Methods and apparatus for optimal journaling for continuous data replication
View Patent ↗Methods and systems for continuous data replication, including receiving a write transaction from a production site, the write transaction including raw data (RWNEW), a size indicator (SZ 1 ) for the size of the raw data, and a start address (ADD 1 ) for writing the raw data, copying the raw data (RWNEW) into a first journal data stream at a current journal address (JADD 1 ), and copying the size indicator (SZ 1 ), the start address (ADD 1 ) and the journal address (JADD 1 ) into a first journal meta-data stream.
1. A method of replicating data, comprising:
(a) receiving transactions comprising write data from a first data site to be backed tip at a backup storage site;
(b) writing the write data to a first position in a do data stream;
(c) writing metadata for the write data to a first position in a do metadata stream, wherein the metadata for the write data in step (c) comprises a first field to store a size of the write data, a second field to store an address at which the write data is to be stored in the backup storage site and a third field to store a pointer to an offset in the do data stream where the write data is located;
(d) reading data from a second position in the do data stream;
(e) reading metadata from a second position in the do metadata stream;
(f) reading data to be overwritten from a logical unit in the backup storage site, where location and size of the data to be overwritten is determined from the do metadata stream;
(g) writing the data to be overwritten to a first position in an undo data stream;
(h) writing metadata for the data to be overwritten to an undo metadata stream, the metadata for the data to be overwritten comprises a first field to store a size of the data to be overwritten, a second field to store an address at which the data is to be overwritten in the backup storage site and a third field to store a pointer to an offset in the undo data stream where the data to be overwritten is located; and
(i) writing the write data to the logical unit in the backup storage site using the second field in the do metadata stream.
2. The method according to claim 1 , further including undoing a write transaction by:
(j) reading data from the first position in the undo data stream and metadata from the first position in the undo metadata stream;
(k) reading data from the logical unit in the backup storage site that is to be overwritten, wherein location and size of the data is determined from the undo metadata stream;
(l) writing the data read from the logical unit in the backup storage site in step (k) to the second position of the do data stream;
(m) updating the do metadata stream;
(n) writing the data from the first position of the undo data stream to the logical unit in the backup storage site; and
(o) writing the metadata from the first position of the undo metadata stream to the logical unit in the backup storage site.
3. The method according to claim 1 , further including
omitting steps (d) and (e) where a number of the write data transactions from the first data site to be backed up per unit time is less than a predetermined amount.
4. The method according to claim 3 , further including performing steps (d) and (e) again when a queue memory is more than a given percent full.
5. The method according to claim 1 , further including omitting steps (f), (g), and (h) when a size of the do data stream is greater than a given value.
6. The method according to claim 1 , further including omitting steps associated with the undo data stream and undo metadata stream when the do data stream becomes larger than a given value.
7. The method according to claim 1 , further including providing a state machine to transition between a number of states having a given number of processing stages, wherein
a first state includes steps (a)-(i);
a second state includes omitting steps (d) and (e) where a number of the write data transactions from the first data site to be backed up per unit time is less than a predetermined amount; and
a third state includes omitting steps (f), (g), and (h) when a size of the do data stream is greater than a given value.
8. The method according to claim 7 , wherein the state machine has three nodes.
9. The method according to claim 1 , further including omitting steps (b)-(e) to provide a state in a state machine for transitioning between a number of states having a given number of processing stages.
10. An article, comprising:
a storage medium having stored thereon instructions that when executed by a machine result in the following:
(a) receiving write data transactions from a first data site to be backed up at a backup storage site;
(b) writing the write data to a first position in a do data stream;
(c) writing metadata for the write data to a first position in a do metadata stream, wherein the metadata for the write data in step (c) comprises a first field to store a size of the write data, a second field to store an address at which the write data is to be stored in the backup storage site and a third field to store a pointer to an offset in the do data stream where the write data is located;
(d) reading data from a second position in the do data stream;
(e) reading metadata from a second position in the do metadata stream;
(f) reading data to be overwritten from a logical unit in the backup storage site, where location and size of the data to be overwritten is determined from the do metadata stream;
(g) writing the data to be overwritten to a first position in an undo data stream;
(h) writing metadata for the data to be overwritten to an undo metadata stream, the undo metadata comprises a first field to store a size of the data to be overwritten, a second field to store an address at which the data is to be overwritten in the backup storage site and a third field to store a pointer to an offset in the undo data stream where the data to be overwritten is located; and
(i) writing the write data to the logical unit in the backup storage site using the second field in the do metadata stream.
11. The article according to claim 10 , further including instructions for undoing a write transaction by:
(j) reading data from the first position in the undo data stream and metadata from the first position in the undo metadata stream;
(k) reading data from logical unit in the backup storage site that is to be overwritten, wherein location and size of the data is determined from the undo metadata stream;
(l) writing the data read from the logical unit in the backup storage site in step (k) to the second position of the do data stream;
(m) updating the do metadata stream;
(n) writing the data from the first position of the undo data stream to the logical unit in the backup storage site; and
(o) writing the metadata from the first position of the undo metadata stream to the logical unit in the backup storage site.
12. The article according to claim 10 , further including instructions for
omitting steps (d) and (e) where a number of the write data transactions from the first data site to be backed up per unit time is less than a predetermined amount.
13. The article according to claim 12 , further including instructions for performing steps (d) and (e) again when a queue memory is more than a given percent full.
14. The article according to claim 10 , further including instructions for omitting steps (f), (g), and (h) when a size of the do data stream is greater than a given value.
15. The article according to claim 10 , further including instructions for omitting steps associated with the undo data stream and undo metadata stream when the do data stream becomes larger than a given value.
16. The article according to claim 10 , further including instructions providing a state machine to transition between a number of states having a given number of processing stages, wherein
a first state includes steps (a)-(i);
a second state includes omitting steps (d) and (e) where a number of the write data transactions from the first data site to be backed up per unit time is less than a predetermined amount; and
a third state includes omitting steps (f), (g), and (h) when a size of the do data stream is greater than a given value.