicate with the data storage device by way of the host containing the data device. Typically, communication between the hosts occurs via a network file system. In general, a file system (for example, the NT File System (NTFS) employed by Microsoft Windows®) is the portion of an operating system responsible for the storage and tracking of files, and a file system that cooperates with other file systems over a network is termed a network file system. Typically, maintaining a stable state in such a system requires that all file write operations by one host, including the writing of any file directory information and other file “meta-data,” be allowed to complete prior to allowing access by another host.
However, in such a network a complete computer system or host providing a network file system is required for each data storage device added to the network, thus significantly increasing the cost of the system. More importantly, the multiple hosts possess a file system dependency when sharing data at the file system level, as all of the hosts involved must agree on a particular network file system protocol. Further, when a change in the network file system is required, that change must be implemented in all hosts involved. In addition, the usage of files remotely accessible through network file systems typically is limited compared to what is possible by way of direct access of the files from a data storage device, such as a hard disc drive.
Sharing a storage device directly at the device level through a network provides certain advantages over indirect sharing of the storage device via network file systems. Direct sharing tends to be more efficient in terms of latency and access times. Direct sharing is more cost effective because less expensive hardware, in the form of a network device controller may be used instead of an entire computer system, which allows direct connection of each storage device via a network. No additional operating system or file system software is required, which also eliminates the file system dependency problems and limitations identified above.
However, given that such a system provides no centralized control of data transfers between the hosts and the storage device, data integrity is a potential problem. For example, with each host of the system writing and reading various portions of the storage device, one host may easily overwrite portions of files previously written by another host, thus possibly causing partial or total loss of the data contained in the files.
To further explain, sharing a data storage device over a network presents unique challenges compared to, for example, those involved with sharing a network printer. A network printer is often shared by more than two host computers, but the nature of the data being transferred over the network necessitates the two situations be treated differently.
Print commands from computers to network printers apply only to complete files. As a result, all commands issued to a network printer are guaranteed to be serialized at the file level so that no overlapped or interleaved files may be printed. In other words, a file in the shared network printer environment cannot be divided into smaller portions to be interleaved with portions of other files to be printed.
However, files intended for a data storage device, such as a hard disk drive, are ultimately translated into one or more physical sectors of the data device by way of file system software. Further, no guarantee exists that the file will not occupy several discontinuous series of sectors on the data storage device. Therefore, different files from various hosts sharing the storage device may possibly be mapped onto overlapping sectors unless the file systems of the hosts cooperate in some manner.
Given the foregoing, systems and methods that allow multiple hosts to access a shared data storage device in an efficient manner without loss of data integrity would be advantageous.
BRIEF SUMMARY OF THE INVENTIONGenerally, embodiments of the present invention allow read and/or write access by multiple hosts, such as computers or other information appliances, to a data storage device by way of a network while maintaining the data integrity of the data storage device. In one particular embodiment, a method for accessing the data storage device provides, in part, acquiring a resource lock, which provides exclusive access to one of the multiple hosts at a time. The host holding the lock may then directly access the storage device without interference from the other hosts. After accessing the storage device, the accessing host releases the lock on that storage device so that other hosts may then be allowed to access the storage device. The lock may be implemented entirely in software, hardware, or a combination thereof. In one embodiment, the lock is implemented within the data storage device, and the data storage device accepts and executes lock access commands issued by the multiple hosts.
In another embodiment, a networked system is provided which includes a data storage device and a plurality of hosts coupled to the storage device by way of a digital network. In addition, a resource lock is included which provides exclusive access to the data storage device to one of the plurality of hosts at a time. Digital networks employable for coupling the data storage device with the multiple hosts include, but are not restricted to, a local area network such as Ethernet (LAN), a wide area network (WAN), the Internet, a virtual private network (VPN), and any other digital network.
In another embodiment of the invention, a networked system with a data storage device shared by a plurality of hosts over a network utilizes a file system providing a “check out” attribute for each cluster of free blocks available for file storage. A host attempting to claim a cluster of free blocks analyzes the associated check out attribute to determine if another host has already claimed the cluster. If not, the host sets the check out attribute with a value indicating that it has claimed the cluster, thereby providing the host exclusive access to the cluster. In a further embodiment, a resource lock may be employed to protect the access to the check out attribute. Mutually exclusive access to other file system data structures may be provided in a similar fashion.
In a further embodiment, a first host has exclusive direct access to a data storage device, which is accessed by way of the host's local file system over a digital network. A second host requiring access to the data storage device communicates with the first host by way of the digital network. File access requests generated by the second host are redirected away from its own local file system to the first host by a redirection filter driver. The first host maintains direct access to the storage device while the second host communicates with the device through the first host's file system. In one embodiment, the first and second hosts each include a file network system for transferring file access requests from the second to the first host. In another embodiment, each of the first and second hosts employ a network connection (such as a socket connection program) to allow the second host to issue file access requests through the first host.
