Transactional replicator

• US 10,282,364 B2
• Filed: 10/22/2015
• Issued: 05/07/2019
• Est. Priority Date: 04/28/2015
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Kakivaya, Gopal Krishna R., Gupta, Anurag, Shivaprakash, Sumukh, Tarta, Mihail G., Coskun, Mert
• Primary Examiner(s): Lewis, Cheryl
• Application Number:
  US14/920,351
• Publication Number:
  US 20160321296A1
• Time in Patent Office: 1,293 Days
• CPC Class Codes: G06F 16/275 Synchronous replication

Abstract

A transactional replicator applying group commit and barrier concepts is disclosed. Group commit means that the transactional replicator commits multiple transactions in a group and is not restricted to committing single transactions one-at-a-time and is not limited to operating on a single state provider. Barrier means that the transactional replicator does not move forward to commit additional transactions until the previous group of transactions are completed. All state providers must apply their transactions and update state before additional transactions will be committed. A quorum acknowledgement “unlocks” any locks that were acquired to update the state within a state provider. However, as long as there are no lock conflicts, additional transactions can continue to create new group commits as necessary.

Claims

1. A system, comprising:

a plurality of nodes of a distributed cloud computing system, each node comprising a processor and memory and hosting one or more state providers, the state providers configured to provide transactional semantics for interactions with users; and

a transactional replicator configured to replicate states from a primary state-provider replica to at least one secondary state-provider replica, the transactional replicator further configured to;

identify when a first transaction has been committed;

initiate a first group commit by applying a first group of transaction events to the primary state-provider replicas, wherein the first group of transaction events comprise all transaction events received during a first period;

replicate the first group of transaction events to the secondary state-provider replicas; and

allow parallel dispatch of transactions across the state providers to achieve improved throughput.

2. The system of claim 1, wherein the transactional replicator is further configured to:

receive a quorum acknowledgement for the first group commit when a quorum of the primary and secondary state-provider replicas have applied the first group of transaction events.

3. The system of claim 2, wherein the transactional replicator is further configured to:

identify when a second transaction has been committed;

if the quorum acknowledgement has been received, execute a second group commit by applying a second group of transaction events to the primary state-provider replicas, wherein the second group of transaction events comprise all transaction events received during a second period that begins at the end of the first period; and

replicate the second group of transaction events to the secondary state-provider replicas.

4. The system of claim 2, wherein the transactional replicator is further configured to:

send an unlock command to the state providers after the quorum acknowledgement has been received.

5. The system of claim 2, wherein the transactional replicator is further configured to:

after initiating the first group commit, instruct the state providers to prepare to checkpoint their state;

after receiving the quorum acknowledge for the first group commit, instructing the state providers to perform checkpointing; and

instructing the state providers to complete checkpointing.

6. The system of claim 5, wherein the transactional replicator is further configured to:

generate log entries when a prepare checkpoint, preform checkpoint, or complete checkpoint instruction is sent to the state providers.

7. The system of claim 2, wherein the state providers are further configured to:

receive a prepare checkpoint instruction from the transactional replicator; and

identify all state to be documented in a current checkpoint.

8. The system of claim 7, wherein the state providers are further configured to:

receive a perform checkpoint instruction from the transactional replicator; and

store the state to be documented to a temporary file.

9. The system of claim 8, wherein the state providers are further configured to:

receive a complete checkpoint instruction from the transactional replicator; and

rename the temporary file to a permanent checkpoint filename after saving the state to be documented.

10. The system of claim 2, wherein the transactional replicator is further configured to:

create a complete backup file documenting all state in a current primary state-provider replica;

create a series of incremental backup files documenting state changes on the current primary state-provider replica since a prior full or incremental backup, the incremental backup files including checkpoint logs for the primary state-provider replica; and

provide the complete and incremental backup files to a user for storage outside the system.

11. The system of claim 10, wherein the current primary state-provider replica changes over time among a group of primary and secondary state-provider replicas, and wherein the complete and incremental backup files are created from more than one individual replica.

12. The system of claim 10, wherein the transactional replicator is further configured to:

receive a data-loss recovery instruction from a user;

obtain a group of complete and incremental backup files from a user;

apply state from the complete backup file to a new primary state-provider replica; and

apply state changes from the series of incremental backup files to the new primary state-provider replica, wherein the state changes from the incremental backup files are applied in order created, and wherein the checkpoint logs are referenced to ensure no false progress is applied to the new primary state-provider replica.

13. The system of claim 1, wherein the state providers are further configured to:

receive a transaction event;

lock an object to be modified by the transaction event;

apply an updated state to the object, wherein the updated state is indicated by the transaction event;

notify the transactional replicator when the object has been updated;

receive an unlock instruction from the transactional replicator; and

unlock the object after receiving the unlock instruction.

14. A computer-implemented method, comprising:

receiving, at a state provider running on a node of a distributed computing system, a prepare-checkpoint instruction from a transactional replicator;

identifying all current state changes being applied by the state provider;

receiving, at the state provider, a perform-checkpoint instruction from the transactional replicator;

storing the current state changes to a temporary checkpoint file;

receiving, at the state provider, a complete-checkpoint instruction from the transactional replicator;

designating the temporary checkpoint file as a completed checkpoint file; and

ensuring that transactions within a group commit boundary are applied to a quorum of replicas before checkpointing is completed by the state provider to prevent false progress of data.

15. The computer-implemented method of claim 14, further comprising:

logging, by the transactional replicator, the prepare-checkpoint, perform-checkpoint, and complete-checkpoint instructions as they are provided to the state provider to document checkpoint progress.

16. A computing node in a distributed computing system, the computing node comprising a processor and memory and further comprising:

one or more state providers, the state providers configured to provide transactional semantics for interactions with users; and

a transactional replicator configured to replicate states from a primary state-provider replica to at least one secondary state-provider replica, the transactional replicator further configured to;

create a complete backup file documenting all states in a current primary state-provider replica;

create a series of incremental backup files documenting state changes on the current primary state-provider replica since a prior full or incremental backup, the incremental backup files including checkpoint logs for the primary state-provider replica;

provide the complete and incremental backup files to a user for storage outside the system; and

reference the checkpoint logs to ensure no false progress is applied to a new primary state-provider replica.

17. The computing node of claim 16, wherein the current primary state-provider replica changes over time among a group of primary and secondary state-provider replicas, and wherein the complete and incremental backup files are created from more than one individual replica.

18. The computing node of claim 16, wherein the transactional replicator is further configured to:

receive a data-loss recovery instruction from a user;

obtain a group of complete and incremental backup files from a user;

apply state from the complete backup file to a new primary state-provider replica; and

apply state changes from the series of incremental backup files to the new primary state-provider replica, wherein the state changes from the incremental backup files are applied in order created.