DATABASE SYSTEM WITH DATABASE ENGINE AND SEPARATE DISTRIBUTED STORAGE SERVICE

• US 20140279929A1
• Filed: 03/07/2014
• Published: 09/18/2014
• Est. Priority Date: 03/15/2013
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: GUPTA, ANURAG WINDLASS, FACHAN, NEAL, MCKELVIE, SAMUEL JAMES, BURCHALL, LAURION DARRELL, MADHAVARAPU, PRADEEP JNANA, NEWCOMBE, CHRISTOPHER RICHARD, TOBLER, BENJAMIN, COREY, JAMES MCCLELLAN
• Granted Patent: US 9,514,007 B2
• Time in Patent Office: Days
• US Class Current: 707/683
• CPC Class Codes: G06F 11/1451 by selection of backup contents
• CPC Class Codes: G06F 11/1471 involving logging of persistent data for recovery
• CPC Class Codes: G06F 11/2094 Redundant storage or storage space G06F11/2056 takes precedence
• CPC Class Codes: G06F 16/23 Updating
• CPC Class Codes: G06F 16/2365 Ensuring data consistency and integrity
• CPC Class Codes: G06F 2201/80 Database-specific techniques

Abstract

A database system may include a database service and a separate distributed storage service. The database service (or a database engine head node thereof) may be responsible for query parsing, optimization, and execution, transactionality, and consistency, while the storage service may be responsible for generating data pages from redo log records and for durability of those data pages. For example, in response to a write request directed to a particular data page, the database engine head node may generate a redo log record and send it, but not the data page, to a storage service node. The storage service node may store the redo log record and return a write acknowledgement to the database service prior to applying the redo log record. The server node may apply the redo log record and other redo log records to a previously stored version of the data page to create a current version.

Claims

1. A computing system, comprising:

a plurality of computing nodes, each of which comprises at least one processor and a memory;

wherein one or more of the plurality of computing nodes is configured to implement a database service, and wherein the database service comprises a database engine head node;

wherein two or more other ones of the plurality of computing nodes are configured to implement server nodes of a distributed storage service that stores portions of database tables as respective data pages on one or more storage devices;

wherein the database engine head node is configured to;

receive, from a client of the database service, a write request directed to a given data record in a database table, wherein the write request specifies a modification to be made to the given data record;

generate a redo log record representing the modification to be made to the given data record;

send the redo log record, but not a data page comprising the given data record, to a particular server node of the distributed storage service that stores a version of the data page comprising the given data record;

wherein the particular server node of the distributed storage service is configured to;

receive the redo log record from the database engine head node;

write the redo log record to one or more storage devices;

return, to the database engine head node, an acknowledgement that the redo log record was written; and

subsequent to returning the acknowledgement;

generate a current version of the data page comprising the given data record, wherein to generate the current version of the data page, the particular server node of the distributed storage service is configured to apply the received redo log record and one or more other redo log records representing modifications to the data page to a previously stored version of the data page; and

write the current version of the data page to one or more storage devices.

2. The computing system of claim 1,

wherein the database engine head node is further configured to:

receive, from a database client, a request to read the data page comprising the given data record; and

in response to receiving the request to read the data page, send a request for a current version of the data page to the particular server node of the distributed storage service.

3. The computing system of claim 2,

wherein the particular server node of the distributed storage service is further configured to:

receive, from the database engine head node, the request for the current version of the data page comprising the given data record; and

in response to receiving the request for the current version of the data page, return the current version of the data page to the database engine head node.

4. The computing system of claim 3, wherein generating the current version of the data page is performed in response to receiving the request for the current version of the data page.

5. The computing system of claim 1,

wherein the database engine head node comprises a client-side driver for the distributed storage system;

and

wherein to send the redo log record to the particular server node of the distributed storage system, the client-side driver is configured to;

determine the particular server node of the distributed storage service that stores a version of the data page comprising the given data record; and

send the redo log record to the particular server node of the distributed storage system on behalf of the database service.

6. A system, comprising:

one or more computing nodes, each of which comprises at least one processor and a memory, wherein the one or more computing nodes are configured to collectively implement a database service, and where the database service comprises a database engine head node and an interface to a distributed storage system;

wherein the database engine head node is configured to;

receive, from a client of the database service, a write request directed to a given data record in a database table, wherein the write request specifies a modification to be made to the given data record;

generate a redo log record representing the modification to be made to the given data record;

send, via the interface, the redo log record, but not a data page comprising the given data record, to a particular server node of the distributed storage service that stores a version of the data page comprising the given data record;

receive, via the interface, an acknowledgment that the redo log record has been written to the distributed storages service; and

return, to the database client, a response indicating that the requested write has been performed.

7. The system of claim 6, wherein the database engine head node is further configured to:

receive, from a client of the database service, a request to read the data page comprising the given data record; and

in response to receiving the request to read the data page;

send, via the interface, a request for a current version of the data page to the particular server node of the distributed storage service;

receive, via the interface, the current version of the data page; and

return, to the client from which the request to read the data page was received, the current version of the data page.

8. The system of claim 7,

wherein the database engine head node comprises a cache that stores recently accessed data pages;

wherein the database engine head node is configured to send the request for the current version of the data page to the particular server node in response to determining that a current version of the data page is not present in the cache.

9. The system of claim 6,

wherein the database engine head node comprises a cache that stores recently accessed data pages;

wherein the cache stores a copy of the data page comprising the given data record; and

wherein the database engine head node is further configured to apply the modification specified in the write request to the given data record in the cached copy of the data page.

10. The system of claim 6, wherein the database engine head node is further configured to:

prior to sending the redo log record, determine the particular server node of the distributed storage service that stores a version of the data page comprising the given data record, wherein determining the particular server node is performed by the interface to the distributed storage system.