Additional embodiments and advantages of the invention will be realized by those skilled in the art upon reading the detailed description of the invention.
One embodiment of the invention allows direct connection of one or more hosts to one or more data storage devices, as illustrated in
Most host computer systems also utilize a data buffer, or “cache,” normally implemented inside the main logical memory of the computer system and logically residing within the data hierarchy between the file system 101 and the input/output system 102. The buffer is usually employed to increase system performance by saving a copy of a portion of often-used (or recently used) data stored on the data storage device 103 for future computer CPU or main memory accesses to shorten the time required to access that data. Due to the limited size of the buffer compared to the data storage device 103, the buffer typically is able to hold only a small percentage of the total amount of data residing in the data storage device 103 at any given time.
In the example of
As illustrated in
In operation 305, with a valid copy 107 of the meta-data now available in the buffer 105, the file system 101 reads the copy 107. Continuing with operation 306, the file system 101 determines whether the requested data access requires a read or write of file data. In the case of a read operation, operation 307 is executed, in which the file system 101 determines the proper blocks 202 of the actual user data desired and attempts to access a valid copy 106 of the user data in the buffer 105. In operation 308, the file system 101 determines if the copy 106 of user data is not valid or is nonexistent in the buffer 105. If the copy 106 of user data is invalid or not present, operation 309 is performed, in which the input/output system 102 converts the location of the blocks 202 holding that data into the corresponding location of physical sectors 203 of the data storage device 103, reads the user data 108 from the physical sectors 203 of the data storage device 103, and copies the user data 108 into the buffer 105, resulting in a valid copy 106 of the user data in the buffer 105. Operation 308 then may be executed once again to ensure the copy 106 of the user data in the buffer 105 is valid. In operation 310, the file system 101 then reads the copy 106 of the user data from the buffer 105, thus completing the read request from the user process 100.
If, instead, the file system 101 determined in operation 306 that a write operation is involved, operation 311 is executed, in which the file system 101 uses the copy 107 of meta-data previously read from the buffer 105 in operation 305 and transfers the user data 106 and associated meta-data 107 to be written to an appropriate location in the buffer 105, thus making those portions of the buffer 105 valid. At some later time, when the buffer 105 is to be “flushed” (i.e., data in the buffer 105 is to be written to the data storage device 103), operation 312 is performed, in which the user data 106 and associated meta-data 107 in the buffer 105 are written to the data storage device 103 as user data 108 and meta-data 109 by way of the input/output system 102, thereby completing the write operation.
As mentioned earlier, allowing multiple hosts concurrent direct access to the data storage device 103 may cause data integrity problems in both the meta-data 109 and the file data 108 located on the data storage device 103, as well as any copies 106, 107 of that data in a buffer 105 of each host. For example, referring to
In addition, the use of a buffer 105 within each of the hosts exacerbates any potential data integrity problems. For example, if each host is accessing copies 106, 107 of meta-data and file data from its own buffer 105, updates to those copies 106, 107 will not be seen by other hosts until that information is flushed from the buffer 105 and written to the data storage device 103. Accordingly, each host may be attempting to update the same data file in different ways, completely unaware that multiple, dissimilar copies of the same file exist, thus destroying the data integrity of that file.
To address this problem, one embodiment of the invention involves the use of a resource “lock” to prevent access to the data storage device 103 at the device level by more than one host at any particular time. Generally speaking, the lock is acquired by a host attempting access the storage device, including any reading or writing of a data file to the device 103, and is released after the access operation has been completed. In most embodiments, completion of a write command would include the host in possession of the lock flushing the contents of its buffer 105, thus ensuring the meta-data and file data of the data storage device 103 has been updated. Only one host may possess the lock at any one time, thereby prohibiting access to the data storage device by any other host. The lock may also be implemented as a “semaphore” or similar construct known in the art. Generally, a semaphore is a flag or similar indicator that is writable and readable by one or more hosts, and is used to relay a simple message between those hosts.
The lock itself may be implemented in several different ways. The lock may be implemented entirely in software (such as device driver or network protocol), although hardware implementations are possible as well, as are hybrid hardware/software implementations. In one embodiment, the data storage device 103 itself may internally store the value of the lock for access by each of the hosts using the device 103. All access and manipulation of the lock by the host would then be controlled, for example, by a device-level controller within or operably connected to the data storage device 103. In addition to lock control, the device-level controller may process standard device-level commands normally targeted for a data storage device, such as the commands associated with the Small Computer Systems Interface (SCSI) or Integrated Drive Electronics (IDE) interfaces known in the art.
