Disconnected operation for systems utilizing cloud storage

Abstract

While connected to cloud storage, a computing device writes data and metadata to the cloud storage, indicates success of the write to an application of the computing device, and, after indicating success to the application, writes the data and metadata to local storage of the computing device. The data and metadata may be written to different areas of the local storage. The computing device may also determine that it has recovered from a crash or has connected to the cloud storage after operating disconnected and reconcile the local storage with the cloud storage. The reconciliation may be based at least on a comparison of the metadata stored in the area of the local storage with metadata received from the cloud storage. The cloud storage may store each item of data contiguously with its metadata as an expanded block.

Description

BACKGROUND

Increasingly, data is stored on and retrieved from the cloud. Cloud computing enables storage and retrieval operations to be spread across a large number of devices. By involving a large number of devices in a given read from or write to the cloud, the cloud may even offer faster storage and retrieval to a computing device than is available on local storage of that computing device.

To utilize cloud storage, a computing device must be connected to the cloud, typically via a network. While connected, the computing device reads data from the cloud and writes data to the cloud. This dependence on connectivity to the cloud poses problems when the computing device is not connected to the cloud. One approach to disconnected operation is for the computing device to locally store writes while disconnected from the cloud and provide those writes to the cloud upon reconnecting. This approach provides no ability to read data that was stored on the cloud during disconnected operation, however. Another approach to disconnected operation is for the computing device to locally mirror writes to the cloud while connected and locally store writes while disconnected. This approach enables reads during disconnected operation but to make the techniques resilient to local mirror crashes, clients need to compare the entirety of the local mirror contents to the corresponding data in the cloud. Further harming performance, locally mirroring every write in a synchronous fashion degrades performance if writing to the mirror is slower than writing to the cloud, causing applications to wait longer for indications of the success or failure of their writes.

SUMMARY

While connected to cloud storage, a computing device writes data and metadata to the cloud storage and asynchronously mirrors the data and the metadata to local storage of the computing device. By asynchronously mirroring the write, the application requesting the write is enabled to continue processing without awaiting a result of the mirrored write to the local storage. When mirroring the write to the local storage, the computing device writes the data and metadata to separate areas of the local storage. During reconciliation of the local storage with the cloud storage following a crash or disconnected mode, the computing device only needs to examine the area of local storage that includes the metadata and compare the metadata from that area to metadata retrieved from the cloud storage. Because the cloud storage may be significantly faster than the local storage, the cloud storage may store the data and metadata differently than the local client stores those items. For example, the client may keep data and metadata in two different areas of the local storage, while the cloud storage may store expanded blocks that each include a data item and its associated metadata in contiguous storage. These different techniques for storing data and metadata in the local and cloud storages improve the speed of reconciling the local and cloud storages.

This Summary is 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 or essential features of the claimed subject matter; nor is it to be used for determining or limiting the scope of the claimed subject matter.

BRIEF DESCRIPTION OF THE DRAWINGS

The detailed description is set forth with reference to the accompanying figures. In the figures, the left-most digit(s) of a reference number identifies the figure in which the reference number first appears. The use of the same reference numbers in different figures indicates similar or identical items or features.

FIG. 1 illustrates an example computing device and cloud storage devices configured to store metadata and data in different manners to improve the speed of reconciling cloud and local storage. The computing device is also configured to write to the cloud storage devices and asynchronously mirror writes to local storage of the computing device.

FIG. 2 illustrates an example process for writing to cloud storage and asynchronously mirroring writes to local storage of the computing device.

FIG. 3 illustrates an example process for reconciling local storage of a computing device with cloud storage following a crash of the computing device.

FIG. 4 illustrates an example process for disconnected operation of a computing device and reconciliation of local storage of the computing device to cloud storage upon reconnecting to the cloud storage.

FIG. 5 illustrates an example process for reconciling local storage of a computing device with cloud storage following disconnected operation during which the computing device crashed.

