Data mobility, accessibility, and consistency in a data storage system

A method for performing a single-phase commit in a data storage system, the method comprising:

• receiving a plurality of write requests, each write request originating from a client and providing a logical address for storing data;
• atomically committing the plurality of write requests in a single-phase transaction, wherein each write request is committed to two or more physical locations in the data storage system, and wherein the two or more physical locations are indicated by the logical address associated with the write request;
• receiving commit responses from each write request in the transaction, the commit responses indicated as positive, negative, or null;
• and
• resolving the transaction as positive or negative based on the commit responses, wherein the transaction is resolved as positive where at least one write request returned a positive response and no write request returned a negative response, and wherein the transaction is resolved as negative where at least one write request returned a negative response.

A method for restoring data associated with an unavailable block storage unit in a data storage system, the method comprising:

• identifying a block storage unit that becomes unavailable during a transaction of a plurality of write requests atomically committed to the data storage system, wherein one of the plurality of write requests is addressed to the unavailable block storage unit;
• indicating a new block storage unit for storing data intended for the unavailable block storage unit;
• causing the data intended for the unavailable block storage unit to be stored in the indicated block storage unit;
• and
• updating the unavailable block storage unit with the data stored in the indicated block storage unit when available.

A method for providing data mobility with continuous access to data in a data storage system, the method comprising:

• adding a new layer to a protected data storage unit, the new layer identifying two or more block storage units for storing data contained in any new write request intended for the protected data storage unit;
• copying data from a first block storage unit located on a first hardware storage device associated with a first computer node to a second block storage unit located a second hardware storage device associated with a second node, wherein the first block storage unit and the second block storage unit are identified in a first layer of the protected data storage unit, and wherein the first layer is below the new layer;
• and
• updating an entry in a data storage map to reference the second block storage unit in the first layer.

A method of merging data within a data storage system, the method comprising:

• adding a top layer in a protected data storage unit, wherein the top layer identifies two or more block storage units, each block storage unit located on a different hardware storage device, and each hardware storage device associated with a node in the data storage system, and wherein the protected data storage unit includes two or more layers other than the top layer;
• merging data in a first block storage unit referenced in a first underlying layer of the protected data storage unit with data in a second block storage unit referenced in a second underlying layer of the protected data storage unit, wherein the first underlying layer is above the second underlying layer, and wherein the first block storage unit and the second block storage unit are located on hardware storage devices associated with a same node;
• and
• updating a data storage map to reference a new block storage unit in the first underlying layer, the new block storage unit including the merged data from both the first block storage unit and the second block storage unit.

A transactional storage system providing data mobility, the system comprising:

• one or more computer nodes, each computer node associated with one or more hardware storage devices;
• a data storage system for providing user access to data stored on two or more of the hardware storage devices;
• a data storage map for linking protected data storage units referenced by the data storage system to data stored on the hardware storage devices, wherein each protected data storage unit includes one or more layers, each layer indicating two or more block storage units containing the same data, and wherein each block storage unit is associated with a hardware storage device on which data is stored;
• and
• an address abstraction layer adapted to:
• translate a write request from the data storage system to a protected data storage unit into two or more write requests, each of the two or more write requests intended for the two or more differing block storage units identified in a layer of the protected data storage unit, wherein only a top layer of the protected data storage unit is capable of receiving a write request;
• and
• modify data identified in the protected data storage unit, wherein the modification includes copying data contained by a first block storage unit of a first underlying layer of the protected storage unit to a second block storage unit identified in a second underlying layer of the protected storage unit.

A transactional storage system, configured to operate with respect to a plurality of pstores, each pstore comprising one or more bstores, capable of supporting single phase commit comprising, the transactional storage system comprising:

• a component configured to initiate a transaction;
• a component configured to identify every bstore participating in the transaction;
• a component configured to transmit write requests associated with the transaction to each identified bstore participating in the transaction;
• a component configured to, send a commit request to each identified bstore participating in the transaction, wherein a commit response is sent to the initiator each time a commit operation associated with the transaction is performed at any identified bstore participating in the transaction;
• a component configured to, in response to a determination that the transaction has failed at at least one bstore, invoke a resolution protocol to determine a state of the transaction, the component configured to invoke the resolution protocol comprising:
• a component configured to, in response to a determination that at least one identified bstore participating in the transaction has sent a negative commit response in association with the transaction, abort the transaction, and
• a component configured to, in response to a determination that a positive or unknown commit response has been received from each identified bstore participating in the transaction in association with the transaction,
• provide an indication that the transaction has been committed, and
• for each unknown commit response, create a cleaning kit.