Generally, a device-level controller is implemented by way of an embedded microcontroller system designed and employed to perform tasks specific to the control and maintenance of the associated data storage device, including the processing of device-level commands, as described above. Typically, such a system employs resources, in terms of algorithmic capability, speed and logical memory, minimally sufficient to perform its assigned duties in terms of data storage device control, but not to implement commands beyond device control. (For example, the microcontroller may respond to device-level command, but would not implement and run an operating system.) Accordingly, the term “computer,” as is generally employed in the art, does not embrace a device-level controller of the embodiments of the present invention.
In some embodiments, the data storage device 103 (or associated controller) may provide lock control and access for the hosts by way of lock-specific commands that facilitate the checking and setting of the lock. For example, the data storage device 103 may provide an “acquire lock” command that provides both a checking and setting of the lock in one operation. In other words, if the lock is already held by another host, the data storage device 103 will return a failure indication to the requesting host. Otherwise, a success indication can be returned, indicating to the requesting host that the lock acquisition was successful, in which case the requesting host may proceed with accessing data on the data storage device 103. In another implementation, the requesting host may be able to specify a timeout period as a parameter of the acquire lock command so that the data storage device 103 attempts to acquire the lock for the requesting host for the timeout period if the lock is held by another host before returning a failure indication. In addition, the data storage device 103 would also support a “release lock” command to allow the host holding the lock to allow access by the other hosts. In another embodiment, a network device controller 34, as shown in
In yet another implementation, the lock may be implemented by one of the hosts accessing the data storage device 103. Other hosts would then access the lock by making requests via the network to the host implementing the lock. In another embodiment, responsibility for maintaining the lock may be distributed among all or some subset of the hosts involved, with each host possessing a local copy of the state of the lock, including the identity of the host currently in possession of the lock. In that particular case, care must be taken to ensure that each of the hosts maintains a current copy of the state of the lock.
Once the lock is acquired, operation 403 is performed, in which the file system 101 checks for a valid copy 107 of the meta-data in the buffer 105. In some cases, the file system 101 may assume beforehand that the meta-data 107 in the buffer 105 is invalid if the previous lock holder was another host, thereby circumventing an exhaustive check of the buffer 105. In such a case, an identification of the last host to hold the lock may also be implemented in conjunction with the lock itself. For example, a host may write a specific file system data structure on the data storage device 103 after acquiring a lock indicating it is the most recent holder of the lock.
If the meta-data in the buffer 105 is not present or valid, operation 404 is executed, in which the corresponding meta-data 109 is read from the data storage device 103 (i.e., the data storage device 103) using the input/output system 102. Otherwise, the file system 101 may proceed to operation 405. Once operation 404 is complete, the file system 101 may again check if validity of the meta-data in the buffer 105 is valid in operation 403. With a valid copy 107 of the meta-data now in the buffer 105, operation 405 is performed, in which the file system 101 reads that copy 107 of the meta-data, translates the information for the file 201 into a set of blocks 202 containing the desired user data, and then writes the new data into the proper location in the buffer 105. In addition, the meta-data is updated based on any changes necessary due to the new user data to be written.
The file system 101 then determines in operation 406 if the host will hold the lock after the write operation. If not, in operation 408 the host flushes all file data and meta-data in the buffer 105 to the data storage device 103 using the input/output system 102 to perform the logical-to-physical sector translation before it releases the lock in operation 409. If the host holds the lock, the host then determines in operation 407 if another write request is pending. If so, the write procedure resumes by returning to operation 403. If there is no pending write request, the host determines if it will still hold the lock in operation 406 based on other criteria, such as anticipated near-term data storage device access requirements.
As a result, the lock mechanism prevents multiple hosts from simultaneously or concurrently writing to the data storage device 103 by allowing only one host at any time to hold the lock for writing operations. Accordingly, the writing by any host holding a lock is likely valid, because all previous write operations to the shared data storage device 103 by other hosts are fully recorded by flushing all user data 106 and meta-data 107 from the buffer 105 to the data storage device 103 before another host can acquire a lock.
In an alternative embodiment, all hosts connected to the data storage device 103 may cooperate to time-share access to the data storage device 103 by allocating a limited lock hold time to each host, thus preventing monopolization of the lock by any single host. The limited hold time may be invariant (for example, ten seconds per host), prioritized by host (for example, host A has a ten second lock while host B has a five second lock), or varying with some other parameter (for example, larger files may permit longer locks, lock duration may be determined by data throughput speed of the storage device and the host, and so forth).
In one implementation, the reading procedures of
The write and read operations shown in
In another embodiment of the invention, the data integrity of a data storage device shared among multiple hosts over a network at the device level is maintained by a set of features or data structures provided by the file systems of the hosts. These structures provide a mechanism by which free blocks (i.e., blocks not currently written with valid user data or meta-data) of the data storage device are allocated to each host in a mutually exclusive manner. Considering this particular embodiment of the invention involves enhancements to the typical file system currently available, a more detailed discussion of file systems and their operation is provided.