DETAILED DESCRIPTION

This disclosure describes, in part, techniques for enabling continued operation during network outages and improved reconciliation between a computing device and the cloud following a crash or disconnected operation. In order for the computing device to satisfy reads when the network is down, a cloud client of the computing device writes to cloud storage and asynchronously mirrors those writes to local storage of the computing device. As used herein, the term “mirror” refers to writing substantially the same data metadata to two different storage components. The data and metadata may, however, be stored differently in the different storage components. The term “asynchronously mirror” refers to providing an indication to the application which generated the write that the write was successful without awaiting a result of mirroring the write to the local storage. From the perspective of the application that generated the write, the write operation is finished when the cloud acknowledges that the write was successful. The application does not have to wait for the local data and metadata writes to complete, because the cloud client handles those events asynchronously, without blocking the application.

In various embodiments, when the cloud client mirrors the write to the local storage, it writes data and metadata associated with that data to different areas of the local storage. For example the cloud client may write multiple data items to a first area of the local storage and multiple corresponding metadata records to a separate area of the local storage. Because the metadata is much smaller than its associated data, the second area may be much smaller in size than the first area. In contrast to this, the cloud storage may store each data item contiguously with its corresponding metadata record as an expanded block. The cloud storage may store the data and metadata in this manner to minimize the number of seeks performed on the cloud storage if the cloud storage uses a seek-based storage medium like a hard disk.

Storing metadata records in a separate area of local storage and in expanded blocks of cloud storage improves the speed of reconciliation between the local storage and cloud storage. After a crash or following disconnected operation, the cloud client need only use the much smaller second area of the local storage rather than the entire local storage. This may be advantageous if the local storage is slow relative to the cloud storage. This area of metadata records is compared to metadata retrieved from the cloud storage. Based on the result of the comparison, the cloud client then reconciles the local storage and cloud storage.

In various embodiments, the cloud client ensures the integrity of the data mirrored to the local storage by writing the data before the metadata in connected mode, and writing the metadata before the data in disconnected mode. For example, the cloud client operating in connected mode may write a data item and its metadata record to the cloud, mirror the write of the data item to local storage, and crash. The cloud storage now has the updated metadata record and the local storage has either a previous metadata record or a default record. The cloud client will reconcile the local storage and cloud storage, comparing version identifiers in the metadata records which will indicate that the cloud storage has a more recent version of the data item (note that this is incorrect but harmless; the local storage has updated data, but not updated metadata, and the updated data of the local storage will be overwritten with the same updated data stored in the cloud). Based on the comparison, the cloud client will update the local storage with the data item and metadata record from the cloud storage. If the initial writes of metadata and data to the local storage were reversed, and the computing device crashed after writing the metadata record to local storage but before writing the data item, the reconciliation would have wrongly indicated that the local storage has the same version of the data item as the cloud storage.

The order of writing data and metadata to the local storage is reversed in disconnected mode because, in that mode, the local storage has the most recent version of the data. For example, a computing device operating in disconnected mode may write a metadata record for a data item to local storage and, before writing the data item to local storage, crash and then immediately reconnect to the cloud storage. Upon entering connected mode, the cloud client may reconcile the local storage with the cloud storage. In this example the reconciliation would indicate (incorrectly, but harmlessly) that the local storage has the more recent version of the data item, resulting in the data item being used to update the data item stored in cloud storage. If the initial writes of metadata and data to the local storage were reversed, and the computing device wrote the data but not the updated metadata record to local storage, the reconciliation would have wrongly indicated that the local storage has the same version of the data item as the cloud storage, resulting in an inconsistency between the cloud storage and local storage.

Example Devices

FIG. 1 illustrates an example computing device and cloud storage devices configured to store metadata and corresponding data in different manners to improve the speed of reconciling cloud and local storage. The computing device is also configured to asynchronously mirror writes to the cloud storage devices to local storage of the computing device. As illustrated, a computing device 102 is connected to cloud storage devices by a network 106. The computing device includes a memory 108, which stores applications 110, a file system 112, a virtual disk driver 114, and a cloud client 116. The cloud client 116 writes data items and metadata records to cloud storage 118 of the cloud storage devices 104, which store the data items and metadata records as expanded blocks 120. The cloud client 116 then asynchronously mirrors the write to local storage 122 of the computing device, writing the data items to a first area 124 of the local storage 122 and the metadata records to a second area 126 of the local storage 122. To operate the cloud client 116 and other modules 110-114, the computing device 102 includes a processor 128, and to communicate with the cloud storage devices 104 via the network 106, the computing device 102 includes network interfaces 130. Also, the computing device 102 may include removable storage 132, non-removable storage 134, and input/output (I/O) devices 136.