11. The system of claim 6, wherein the database engine head node is further configured to:

receive, from one or more clients of the database service, two or more additional write requests directed to the database table; and

service the two or more additional write requests, wherein servicing the two or more additional write requests comprises enforcing atomicity, isolation, and consistency properties of transactions that target the database table.

12. A system, comprising:

one or more computing nodes, each of which comprises at least one processor and a memory, wherein the one or more computing nodes are configured to collectively implement a distributed storage service that stores data blocks on one or more storage devices on behalf of a client of the distributed storage service;

wherein a given one of the one or more computing nodes is configured to;

receive, from the client, an update record for a particular data block that is stored on the given computing node, wherein the update record specifies a modification to be made to the particular data block;

write the update record to one or more of the one or more storage devices; and

return, to the client, an acknowledgement that the update record was written, without waiting for a current version of the particular data block to be generated or for the modification specified in the update record to be applied to the particular data block.

13. The system of claim 12, wherein, subsequent to returning the acknowledgement, the given one of the one or more computing nodes is further configured to:

generate a current version of the particular data block, dependent on the update record; and

write the current version of the particular data block to one or more of the one or more storage devices.

14. The system of claim 13, wherein to generate the current version of the particular data block, the given computing node is configured to apply the update record to a previously stored version of the particular data block.

15. The system of claim 14, wherein to generate the current version of the particular data block, the given computing node is further configured to apply, to the previously stored version of the particular data block, one or more other update records representing modifications to the particular data block that were received since the previously stored version of the particular data block was written.

16. The system of claim 13, wherein to generate the current version of the particular data block, the given computing node is configured a execute a background process to apply the update record and one or more other update records to a previously stored version of the particular data block while a foreground process receives read or write requests targeting data blocks stored by the distributed storage service.

17. The system of claim 13,

wherein the given one of the one or more computing nodes is further configured to receive, from the client, a read request targeting the data block;

and

wherein generating the current version of the particular data block is performed in response to receiving the read request.

18. The system of claim 12,

wherein another one of the one or more computing nodes stores the particular data block;

and

wherein the other one of the one or more computing nodes is configured to;

receive, from the client, the update record for the particular data block;

write the update record to another one of the one or more storage devices;

return, to the client, another acknowledgement that the update record was written; and

subsequent to returning the other acknowledgement;

generate a current version of the particular data block, wherein to generate the current version of the particular data block, the given computing node is configured to apply the update record to a previously stored version of the particular data block; and

write the current version of the particular data block to another one of the one or more storage devices.

19. A non-transitory computer-readable storage medium storing program instructions that when executed on one or more computers cause the one or more computers to perform:

receiving a query comprising a write request targeting a data block that is stored in a distributed storage system, wherein the write request specifies a modification to be made to the data block;

generating a redo log record representing the modification to be made to the data block;

determining a node in the distributed storage system that stores the data block;

sending the redo log record, but not a copy of the data block, to the determined node in the distributed storage system; and

receiving an acknowledgement from the determined node in the distributed storage system that the redo log record has been written to a storage device on the determined node.

20. The non-transitory computer-readable storage medium of claim 19, wherein when executed on the one or more computers, the program instructions further cause the one or more computers to perform, prior to said generating a redo log:

parsing the query; and

generating a query execution plan.

21. The non-transitory computer-readable storage medium of claim 19, wherein when executed on the one or more computers, the program instructions further cause the one or more computers to perform:

determining one or more other nodes in the distributed storage system that store the data block, wherein the one or more other nodes are part of a same protection group as the determined node; and

sending the redo log record, but not a copy of the data block, to the one or more other nodes in the distributed storage system.

22. The non-transitory computer-readable storage medium of claim 21, wherein when executed on the one or more computers, the program instructions further cause the one or more computers to perform:

receiving an acknowledgement from a quorum of nodes in the protection group that the redo log record has been written to a storage device on each of the nodes in the quorum of nodes; and

in response to receiving an acknowledgement from the quorum of nodes, returning, to a client from which the query was received, a query response indicating that the write request has been committed.

23. The non-transitory computer-readable storage medium of claim 19, wherein the query is received from a database service.

24. A non-transitory computer-readable storage medium storing program instructions that when executed on one or more computers cause the one or more computers to perform:

storing a plurality of data pages comprising data records of a database table on one or more storage system server nodes, each comprising one or more storage devices;

receiving a redo log record directed to a particular data page;

storing the redo log record on one or more of the storage devices;

determining that a coalescing operation is warranted for the particular data page;

in response to said determining;

performing a coalescing operation for the particular data page, wherein performing the coalescing operation comprise applying the redo log record and one or more other redo log records to a previously generated version of the particular data page to generate a current version of the particular data page; and

storing the current version of the particular data page one or more of the storage devices.

25. The non-transitory computer-readable storage medium of claim 24, wherein when executed on the one or more computers, the program instructions further cause the one or more computers to perform:

receiving a read request directed to the particular data page; and

in response to receiving the read request, returning the current version of the particular data page.

26. The non-transitory computer-readable storage medium of claim 24, wherein said determining that a coalescing operation is warranted comprises determining that the number of redo log records directed to the particular data page that have been received but not yet applied to the particular data page exceeds a pre-determined threshold for a coalescing operation.

27. The non-transitory computer-readable storage medium of claim 24, wherein said determining that a coalescing operation is warranted comprises determining that a node or storage device has failed.

28. The non-transitory computer-readable storage medium of claim 24, wherein when executed on the one or more computers, the program instructions further cause the one or more computers to perform:

receiving another redo log record directed to another data page;

storing the other redo log record on one or more of the storage devices;

determining that a coalescing operation is not warranted for the other data page; and

in response to said determining, refraining from performing a coalescing operation for the other data page.