A file system, which typically is the largest unit of data structure identified with a persistent data storage device, normally includes a collection of files, file descriptors, directories and other file-system-related information. The file system is stored on a “logical disk,” which can be a physical disk (e.g., a hard disk drive), a disk partition (i.e., some identified portion of a physical disk), several physical disks, or some other physical manifestation of a data storage device. From the file system's standpoint, a logical disk is composed of a large, one-dimensional array of logical blocks, as described above.
File descriptors 603 (often called “inodes” in UNIX) contain all meta-data associated with a particular file, including the actual block addresses on the data storage device where the file data is located. Other meta-data in the file descriptors 603 normally include the owner of the file, file protection information, time of creation, time of last modification, time of last use, and other information related to the specific file.
In most file system implementations, directories, which essentially are logical collections of files and other directories, are actually implemented as files, so each directory will have a file descriptor in the file descriptor area and occupy some data blocks in the data block area 604. In other words, no special areas of the storage device are reserved within the file system for directories.
Generally, at some point in time, some plurality of the data blocks on a data storage device are allocated or occupied by file data and file descriptors, while some blocks remain unallocated, or “free.” All allocated blocks are linked to a file descriptor so that file data can be traced by starting with the corresponding file descriptor. As files grow, more free blocks are allocated to the files as necessary. Accordingly, file systems typically maintain a list of free blocks for quick allocation of those blocks to files.
Different file systems may also define varying structures containing the information required to implement the file system. For example,
Traditional file systems are designed such that all free blocks are managed by a single host because the data storage device is normally attached to the inner system bus of the host. As a result, only the single host may directly access the device. However, if two different hosts are able to access the device directly at the device level, as would be the case when a storage device is coupled with the hosts directly via a network, each host may allocate logical blocks from the same free block list independently, thus potentially allocating identical blocks to different files. This, in turn, would corrupt the consistency and integrity of the entire file system.
For example, some distributed file systems, such as xFS (“x” File System) and GFS (Global File System) utilize a server for managing the meta-data of the file system, including the free block list. Each client host consults the server before the client acquires free blocks for writing a file. The server allocates free blocks, but does not control actual data written to the storage device. This mechanism allows the clients to write and read data blocks onto and from the data storage device directly without relaying the user data to and from the server. However, this mechanism does not allow the hosts to share the storage device directly at the device level without server intervention. Thus, since only the server controls the allocation of free blocks, and each client host must consult the server to obtain free blocks, the file system cannot be corrupted on the basis of allocating free blocks to multiple clients. However, such a mechanism suffers from scalability and performance overhead limitations since a single server intervenes in all free block allocation to provide proper meta-data management and cache coherency. Thus, computer networks employing a distributed file system generally lack the performance associated with a network in which multiple computers access a shared data storage device directly at the device level without the assistance of a file server.
In order for multiple hosts sharing the same storage device to maintain free blocks of the storage device in a manner such that no block is allocated to more than one particular file at a time, the total number of free blocks may be divided into multiple, mutually exclusive sets of free blocks, with each set being accessible to only one particular host at a time. For example,
In further reference to
Any checked out blocks not ultimately allocated to a file are subsequently returned by the accessing host to the free block list. To return the unallocated free blocks, the host inserts new nodes of free block clusters into the free block list and leaves the check out attribute 624, 625, 626, 627 of the newly inserted cluster nodes blank.
In addition to setting the check out attribute, the accessing host also sets a timestamp attribute 628, 629, 630, 631 with a value indicating when the free blocks were checked out. The timestamp 628, 629, 630, 631 is employed to prevent a host from holding unallocated free blocks indefinitely. This situation can occur, for example, when a host has checked out one or more clusters of free blocks and then becomes inoperative, or “crashes,” before the host has the opportunity to return the unallocated blocks to the free block list. If the timestamp 628, 629, 630, 631 is older than some predetermined value, other hosts may then claim the unallocated blocks from the outdated cluster of free blocks using the standard check out procedure.
In the specific example of
The file system stores within its file system descriptor the location of the file descriptor of the root directory. As the root directory and subdirectories accumulate files and other directories, links are provided within each directory pointing to block locations in the file data block area where the files and file descriptors associated with the directory are stored. Therefore, the file system can trace the entirety of the directory structure starting from the root directory.
In one embodiment of the invention, a separate set of file descriptors for the root directory are provided for each host that shares the device, in contrast to the single set of file descriptors normally employed.