The computing device 102 and the cloud storage devices 104 may each be implemented as any one or more computing devices, such as personal computers (PCs), laptop computers, workstations, server systems, mainframes, server farms, cellular phones, smart phones, tablet computers, media centers, media devices, or any other computing devices. Also, modules and data of the each of the computing device 102 and cloud storage devices 104 may be implemented in a single computing device or distributed among multiple computing devices. In some embodiments, one or more of the computing device 102 and cloud storage devices 104 may be implemented as a virtual machine on a computing device.

In some embodiments, the cloud storage devices 104 form a datacenter-scale store which leverages a full bisection bandwidth network that matches cloud storage disk bandwidth to network interface card (NIC) bandwidth, enabling the cloud storage devices 104 to exchange data at full NIC speeds. In further embodiments, the cloud storage devices 104 store data in units referred to as “tracts” which may have a uniform size (e.g., 8 MB). A plurality of these tracts may form a binary large object (BLOB) or other data container and may be distributed across a plurality of the cloud storage devices 104. In one embodiment, the virtual disk driver 114 may be mapped to a BLOB or other data container by the cloud storage devices 104. An example set of cloud storage devices 104 is described in greater detail in U.S. Pat. No. 8,438,244, entitled “Bandwidth-Proportioned Datacenters” and issued on May 7, 2013.

The network 106 may be or include a public or private packet switched network, such as the Internet, circuit switched network, or combination of packet switched and circuit switched networks. The network 106 may include a plurality of computing devices connected, for example, by one or more wide area networks (WAN), one or more local area networks (LAN), and/or one or more personal area networks (PAN). Communication between these ones of these computing devices of the network 106 may be wired, wireless, or both. These communications may utilize any sort of communication protocol known in the art for sending and receiving messages, such as the Transmission Control Protocol/Internet Protocol (TCP/IP), the Hypertext Transfer Protocol (HTTP), Extensible Messaging and Presence Protocol (XMPP), and/or the Session Initiation Protocol (SIP).

In various embodiments, the memory 108 is volatile (such as RAM), non-volatile (such as ROM, flash memory, etc.) or some combination of the two. As mentioned, the system memory 108 may store applications 110, a file system 112, a virtual disk driver 114, and a cloud client 116. In addition, the memory 108 may also store other modules and data, such as an operating system of the computing device 102. Alternatively, any of modules 110-116 may be implemented in hardware. For example, the cloud client 116 may be implemented in a storage controller.

In some embodiments, applications 110 may be any sort of applications, such as operating system components or third party applications. The applications 110 may be kernel-mode applications or user-mode applications. As described further herein, applications 110 may utilize the file system 112 to make write requests or read requests and may receive in response indications of the success or failure of the write requests or data sought by the read requests.

The file system 112 may be any sort of file system known in the art, such as the NT file system (NTFS) of Microsoft® or the third extended file system (ext3) associated with the Linux® kernel. The file system 112 provides access to application 110 and operating system components seeking to read from or write to a disk, whether actual or virtual. The file system 112 interfaces with a disk driver or virtual disk driver, providing write requests or read requests from applications 110 and other components and receiving responses to those requests.

The virtual disk driver 114 appears to the file system 112 as a driver for a local disk drive, such as a standard serial advanced technology attachment (SATA) drive. Thus, the virtual disk driver 114 enables the applications 110 and file system 112 to be oblivious to the location of the storage. The virtual disk driver 114 may be a kernel-mode component of the operating system of the computing device 102 and may forward read and write requests from the file system 112 to the cloud client 116 and forward responses from the cloud client 116 to the file system 112. While FIG. 1 only shows one instance of a virtual disk driver 114, the memory 108 may store multiple instances of the virtual disk driver 114, one for each virtual disk.

The virtual disk driver 114 may be implemented as a kernel-mode component that is separate from the cloud client 116 to enable the cloud client 116 to operate in user-mode. To minimize the overhead of exchanging data across the user-kernel boundary, the virtual disk driver 114 and cloud client 116 may utilize zero-copy inter-process communication. In other embodiments, the virtual disk driver 114 and cloud client 116 may be the same component.

