Mechanism to run oltp workload on in-memory database under memory pressure

Patent metadata

Field	Value
Publication number	US-2016140206-A1
Application number	US-201615005610-A
Country	US
Kind code	A1
Filing date	Jan 25, 2016
Priority date	Sep 21, 2013
Publication date	May 19, 2016
Grant date	—

Abstract

Techniques are provided for maintaining data persistently in one format, but making that data available to a database server in more than one format. For example, one of the formats in which the data is made available for query processing is based on the on-disk format, while another of the formats in which the data is made available for query processing is independent of the on-disk format. Data that is in the format that is independent of the disk format may be maintained exclusively in volatile memory to reduce the overhead associated with keeping the data in sync with the on-disk format copies of the data.

First claim

What is claimed is:

1 . A method comprising:
maintaining, by a database server, at least two copies of each item in a particular set of data; wherein, for each item in the particular set of data, the at least two copies include:
a first copy, stored on persistent storage, in a persistent format, and
a second copy, stored in volatile memory, in a mirror format;
wherein the particular set of data includes a first item and a second item; when a particular transaction performs a first operation that makes a change to the first item, the database server:
making the change to a copy of the first item that is in the persistent format; and
storing, in an in-memory journal, an entry that indicates the change without updating any copy of the first item that is in the mirror format;
after the particular transaction performs the change, when the particular transaction performs a second operation that involves the first item and the second item, the database server:
obtaining a copy of the first item, which is a current version of the first item, based on the entry in the in-memory journal; and
obtaining a copy of the second item, which is a current version of the second item, from volatile memory in the mirror format.

2 . The method of claim 1 , wherein:
the in-memory journal is a private journal; and the method further comprises, upon commit of the particular transaction, moving one or more entries, from the in-memory journal, to a global journal.

3 . The method of claim 2 , further comprising:
after commit of the particular transaction, a second transaction obtaining a copy of the first item, that is the current version of the first item, based on a particular entry in the global journal; wherein the second transaction is distinct from the particular transaction.

4 . The method of claim 1 , wherein:
the second operation involves a third item; and the method further comprises, when the particular transaction performs the second operation that involves the first item and the second item, the database server obtaining a copy of the third item, which is a current version of the third item, from persistent format data.

5 . The method of claim 4 , wherein obtaining the copy of the third item from persistent format data comprises obtaining a cached copy of the third item from volatile memory.

6 . The method of claim 1 , wherein making the change to the copy of the first item that is in the persistent format comprises making the change to a cached copy of the first item stored in volatile memory.

7 . A method comprising:
maintaining, by a database server, at least two copies of each item in a particular set of data; wherein, for each item in the particular set of data, the at least two copies include:
a copy, stored on persistent storage, in a persistent format; and
a copy, stored in volatile memory, in a mirror format;
maintaining an in-memory journal with particular entries that reflect committed changes, wherein each change of the committed changes:
was made to a copy of an item, in the particular set of data, in the persistent format; and
was not made to any copy of the item, in the particular set of data, in the mirror format;
purging a portion of the particular entries from the in-memory journal; wherein the portion of the particular entries includes information for one or more changes to one or more items of the particular set of data; wherein purging the portion of the particular entries from the in-memory journal is performed without having applied the one or more changes to any copies, of the one or more items, in the mirror format; after purging the portion of the particular entries from the in-memory journal, providing, in response to a query, a copy of a particular item, of the particular set of data, in the mirror format.

8 . The method of claim 7 , wherein purging the portion of the particular entries from the in-memory journal is performed based, at least in part, on memory availability within the volatile memory.

9 . The method of claim 7 wherein:
the mirror format copies of items, in the particular set of data, belong to a set of mirror format data;
the set of mirror format data reflects values that were current as of a particular time;
the portion of the particular entries includes entries for changes committed between a first time and the particular time;
the first time precedes the particular time; and
the method further comprises:
before purging the portion of the particular entries from the in-memory journal, using the in-memory journal in conjunction with the set of mirror format data as a source for data items for transactions with snapshot times between the first time and the particular time; and
after purging the portion of the particular entries from the in-memory journal, using persistent format data, that includes persistent format copies of items in the particular set of data, as a source for data items for transactions with snapshot times between the first time and the particular time.

10 . The method of claim 7 wherein:
the mirror format copies of items, in the particular set of data, belong to a set of mirror format data;
the set of mirror format data reflects values that were current as of a particular time;
the portion of the particular entries includes entries for changes committed between a first time and a second time;
the particular time precedes both the first time and the second time; and
the first time precedes the second time;
the method further comprises:
before purging the portion of the particular entries from the in-memory journal, using the in-memory journal in conjunction with the set of mirror format data as a source for data items for transactions with snapshot times between the particular time and the second time; and
after purging the portion of the particular entries from the in-memory journal:
using the in-memory journal in conjunction with the set of mirror format data as a source for data items for transactions with snapshot times between the particular time and the first time, and
using persistent format data, that includes persistent format copies of items in the particular set of data, as a source for data items for transactions with snapshot times between the first time and the second time.

11 . The method of claim 10 further comprising after purging the portion of the particular entries from the in-memory journal:
identifying, based on a delete vector for a particular transaction with a snapshot time of a third time, one or more particular items that have not changed between the first time and the third time;
wherein the third time is before the second time and the third time is after the first time;
in response to identifying the one or more particular items that have not changed between the first time and the third time, using the in-memory journal in conjunction with the set of mirror format data as a source, for the one or more particular items, for the particular transaction.

12 . One or more non-transitory computer-readable media storing one or more sequences of instructions which, when executed by one or more processors, cause:
maintaining, by a database server, at least two copies of each item in a particular set of data; wherein, for each item in the particular set of data, the at least two copies include:
a first copy, stored on persistent storage, in a persistent format, and
a second copy, stored in volatile memory, in a mirror format;
wherein the particular set of data includes a first item and a second item; when a particular transaction performs a first operation that makes a change to the first item, th

Assignees

• Oracle Int Corp

Inventors

• Hase Sanket
• Raja Vivekanandhan
• Ganesh Amit
• Marwah Vineet
• Pendse Sukhada
• Su Shuang
• Mullick Atrayee

Classifications

• G06F16/23 Primary

  Updating · CPC title

• G06F17/30592 Primary

  Physics · mapped topic

• G06F17/30088

  Physics · mapped topic

• G06F17/30185

  Physics · mapped topic

• G06F16/283 Primary

  Multi-dimensional databases or data warehouses, e.g. MOLAP or ROLAP · CPC title