When a host accesses the file system to view the directory structure, the host peruses the entire directory structure by following the links to the corresponding file descriptors and data starting from the complete set of root directory file descriptors 654, 656, 658 allocated to the hosts. In the particular example of
In further reference to the exemplary file directory structure depicted in
As is evident from the foregoing discussion, any host can read directory structures created by other hosts in order to obtain a complete view of the entire directory structure by following the links starting from the root directory file descriptors allocated for the hosts involved. As a result, any host can create its own files under directories created by other hosts.
Further, if the links starting from one of the root directory file descriptors identified with a particular host (for example, the root directory file descriptor for Host 1) are followed, the directory structure created by that host may be retrieved. In other words, each of the directory structures created by each host can be retrieved by following the links starting from the root directory file descriptor allocated for the corresponding host.
Based on the foregoing, the directory structure created by each host constitutes a portion of the total directory structure of the entire file system. The total directory structure may thus be obtained by superimposing the partial directory structures created by each of the hosts. For example,
The file system maintains attributes of a file, as well as the file data itself, to represent related information necessary for management of the file. One such attribute in one embodiment indicates ownership of the file. Because embodiments of the invention allow multiple independent hosts to share the same device directly at the device level, the file system may maintain an “ownership” attribute within the meta-data of each file to distinguish which host maintains ownership of the file. This host identification prevents a host from exercising impertinent access rights to files owned by other hosts.
Another file attribute in another embodiment of the invention is a “check out” attribute of a file, which is distinguished from the “check out” attribute associated with each free block cluster, described in detail above. When a host having write permission for a particular file accesses that file, the file system marks the check out attribute of the file with an identification of the accessing host. At that point, other hosts may not check out the file with write permission. This mechanism prevents more than one host from writing the same file at the same time, which would likely corrupt the data in the file. However, in some embodiments hosts may read a file that is currently checked out by another host holding write permission.
In one embodiment, the file system may require a host to acquire a resource lock before it can check out the file in order to prevent race conditions created by multiple hosts vying to check out the same file. Alternatively, if the data integrity level of a file is relaxed, use of a resource lock can be avoided. For example, some video and audio data files may remain viable even if the data integrity of the files has been compromised to a degree.
Conventional file systems typically maintain in-memory data structures, instead of on-disk data structures (such as the file check out attribute described above), for managing the consistency of files opened by system processes of the host. Storing such data structures in volatile memory, like many forms of random access memory (RAM) (or other non-persistent storage devices), may be appropriate in an environment in which a single host possesses exclusive control of the storage device. In the embodiments described herein, however, multiple hosts may share control of the data storage device. Therefore, data structures relevant to file consistency management that are maintained only within the volatile memories of each host may have limited utility where multiple hosts share the same storage device directly at the device level unless the hosts share the in-memory data structures spread over multiple hosts.
A file system generally maintains files containing information for effective handling of the blocks of the storage device on which the file system is implemented. These files are termed “file system files.” Typically, two of these files are the free block file and the bad block file. The free block file contains a list of blocks that are free, and possibly a complementary list of blocks that are written. The bad block file contains a list of “bad,” or defective, blocks on which no data can be written to or read from correctly, generally due to defective portions of the recordable medium within the data storage device. Conventional file systems have a single host manage those particular file system files.
In various embodiments of the present invention, each host of a multiple-host system accesses the free block file in a mutually exclusive manner by way of the cluster check out attribute, as described in detail earlier. In addition, resource lock acquisition may be required prior to setting the check out attribute in some embodiments, thereby providing additional data integrity for the free block file. Access to the bad block file may be regulated in a similar fashion so that hosts may add bad blocks to the bad block list in a secure manner as they are encountered in the course of disk operation. Further, the same control mechanism may be applied to maintain the data integrity of any file system files.
Ordinarily, the use of multiple hosts sharing a single data storage device would produce the possibility of file name conflicts among files created by the multiple hosts. For example, if the hosts run the same operating system and use default swap files set automatically by the operating system, two or more swap files associated with different hosts could have the same file name under the same directory. However, in embodiments of the invention, the host ownership attribute (described above) may be utilized to distinguish such files.
The file system features presented herein may also be implemented in a system employing multiple shared storage devices.
The controller logic 674 may be implemented as hardware, software, or some combination thereof. In one embodiment, a combination of hardware and software could be employed to process device-level commands for each of the data storage devices 671, 672, 673, such as the device-level controller discussed earlier. (For example, the aforementioned device-level controller may implement, or serve as an example of, the controller logic 674.) Alternatively, the controller logic 674 may be a software driver executed by a microcontroller system capable of transforming and/or relaying device-level commands received from a host to one of the data storage devices 671, 672, 673. In that case, the driver may determine which of the three storage devices 671, 672, 673 is the target of the command received and performs any block location translation necessary from the addressing scheme utilized by the host to the scheme employed by the particular storage device 671, 672, 673. The driver would then relay the modified command to the appropriate target storage device 671, 672, 673, which is capable of processing the modified command itself.