In various embodiments, the cloud client 116 receives read and write requests as block requests from one or more virtual disk drivers 114. The cloud client 116 translates these block requests into input/output operations that are recognizable by the cloud storage devices 104. Specifically, the cloud client 116 translates offsets representing locations of data in a linear range of bytes to tract identifiers and tract offsets. As mentioned, each virtual disk may be associated with a BLOB or data container that includes multiple tracts. The tract identifier identifies a specific tract that the data referenced by the block request is associated with, and the tract offset identifies a specific location within a tract that is associated with the data. In some embodiments, the cloud client 116 maps the offsets in such a way that data blocks with contiguous offsets on the virtual disk are distributed among multiple different tracts, thus increasing the number of tracts referenced for a given block request and the number of cloud storage devices 104 involved in the input/output operations associated with that block request. Further details on the translation capabilities of the cloud client 116 may be found in U.S. patent application Ser. No. 13/598,990, which is entitled “Block-level Access to Parallel Storage,” and was filed on Aug. 30, 2012.

In further embodiments, the cloud client 116 determines whether it is operating in connected mode or disconnected mode. “Connected mode” refers to a connection between the computing device 102 and cloud storage devices 104, and “disconnected mode” refers to the lack of a connection between the computing device 102 and cloud storage devices 104. The cloud client 116 may determine whether it is operating in connected mode or disconnected mode periodically, in response to a notification of an event, such as a connection event or a loss-of-connection event.

When operating in connected mode, the cloud client 116 receives a write request, and translates that request for the cloud storage devices 104, the cloud client 116 first writes the data and metadata associated with that request to the cloud storage 118 of one or more of the cloud storage devices 104. The cloud storage 118 may be memory, removable storage, or non-removable storage of any of the types described in further detail herein. The metadata written may include a unique identifier, such as a globally unique identifier (GUID), a tract identifier, a version identifier for the data, and a checksum for the data. The cloud storage devices 104 store the metadata and data as expanded blocks 120 in the cloud storage 118, each expanded block including one data item and its associated metadata record. The cloud storage device(s) 104 receiving the data then respond to the cloud client 116 with an indication of the success of the write, and the cloud client 116 responds to the requesting application 110 through the virtual disk driver 114 and file system 112, providing the indication.

After providing the indication to the requesting application 110, the cloud client 116 mirrors the write to local storage 122 of the computing device 102. The local storage 122 may be memory, removable storage, or non-removable storage of any of the types described in further detail herein. The cloud client 116 partitions the local storage 122 into two areas: a first area 124 for data and a second area 126 for metadata. These areas may or may not be contiguous locations in the local storage 122. In mirroring the write, the cloud client 116 first writes the data to the first area 124 and then writes the metadata to the second area 126. The metadata written to the second area 126 may include the version identifier for the data and a checksum for the data.

When the cloud client 116 is operating in connected mode and receives a read request, the cloud client 116 retrieves the data sought by the read request from the cloud storage 118 of the cloud storage devices 104.

In various embodiments, when the cloud client 116 is operating in disconnected mode and receives a write request, the cloud client 116 writes the data and metadata for that request to local storage 122. Because the cloud client 116 is disconnected from the cloud storage devices 104, it must provide the data and metadata to the cloud storage devices 104 until the cloud client 116 reconnects to the cloud storage devices 104. In writing to the local storage 122 in disconnected mode, the cloud client 116 reverses the order in which it writes the data and metadata. In disconnected mode, the cloud client 116 first writes the metadata to the second area 126 and then writes the data to the first area 124. The cloud client 116 reverses the order to avoid issues of data integrity when reconciling the cloud storage 118 and local storage 122.

When the cloud client 116 is operating in disconnected mode and receives a read request, the cloud client 116 retrieves the data sought by the read request from the local storage 122.

In some embodiments, the cloud client 116 determines that the computing device 102 has recovered from a crash and that the crash occurred while the computing device 102 was operating in connected mode. In response, the cloud client 116 initiates a reconciliation process between the local storage 122 and the cloud storage 118. As part of the reconciliation process, the cloud client 116 retrieves metadata records from the cloud storage 118 of the cloud storage devices 104. The cloud client 116 then compares version identifiers from the retrieved metadata to version identifiers from metadata records stored in the second area 126 of the local storage 122. Because the cloud client 116 always writes to the cloud storage 118 in connected mode before mirroring those writes, the cloud storage 118 will either have a more recent version of a given data item than the local storage 122 or the same version. The cloud client 116 then updates ones of the data items and their corresponding metadata records on the local storage 122 with counterparts retrieved from the cloud storage 118 when the comparisons for those ones of the data items indicate that the cloud storage 118 has the more recent versions.

