Stateless datastore-independent transactions

Patent metadata

Field	Value
Publication number	US-9519674-B2
Application number	US-201414482661-A
Country	US
Kind code	B2
Filing date	Sep 10, 2014
Priority date	Sep 10, 2014
Publication date	Dec 13, 2016
Grant date	Dec 13, 2016

Abstract

At a client-side component of a storage group, a read descriptor generated in response to a read request directed to a first data store is received. The read descriptor includes a state transition indicator corresponding to a write that has been applied at the first data store. A write descriptor indicative of a write that depends on a result of the read request is generated at the client-side component. The read descriptor and the write descriptor are included in a commit request for a candidate transaction at the client-side component, and transmitted to a transaction manager.

First claim

What is claimed is:

1. A system, comprising:
one or more computing devices configured to:
receive, at a client-side component of a heterogeneous storage group comprising a plurality of data stores including a first data store and a second data store, respective read descriptors corresponding to one or more reads, including a first read descriptor corresponding to a first read directed to the first data store, wherein the first read descriptor comprises an indication of a state transition that has been applied at the first data store, and wherein, in accordance with a stateless protocol, the first data store does not maintain session metadata pertaining to the client-side component after the first read descriptor is provided to the client-side component;
store, in one or more buffers accessible to the client-side component, respective write descriptors of one or more writes, including a first write descriptor of a first write directed to a particular data store of the heterogeneous storage group, wherein a payload of the first write is based at least in part on a result of the first read, and wherein the first write descriptor indicates an object to be modified at the particular data store;
generate, at the client-side component, a commit request for a candidate transaction comprising the one or more writes, wherein the commit request comprises at least the first read descriptor and the first write descriptor;
transmit, from the client-side component to a transaction manager, the commit request;
determine, at the transaction manager, that the candidate transaction is accepted for commit based at least in part on an analysis of the first read descriptor in accordance with a concurrency control protocol; and
initiate a modification of the object indicated in the first write descriptor.

2. The system as recited in claim 1 , wherein the client-side component comprises a process executing at a request handler node of a storage service implemented at a provider network.

3. The system as recited in claim 1 , wherein the first read descriptor comprises read repeatability verification metadata for at least the first read.

4. The system as recited in claim 1 , wherein the commit request includes a second read descriptor corresponding to a second read directed to the second data store, wherein a payload of at least one write of the one or more writes is based at least in part on a result of the second read.

5. The system as recited in claim 1 , wherein the one or more writes comprise a second write directed to a different data store than the particular data store.

6. A method, comprising:
performing, by one or more computing devices:
receiving, at a client-side component of a storage group comprising one or more data stores including a first data store, respective read descriptors corresponding to one or more reads, including a first read descriptor corresponding to a first read directed to the first data store, wherein the first read descriptor comprises an indication of a state transition that has been applied at the first data store;
generating, at the client-side component, respective write descriptors of one or more writes, including a first write descriptor of a first write directed to a particular data store of the one or more data stores, wherein a payload of the first write is based at least in part on a result of the first read, and wherein the first write descriptor indicates an object to be modified at the particular data store;
transmitting, from the client-side component to a transaction manager, a commit request for a candidate transaction comprising one or more writes including the first write, wherein the commit request comprises at least the first read descriptor and the first write descriptor;
determining, at the transaction manager, that the candidate transaction is to be accepted for commit based at least in part on an analysis of the first read descriptor.

7. The method as recited in claim 6 , wherein in accordance with a stateless protocol, the first data store does not maintain session metadata pertaining to the first read after the first read descriptor is provided to the client-side component.

8. The method as recited in claim 6 , further comprising performing, by the one or more computing devices:
storing, in a durable log of the transaction manager, a transaction record indicating that the candidate transaction has been accepted for commit;

wherein said determining that the candidate transaction is to be accepted for commit is based at least in part on an analysis of one or more other transaction records stored in the durable log.

9. The method as recited in claim 8 , wherein the durable log is implemented using a plurality of nodes of a replication graph.

10. The method as recited in claim 6 , wherein said determining that the candidate transaction is to be accepted for commit comprises verifying that a result set of the first read has not been modified since the read descriptor was generated.

11. The method as recited in claim 6 , wherein the indication of the particular state transition comprises a logical timestamp.

12. The method as recited in claim 6 , wherein the first read descriptor comprises read repeatability verification metadata for at least the first read.

13. The method as recited in claim 6 , wherein the one or more data stores comprise a second data store, wherein the first data store implements a first data model and the second data store implements a different data model, wherein the commit request includes a second read descriptor corresponding to a second read directed to the second data store, wherein a payload of at least one write of the one or more writes is based at least in part on a result of the second read.

14. The method as recited in claim 12 , wherein the first data store comprises one of: a non-relational database system, a relational database system, a storage service that implements a web services interface allowing access to unstructured data objects, an in-memory database, an instance of a distributed cache, a component of a queueing service, or a component of a notification service.

15. The method as recited in claim 6 , wherein the first data store does not support atomicity for a transaction that includes more than one write.

16. A non-transitory computer-accessible storage medium storing program instructions that when executed on one or more processors:
receive, at a client-side component of a storage group comprising one or more data stores including a first data store, respective read descriptors corresponding to one or more reads, including a first read descriptor corresponding to a first read directed to the first data store, wherein the first read descriptor comprises an indication of a state transition that has been applied at the first data store;
generate, at the client-side component, respective write descriptors of one or more writes, including a first write descriptor of a first write directed to a particular data store of the one or more data stores, wherein a payload of the first write is based at least in part on a result of the first read, and wherein the first write descriptor indicates an object to be modified at the particular data store; and
transmit, from the client-side component to a transaction manager, a commit request for a candidate transaction comprising one or more writes including the first write, wherein the commit request comprises at least the first read descriptor and the first write descriptor.

17. The non-transitory computer-accessible storage medium as recited in claim

Assignees

• Amazon Tech Inc

Inventors

• Vermeulen Allan Henry
• Rath Timothy Andrew
• Cole Timothy Daniel
• Muniswamy-Reddy Kiran-Kumar

Classifications

• G06F17/30348

  Physics · mapped topic

• G06F17/30377 Primary

  Physics · mapped topic

• G06F17/30545

  Physics · mapped topic

• G06F16/2308

  Concurrency control (transaction processing G06F9/466) · CPC title

• G06F16/2358

  Change logging, detection, and notification (replication G06F16/27) · CPC title