In alternative embodiments of the invention, data integrity of a system of multiple hosts sharing a data storage device can be maintained by utilizing file systems already existing in current hosts. More specifically, instead of allowing all hosts to share a data storage device directly at the device level over a network, one host may have access to the data storage device at device level over a network, while all other hosts are allowed indirect access to the data storage device via their network file systems.
In one embodiment using the structure described in
Sharing the data storage device 742 between the first and second hosts 701, 702 in this manner provides advantages over a system which employs indirect sharing of a data storage device strictly at the network file system level, as described earlier. Although the embodiment of
Similarly, when the network file system 708 of the first host 701 later receives a reply from the local file system 706 to be transferred to the network file system 707 of the second host 702, the network file system 707 of the second host 702 directs the reply to the requesting user process through the redirection filter driver software 700 as though the reply were received from the functional local file system 705.
If no network file system is available or desirable between a host that physically mounts a data storage device (e.g., the first host 701 of
When the network connection 710 of the first host 703 then receives a reply from the local file system 706 to be transferred to its counterpart network connection 709 of the second host 704, the network connection 709 of the second host 704 directs the reply to the requesting user process through the redirection filter driver software 711 as if the reply were received from the functional local file system 712. The network connections 709, 710 can be any program that transfers requests and replies therebetween, such as a conventional socket program.
Certain features of the invention described herein may be implemented as an additional layer in or atop the file and/or operating systems. For example, the aforementioned filter driver software may be added to an existing file system without requiring modification of the file system. Similarly, certain features of the invention may be implemented as an additional attribute of a storage device, storage device controller, or storage device file system/structure. For example, the aforementioned check out attribute may augment a file system or structure to provide added functionality. The basic file system/structure may remain relatively unchanged. In other words, the basic functionality and features of the core file system, structure, operating system, and so forth remain unchanged by the invention, which provides added functionality.
Disclosed herein are several embodiments of systems and methods for ensuring the data integrity of a networked data storage device that is shared among a plurality of hosts. While these embodiments are described in specific terms, other embodiments encompassing principles of the invention are also possible. For example, various features of one embodiment may be combined with features of other embodiments to create a new embodiment not specifically discussed herein. Thus, the scope of the invention is not to be limited to the disclosed embodiments, but is determined by the following claims.
Claims
1. A method for accessing a data storage device shared by a plurality of hosts via a network while preserving the data integrity of the data storage device, comprising:
- acquiring a resource lock, the resource lock providing exclusive access to the data storage device to one of the plurality of hosts;
- directly accessing the data storage device; and
- releasing the resource lock so that another of the plurality of hosts may access the data storage device.
2. The method of claim 1, further comprising:
- prior to releasing the lock, determining if the data storage device is to be accessed again; and
- if the data storage device is to be accessed again, accessing the data storage device.
3. The method of claim 1, the accessing step comprising writing user data to the data storage device.
4. The method of claim 3, the writing step comprising:
- determining if meta-data associated with the user data is available and valid within a data buffer;
- if the meta-data is unavailable and invalid in the data buffer, reading the meta-data from the data storage device into the data buffer;
- reading the meta-data from the data buffer;
- determining the proper location on the data storage device for the user data from the meta-data;
- updating the meta-data based on the user data;
- writing the meta-data and the user data to the data buffer; and
- flushing the meta-data and the user data from the data buffer to the data storage device.
5. The method of claim 1, the accessing step comprising reading user data from the data storage device.
6. The method of claim 5, the reading step comprising:
- determining if meta-data associated with the user data is available and valid within a data buffer;
- if the meta-data is not available and valid in the data buffer, reading the meta-data from the data storage device into the data buffer;
- reading the meta-data from the data buffer;
- determining from the meta-data if the user data is available and valid in the data buffer;
- if the user data is not available and valid in the data buffer, reading the user data from the data storage device into the data buffer; and
- reading the user data from the data buffer.
7. The method of claim 1, wherein the resource lock is acquired and released by way of software commands issued over the network to the data storage device.
8. The method of claim 1, wherein the resource lock is implemented entirely in software.
9. The method of claim 1, wherein the resource lock may only be held for a predetermined period of time before being released.
10. A device implementing the method of claim 1.
11. The device of claim 10, comprising a personal computer.
12. The device of claim 10, comprising a workstation.
13. The device of claim 10, comprising a personal digital assistant (PDA).
14. A digital storage medium storing a software program implementing a method for accessing a data storage device shared by a plurality of hosts via a network, the method comprising:
- acquiring a resource lock, the resource lock providing exclusive access to the data storage device to one of the plurality of hosts;
- accessing the data storage device; and
- releasing the resource lock so that another of the plurality of hosts may access the data storage device.