The cloud client 116 may also be configured to determine that the computing device 102 has transitioned from disconnected mode to connected mode. In response, the cloud client 116 initiates a reconciliation process between the local storage 122 and the cloud storage 118. As part of the reconciliation process, the cloud client 116 retrieves metadata records from the cloud storage 118 of the cloud storage devices 104. The cloud client 116 then compares version identifiers from the retrieved metadata to version identifiers from metadata records stored in the second area 126 of the local storage 122. Because the cloud client 116 writes only to the local storage 122 while in disconnected mode, the local storage 122 will either have a more recent version of a given data item than the cloud storage 118 or the same version. The cloud client 116 then updates ones of the data items and their corresponding metadata records on the cloud storage 118 with counterparts retrieved from the local storage 122 when the comparisons for those ones of the data items indicate that the local storage 122 has the more recent versions.

In some embodiments, the cloud client 116 maintains an in-memory representation that tracks the writes made in disconnected mode. The cloud client 116 may then use this in-memory representation as part of the reconciliation process and simply provide the tracked writes to the cloud storage 118, avoiding the need to retrieve metadata records, iterate through the second area 126, or compare version identifiers.

In various embodiments, the cloud client 116 may determine that the computing device 102 has recovered from a crash and that the crash occurred while the computing device 102 was operating in disconnected mode. In response, the cloud client 116 works with the file system 112 to detect torn local writes using mechanisms of the file system (the file system possesses such mechanisms to recover from crashes that occur while local storage is handling writes). When the cloud client 116 transitions from disconnected mode to connected mode, the cloud client 116 performs the reconciliation process described above for transitioning from disconnected mode to connected mode.

In some embodiments, the processor(s) 128 is a microprocessing unit (MPU), central processing unit (CPU), a graphics processing unit (GPU), or any other sort of processing unit. Among other capabilities, the processor 128 can be configured to fetch and execute computer-readable processor-accessible instructions stored in memory 108, such as the instructions represented by modules 110-116.

In some embodiments, the network interfaces 130 include any sort of network interfaces 130 known in the art. For example, network interfaces 130 may include a wired interface, such as an Ethernet interface, for wired network communications. Network interfaces 130 may also or instead include any one or more of a wireless LAN interface, or a near field interface.

Computing device 102 may also include additional data storage devices (removable and/or non-removable) such as, for example, magnetic disks, optical disks, or tape. Such additional storage is illustrated in FIG. 1 by removable storage 132 and non-removable storage 134. Memory 108, removable storage 132 and non-removable storage 134 are all examples of computer storage media. As used herein, “computer-readable media” includes computer storage media and communication media. Computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read only memory (ROM), electrically erasable programmable ROM (EEPROM), flash memory or other memory technology, compact disk ROM (CD-ROM), digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store information for access by a computing device.

In contrast, communication media can embody computer-readable instructions, data structures, program modules, or other data in a modulated data signal, such as a carrier wave. As defined herein, computer storage media does not include communication media.

In some embodiments, the I/O devices 136 include any sort of output devices known in the art, such as a display, speakers, a vibrating mechanism, or a tactile feedback mechanism. I/O devices 136 also include ports for one or more peripheral devices, such as headphones, peripheral speakers, or a peripheral display. I/O devices 136 further include any sort of input devices known in the art. For example, I/O devices 136 may include a camera, a microphone, a keyboard/keypad, or a touch-sensitive display. A keyboard/keypad may be a push button numeric dialing pad (such as on a typical telecommunication device), a multi-key keyboard (such as a conventional QWERTY keyboard), or one or more other types of keys or buttons, and may also include a joystick-controller and/or designated navigation buttons, or the like.

While example device configurations and architectures have been described, other implementations are not limited to the particular configurations and architectures described herein. Thus, this disclosure can extend to other implementations, as would be known or as would become known to those skilled in the art.