15. The digital storage medium of claim 14, the method further comprising:
- prior to releasing the lock, determining if the data storage device is to be accessed again; and
- if the data storage device is to be accessed again, accessing the data storage device.
16. The digital storage medium of claim 14, the accessing step comprising writing user data to the data storage device.
17. The digital storage medium of claim 16, the writing step comprising:
- determining if meta-data associated with the user data is available and valid within a data buffer;
- if the meta-data is unavailable and invalid in the data buffer, reading the meta-data from the data storage device into the data buffer;
- reading the meta-data from the data buffer;
- determining the proper location on the data storage device for the user data from the meta-data;
- updating the meta-data based on the user data;
- writing the meta-data and the user data to the data buffer; and
- flushing the meta-data and the user data from the data buffer to the data storage device.
18. The digital storage medium of claim 14, the accessing step comprising reading user data from the data storage device.
19. The digital storage medium of claim 18, the reading step comprising:
- determining if meta-data associated with the user data is available and valid within a data buffer;
- if the meta-data is not available and valid in the data buffer, reading the meta-data from the data storage device into the data buffer;
- reading the meta-data from the data buffer;
- determining from the meta-data if the user data is available and valid in the data buffer;
- if the user data is not available and valid in the data buffer, reading the user data from the data storage device into the data buffer; and
- reading the user data from the data buffer.
20. The digital storage medium of claim 14, wherein the resource lock is acquired and released by way of software commands issued over the network to the data storage device.
21. The digital storage medium of claim 14, wherein the resource lock may only be held for a predetermined period of time before being released.
22. A method for claiming a cluster of free blocks of a data storage device shared by a plurality of hosts via a network while preserving the data integrity of the data storage device, comprising:
- reading a check out attribute associated with the cluster of free blocks to determine if the cluster of free blocks has been claimed by one of the plurality of hosts; and
- if the cluster of free blocks has not been claimed by one of the plurality of hosts, setting the check out attribute with a host identification, thereby claiming the cluster of free blocks.
23. The method of claim 22, further comprising:
- setting a timestamp attribute with an indication of the time at which the check out attribute was set with the host identification.
24. The method of claim 22, further comprising:
- acquiring a resource lock prior to reading the check out attribute, the resource lock providing exclusive access by one of the plurality of hosts to the check out attribute; and
- releasing the resource lock after accessing the check out attribute so that another of the plurality of hosts may access the check out attribute.
25. A method for accessing a bad block list of a data storage device shared by a plurality of hosts via a network while preserving the data integrity of the data storage device, comprising:
- reading a check out attribute associated with the bad block list to determine if the bad block list is being accessed by one of the plurality of hosts; and
- if the bad block list is not being accessed by one of the plurality of hosts, setting the check out attribute with a host identification, thereby gaining access to the bad block list.
26. The method of claim 25, further comprising:
- setting a timestamp attribute with an indication of the time at which the check out attribute was set with the host identification.
27. The method of claim 25, further comprising:
- acquiring a resource lock prior to reading the check out attribute, the resource lock providing exclusive access by one of the plurality of hosts to the check out attribute; and
- releasing the resource lock after accessing the check out attribute so that another of the plurality of hosts may access the check out attribute.
28. A method for accessing a file system file imparting information regarding a file system of a data storage device shared by a plurality of hosts via a network while preserving the data integrity of the data storage device, comprising:
- reading a check out attribute associated with the file system file to determine if the file system file is being accessed by one of the plurality of hosts; and
- if the file system file is not being accessed by one of the plurality of hosts, setting the check out attribute with a host identification, thereby gaining access to the file system file.
29. The method of claim 28, further comprising:
- setting a timestamp attribute with an indication of the time at which the check out attribute was set with the host identification.
30. The method of claim 28, further comprising:
- acquiring a resource lock prior to reading the check out attribute, the resource lock providing exclusive access by one of the plurality of hosts to the check out attribute; and
- releasing the resource lock after accessing the check out attribute so that another of the plurality of hosts may access the check out attribute.
31. A method for traversing the entire file structure of a file system of a data storage device shared by a plurality of hosts via a network, comprising:
- reading a file system descriptor of the file system from the data storage device;
- following a link within the file system descriptor associated with a root directory file descriptor for each of the plurality of hosts; and
- traversing a file structure for each of the plurality of hosts, beginning with the root directory file descriptor for each of the plurality of hosts.
32. A method for accessing a user file of a data storage device shared by a plurality of hosts via a network while preserving the data integrity of the data storage device, comprising:
- reading a check out attribute associated with the user file to determine if the user file is being accessed by one of the plurality of hosts; and
- if the user file is not being accessed by one of the plurality of hosts, setting the check out attribute with a host identification, thereby gaining access to the user file.