Example Processes

FIGS. 2-5 illustrate example processes. These processes are illustrated as logical flow graphs, each operation of which represents a sequence of operations that can be implemented in hardware, software, or a combination thereof. In the context of software, the operations represent computer-executable instructions stored on one or more computer storage media that, when executed by one or more processors, perform the recited operations. Generally, computer-executable instructions include routines, programs, objects, components, data structures, and the like that perform particular functions or implement particular abstract data types. The order in which the operations are described is not intended to be construed as a limitation, and any number of the described operations can be combined in any order and/or in parallel to implement the processes.

FIG. 2 illustrates an example process for asynchronously mirroring writes to the cloud storage devices to local storage of the computing device. The process includes, at 202, a computing device receiving a write request from an application of the computing device. At 204, the computing device determines that the computing device is connected to the cloud storage (i.e., is operating in connected mode). In some embodiments, the order of operations 202 and 204 may be reversed.

At 206, the computing device increments a version identifier for a data item associated with the write request. The version identifier may be included in a metadata records associated with the data item.

At 208, the computing device writes the data item associated with the write request, as well as its metadata record, to cloud storage. The writing of data and metadata to the cloud storage may be atomic, with the writing of both data and metadata succeeding or the writing of both data and metadata failing. This writing of data and metadata may include writing the data item, a GUID, a tract identifier, the version identifier, and a checksum. In some embodiments, the computing device determines that it is disconnected from cloud storage responsive to determining that the write to the cloud storage has failed and proceeds to the operation described at block 220.

After writing the data and metadata to the cloud storage, the computing device asynchronously mirrors the write to its local storage. This asynchronous mirroring includes responding, at 210, to the write request of the application by indicating to the application that the write request has succeeded and after or while responding at 210, writing the data and metadata to the local storage. Writing to local storage includes writing, at 212, the data item to a first area of the local storage and writing, at 214, the metadata record for the data item to a second area of the local storage. The computing device performs the writing of the data item before the writing of the metadata record when operating in connected mode.

At 216, the computing device receives a read request and reads one or more data items associated with the read request from cloud storage. While operation 216 is shown as following operations 202-214, it may occur before, during, or after any of the operations 202-214.

At 218, the computing device crashes. This crash may occur before, during, or after any of the operations 202-216. Following the crash, the computing device may perform the operations shown in FIG. 3 and further described herein.

At 220, the computing device disconnects from the cloud storage (i.e., enters disconnected mode). After transitioning to disconnected mode, the computing device may perform the operations shown in FIG. 4 and further described herein.

FIG. 3 illustrates an example process for reconciling local storage of a computing device with cloud storage following a crash of the computing device. The process includes, at 302, a computing device determining that the computing device has recovered from a crash.

At 304, the computing device reconciles the local storage of the computing device with the cloud storage. At 306, the reconciliation includes comparing version identifiers included in metadata records from the second area of the local storage to version identifiers included in metadata records retrieved from the cloud storage. At 308, when the comparison for a given data item indicates that the cloud storage has a more recent version of the data item, the computing device updates the local storage with the data item and corresponding metadata record stored by the cloud storage.

FIG. 4 illustrates an example process for disconnected operation of a computing device and reconciliation of local storage of the computing device to cloud storage upon reconnecting to the cloud storage. The process includes, at 402, a computing device receiving a write request from an application of the computing device. At 404, the computing device determines that the computing device is disconnected to the cloud storage (i.e., is operating in disconnected mode). In some embodiments, the order of operations 402 and 404 may be reversed.

At 406, the computing device increments a version identifier for a data item associated with the write request. The version identifier may be included in a metadata records associated with the data item.

At 408, the computing device then writes the metadata record for the data item associated with the write request to a second area of the local storage. At 410, the computing device writes the data item to a first area of the local storage. The computing device performs the writing of the metadata record before the writing of the data item when operating in disconnected mode. This writing of data and metadata may include writing the data item, the version identifier, and a checksum.

In some circumstances, the computing device may crash following operation 408 or 410. Following the crash, the computing device may perform the operations shown in FIG. 5 and further described herein.

At 414, the computing device receives a read request and reads one or more data items associated with the read request from local storage. While operation 414 is shown as following operations 402-412, it may occur before, during, or after any of the operations 402-412.