33. The method of claim 32, further comprising:
- setting a timestamp attribute with an indication of the time at which the check out attribute was set with the host identification.
34. The method of claim 32, further comprising:
- acquiring a resource lock prior to reading the check out attribute, the resource lock providing exclusive access by one of the plurality of hosts to the check out attribute; and
- releasing the resource lock after accessing the check out attribute so that another of the plurality of hosts may access the check out attribute.
35. A file system data structure providing data integrity to a data storage device shared by a plurality of hosts, the file system data structure comprising:
- a free block list comprising an entry indicating the location of a cluster of free blocks of the data storage device, the free block list residing in at least one block of the data storage device;
- a check out attribute of the entry of the free block list indicating whether the cluster of free blocks has been claimed by one of the plurality of hosts and, if so, the identity of the one of the plurality of hosts;
- wherein the cluster of free blocks may thus be allocated by the one of the plurality of hosts.
36. The file system data structure of claim 35, further comprising:
- a resource lock of the check out attribute of the entry of the free block list allowing only one of the plurality of hosts at a time to access the check out attribute of the entry of the free block list.
37. The file system data structure of claim 35, further comprising:
- a timestamp attribute associated with an entry of the free block list;
- wherein if the check out attribute of the entry of the free block list is set with an identity of one of the plurality of hosts, the timestamp attribute indicates when the check out attribute of the entry of the free block list was set with the identity of the one of the plurality of hosts.
38. The file system data structure of claim 35, further comprising:
- a bad block list comprising an entry indicating the location of a defective block of the data storage device, the bad block list residing in at least one block of the data storage device;
- a check out attribute of the bad block list indicating whether access to the bad block list has been claimed by one of the plurality of hosts and, if so, the identity of the one of the plurality of hosts;
- wherein the bad block list may thus be accessed by the one of the plurality of hosts.
39. The file system data structure of claim 38, further comprising:
- a resource lock of the check out attribute of the bad block list allowing only one of the plurality of hosts at a time to access the check out attribute of the bad block list.
40. The file system data structure of claim 38, further comprising:
- a timestamp attribute associated with the bad block list;
- wherein if the check out attribute of the bad block list is set with an identity of one of the plurality of hosts, the timestamp attribute indicates when the check out attribute of the bad block list was set with the identity of the one of the plurality of hosts.
41. The file system data structure of claim 35, further comprising:
- a file system file imparting information regarding a file system on the data storage device;
- a check out attribute of the file system file indicating whether access to the file system file has been claimed by one of the plurality of hosts and, if so, the identity of the one of the plurality of hosts;
- wherein the file system file may thus be accessed by the one of the plurality of hosts.
42. The file system data structure of claim 41, further comprising:
- a resource lock of the check out attribute of the file system file allowing only one of the plurality of hosts at a time to access the file system file.
43. The file system data structure of claim 41, further comprising:
- a timestamp attribute associated with the file system file;
- wherein if the check out attribute of the file system file is set with an identity of one of the plurality of hosts, the timestamp attribute indicates when the check out attribute of the file system file was set with the identity of the one of the plurality of hosts.
44. The file system data structure of claim 35, further comprising:
- a separate set of file descriptors for a root directory provided for each of the plurality of hosts.
45. The file system data structure of claim 44, wherein each set of file descriptors may be read by each of the plurality of hosts.
46. The file system data structure of claim 44, wherein each of the plurality of hosts may create files and directories under any of the root directories associated with each set of file descriptors.
47. The file system data structure of claim 44, wherein an entire directory structure of the data storage device may be produced by combining all of the sets of file descriptors for each root directory.
48. The file system data structure of claim 35, further comprising:
- an ownership file attribute associated with a file, the ownership file attribute indicating which of the plurality of hosts owns the file.
49. The file system data structure of claim 35, further comprising:
- a check out file attribute associated with a file, the check out file attribute indicating which of the plurality of hosts has exclusive write access to the file.
50. The file system data structure of claim 49, further comprising:
- a resource lock allowing only one of the plurality of hosts at a time to access the check out file attribute associated with the file.
51. A networked system, comprising:
- a plurality of hosts;
- a network coupling the plurality of hosts;
- a first data storage device coupled with, and shared at the device level by, the plurality of hosts over the network, the first data storage device comprising the file system data structure of claim 35.
52. The networked system of claim 51, further comprising:
- a controller logic coupling the first data storage device with the network; and
- a second data storage device coupled with the controller logic;
- wherein the controller logic presents a view of a single data storage device to the plurality of hosts.
Type: Application
Filed: Oct 21, 2008
Publication Date: Feb 12, 2009
Applicant: XiMeta Technology, Inc. (Road Town)
Inventor: Han-gyoo Kim (Irvine, CA)
Application Number: 12/255,433
International Classification: G06F 12/02 (20060101);