At 416, the computing device determines that the computing device is now connected to the cloud storage (i.e., has transitioned to connected mode). This determination at 416 may be performed before, during, or after any of operations 402-416.

After determining that the computing device is connected to cloud storage, the computing device, at 418, reconciles the local storage of the computing device with the cloud storage. At 420, the reconciliation includes comparing version identifiers included in metadata records from the second area of the local storage to version identifiers included in metadata records retrieved from the cloud storage. At 422, when the comparison for a given data item indicates that the local storage has a more recent version of the data item, the computing device updates the cloud storage with the data item and corresponding metadata record stored by the local storage.

FIG. 5 illustrates an example process for reconciling local storage of a computing device with cloud storage following disconnected operation during which the computing device crashed. The process includes, at 502, a computing device determining that the computing device has recovered from a crash that occurred while the computing device was in disconnected mode.

At 504, the computing device scans the local storage to detect torn writes based on checksums. This may involve calculating a checksum for each data item in the local storage and comparing that checksum to the checksum stored in the metadata record for that data item. At 506, a file system of the computing device then attempts to repair any torn writes.

At 508, the computing device determines that the computing device is now connected to the cloud storage (i.e., has transitioned to connected mode). After determining that the computing device is connected to cloud storage, the computing device, at 510, reconciles the local storage of the computing device with the cloud storage. At 512, the reconciliation includes comparing version identifiers included in metadata records from the second area of the local storage to version identifiers included in metadata records retrieved from the cloud storage. At 514, when the comparison for a given data item indicates that the local storage has a more recent version of the data item, the computing device updates the cloud storage with the data item and corresponding metadata record stored by the local storage.

CONCLUSION

Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described. Rather, the specific features and acts are disclosed as exemplary forms of implementing the claims.

Claims

1. A device comprising: one or more processors;memory storing instructions that, when executed by the one or more processors, cause the device to perform operations comprising: receiving a write request from an application while the device is operating in a connected mode in which the device is connected to cloud storage, the connected mode defining a first order in which instances of data and metadata for the instances of data are to be written to local storage of the device;writing a first instance of data specified by the write request to the cloud storage;indicating success of the write request to the application;after indicating the success of the write request to the application, writing the first instance of data to a first area of the local storage;after writing the first instance of data to the first area of the local storage, writing metadata for the first instance of data to a second area of the local storage;determining that the device has been disconnected from the cloud storage and is operating in a disconnected mode, the disconnected mode defining a second order in which the instances of data and the metadata for the instance of data are to be written to the local storage, the second order being reversed compared to the first order to avoid data integrity issues associated with reconciling the cloud storage and the local storage;writing, while the device is operating in the disconnected mode, metadata for a second instance of data to the second area of the local storage; andafter writing the metadata for the second instance of data to the second area of the local storage, writing the second instance of data to the first area of the local storage.

2. The device of claim 1, wherein, while the device is disconnected from the cloud storage, the device reads from and writes to the local storage.

3. The device of claim 1, wherein the operations further comprise reconciling the local storage with the cloud storage based at least on a comparison of metadata stored in the second area of the local storage with metadata received from the cloud storage the reconciling comprises: scanning the local storage to detect torn writes, the detecting being based at least on calculating checksums,repairing the torn writes,comparing version identifiers included in the metadata stored in the second area of the local storage with version identifiers included in the metadata received from the cloud storage, andbased on the comparison indicating that the local storage has a more recent version of the second instance of data than the cloud storage, updating the cloud storage with the second instance of data from the local storage.

4. The device of claim 1, wherein metadata stored in the second area of the local storage and metadata received from the cloud storage include version identifiers.

5. One or more computer storage devices having stored thereon computer-executable instructions configured to program a computing device to perform operations comprising: receiving a first write request from an application of the computing device while the computing device is operating in a connected mode in which the computing device is connected to cloud storage, the connected mode defining a first order in which instances of data and metadata for the instances of data are to be written to local storage of the computing device;writing a first instance of data specified by the first write request to the cloud storage;indicating success of the first write request to the application;after indicating the success of the first write request to the application, writing the first instance of data to a first area of the local storage;after writing the first instance of data to the first area of the local storage, writing metadata for the first instance of data to a second area of the local storage;determining that the computing device has been disconnected from the cloud storage and is operating in a disconnected mode, the disconnected mode defining a second order in which the instances of data and the metadata for the instance of data are to be written to the local storage, the second order being reversed compared to the first order to avoid data integrity issues associated with reconciling the cloud storage and the local storage;receiving, while the computing device is operating in the disconnected mode, a second write request;writing, while the computing device is operating in the disconnected mode, metadata for a second instance of data to the second area of the local storage; andupon completion of the writing the metadata for the second instance of data to the second area of the local storage, writing the second instance of data to the first area of the local storage.

6. The one or more computer storage devices of claim 5, the operations further comprising, prior to writing the first instance of data to the cloud storage, determining that the computing device is connected to the cloud storage.

7. The one or more computer storage devices of claim 5, the operations further comprising reading from the cloud storage when the computing device is connected to the cloud storage and reading from the local storage when the computing device is disconnected from the cloud storage.

8. The one or more computer storage devices of claim 5, the operations further comprising incrementing a version identifier associated with the first instance of data, wherein the writing the first instance of data to the cloud storage includes writing the incremented version identifier as the metadata for the first instance of data to the cloud storage, andwherein the writing the first instance of data to the local storage includes writing the incremented version identifier as the metadata for the first instance of data to the local storage.

9. A system comprising: a cloud storage device having a first processor and first one or more modules that, when operated by the first processor, receive writes including data items and metadata for the data items and store each data item contiguously with its metadata as an expanded block; anda client device having a second processor and second one or more modules that, when operated by the second processor, write the data items and the metadata to the cloud storage device and store the data items and the metadata in local storage of the client device, wherein: while the client device is operating in a connected mode in which the client device is connected to the cloud storage device, a first data item is stored in a first area of the local storage before metadata for the first data item is stored in a second area of the local storage in accordance with a first order in which the data items and the metadata are to be written to the local storage; andwhile the client device is operating in a disconnected mode in which the client device is disconnected from the cloud storage device, a second data item is stored in the first area of the local storage after the metadata for the second data item is successfully stored in the second area of the local storage in accordance with a second order in which the data items and the metadata are to be written to local storage, the second order being reversed compared to the first order to avoid data integrity issues associated with reconciling the cloud storage device and the local storage.

10. The system of claim 9, wherein the client device further includes a memory bitmap and the client device utilizes the memory bitmap to track writes to the local storage while the client device is disconnected from the cloud storage device.

11. The system of claim 9, wherein the second one or more modules include a virtual disk driver and a cloud client.

12. The system of claim 11, wherein the virtual disk driver receives write requests from applications of the client device and provides those write requests to the cloud client.

13. The system of claim 11, wherein the cloud client receives the write requests from the virtual disk driver and performs the writing of the data items and the metadata to the cloud storage device and the storing of the data items and the metadata in the local storage.

14. The device of claim 1, the operations further comprising incrementing a version identifier associated with the first instance of data, wherein the writing the first instance of data to the cloud storage includes writing the incremented version identifier as the metadata for the first instance of data to the cloud storage, andwherein the writing the first instance of data to the local storage includes writing the incremented version identifier as the metadata for the first instance of data to the local storage.

15. The device of claim 1, wherein the device further comprises a memory bitmap and the device utilizes the memory bitmap to track writes to the local storage while the device is disconnected from the cloud storage.

16. The one or more computer storage devices of claim 5, the operations further comprising reconciling the local storage with the cloud storage based at least on a comparison of metadata stored in the first area of the local storage with metadata received from the cloud storage, the reconciling comprises: scanning the local storage to detect torn writes, the detecting being based at least on calculating checksums,repairing the torn writes,comparing version identifiers included in the metadata stored in the first area of the local storage with version identifiers included in the metadata received from the cloud storage, andbased on the comparison indicating that the local storage has a more recent version of the second instance of data than the cloud storage, updating the cloud storage with the second instance of data from the local storage.

17. The system of claim 11, wherein the second one or more modules are configured to provide an indication of a successful write to an application that requested the write without awaiting a result of mirroring the write to the local storage.

18. The system of claim 11, wherein the metadata stored in the local storage and the cloud storage includes version identifiers.

19. The system of claim 11, wherein the data items and the metadata are stored in separate areas of the local storage.