Using ReFS features in Veeam Backup & Replication 9.5
A recent post dealt with Veeam Backup & Replicaion 9.5, which is scheduled for release in the fourth quarter of this year. Today I will focus in more detail on the declared extended support for ReFS.
The Resilient File System (ReFS) file system is recommended for Windows Server 2016, and our developers did not fail to take advantage of its new features. What was the focus of the new version of Veeam Backup & Replication, and what are the results? Very positive: using ReFS in backup repositories significantly improves the reliability of data storage, significantly speeds up block-by-block merging of increments on storage systems, and also allows you to store more recovery points by using links to identical data blocks.
For details, welcome under cat.
')
First of all, ReFS volumes will be supported on internal Direct Attached Storage Systems (DAS) and on Storage Spaces storage systems - both in their traditional form and deployed using Storage Spaces Direct (S2D). This will make it possible to significantly speed up the creation of synthetic full backups and the conversion of backups in the chain, as well as improve reliability and reduce disk space consumption. Let us dwell on these advantages in a bit more detail.
One of the most “intricate” operations in creating a backup, in particular, a synthetic full backup, is a transformation (transformation at the block level), during which the blocks are moved between different files within the incremental chain. For more, see the article of the knowledge base Veeam , as well as a post on Habré .
In the new version of Veeam Backup & Replication, the fast cloning technology (via ReFS API) will be used to create "synthetics" - this will allow not to move blocks between backup files, but to operate with links to the blocks stored on the volume. Thus, all operations associated with the "synthetics" are reduced to updating the meta-data and allow you to do without directly reading and writing data from the backup.
As can be seen from the example below, the creation of a synthetic full backup for a VM backup copy on ReFS took 58 seconds, while 3 incremental backups were processed. If you do the same for the backup stored on NTFS, then it will take more than 36 minutes - as they say, “feel the difference”!
Since the creation of “synthetics” and the transformation operation have now significantly accelerated, and the performance of the backup storage has increased, why not save more full backups to disk? For example, add more GFS backups there. Yes to health! Indeed, thanks to integration with ReFS, it is now possible to share identical blocks for several full backups. This will, of course, reduce the capacity requirements of the backup repository and, to a certain extent, come close in efficiency to storage systems with built-in deduplication (happily avoiding, however, performance degradation due to data rehydration in such systems). And what is most welcome is that storage systems can be built on the basis of a typical Windows Server 2016 using internal disks or disks with direct connections. An example of a synthetic full backup in the ReFS repository:
Let's talk about reliability, which, along with performance and optimization of storage disk space usage, is the key to integration with ReFS. Veeam Backup & Replication 9.5 will automatically activate Data Integrity Streams for all backup files with which the ReFS data integrity scanner works. This will reveal and signal implicit (silent) violations, which is especially important when using the direct incremental-endless backup method, and you need to guarantee the integrity of data with a long shelf life. It can be said that our engineers have "enlarged" the already existing data integrity checking functionality at the storage level, and now it has spread not only to the last restore point, but also to all stored backups.
The following is an example of a Windows Event Log entry when a breakdown of data integrity in a file on a simple ReFS volume is detected:
All of these features will work with regular ReFS volumes - but for those who use Storage Spaces, even more opportunities open up. After all, such systems allow you to take measures to repair damaged blocks, using a Mirror or Parity-copy of the blocks. Thus, it is possible to ensure the integrity of the data in an infinite-incremental backup chain.
How to achieve all this beautiful future? Very simple: configure the backup repository on Windows Server 2016. (Many use the same server as for the Veeam backup server.) As a result, the ReFS API will be available, and, accordingly, all the functionality described above.
» Article on Habré with an overview of new features and products Veeam
» An article about backup methods Veeam
» Veeam Availability Suite 9.5 Announcement
The Resilient File System (ReFS) file system is recommended for Windows Server 2016, and our developers did not fail to take advantage of its new features. What was the focus of the new version of Veeam Backup & Replication, and what are the results? Very positive: using ReFS in backup repositories significantly improves the reliability of data storage, significantly speeds up block-by-block merging of increments on storage systems, and also allows you to store more recovery points by using links to identical data blocks.
For details, welcome under cat.
')
First of all, ReFS volumes will be supported on internal Direct Attached Storage Systems (DAS) and on Storage Spaces storage systems - both in their traditional form and deployed using Storage Spaces Direct (S2D). This will make it possible to significantly speed up the creation of synthetic full backups and the conversion of backups in the chain, as well as improve reliability and reduce disk space consumption. Let us dwell on these advantages in a bit more detail.
Fast cloning technology
One of the most “intricate” operations in creating a backup, in particular, a synthetic full backup, is a transformation (transformation at the block level), during which the blocks are moved between different files within the incremental chain. For more, see the article of the knowledge base Veeam , as well as a post on Habré .
In the new version of Veeam Backup & Replication, the fast cloning technology (via ReFS API) will be used to create "synthetics" - this will allow not to move blocks between backup files, but to operate with links to the blocks stored on the volume. Thus, all operations associated with the "synthetics" are reduced to updating the meta-data and allow you to do without directly reading and writing data from the backup.
As can be seen from the example below, the creation of a synthetic full backup for a VM backup copy on ReFS took 58 seconds, while 3 incremental backups were processed. If you do the same for the backup stored on NTFS, then it will take more than 36 minutes - as they say, “feel the difference”!
Optimization of disk usage when storing full backups
Since the creation of “synthetics” and the transformation operation have now significantly accelerated, and the performance of the backup storage has increased, why not save more full backups to disk? For example, add more GFS backups there. Yes to health! Indeed, thanks to integration with ReFS, it is now possible to share identical blocks for several full backups. This will, of course, reduce the capacity requirements of the backup repository and, to a certain extent, come close in efficiency to storage systems with built-in deduplication (happily avoiding, however, performance degradation due to data rehydration in such systems). And what is most welcome is that storage systems can be built on the basis of a typical Windows Server 2016 using internal disks or disks with direct connections. An example of a synthetic full backup in the ReFS repository:
Data Integrity Streams Protection Method
Let's talk about reliability, which, along with performance and optimization of storage disk space usage, is the key to integration with ReFS. Veeam Backup & Replication 9.5 will automatically activate Data Integrity Streams for all backup files with which the ReFS data integrity scanner works. This will reveal and signal implicit (silent) violations, which is especially important when using the direct incremental-endless backup method, and you need to guarantee the integrity of data with a long shelf life. It can be said that our engineers have "enlarged" the already existing data integrity checking functionality at the storage level, and now it has spread not only to the last restore point, but also to all stored backups.
The following is an example of a Windows Event Log entry when a breakdown of data integrity in a file on a simple ReFS volume is detected:
All of these features will work with regular ReFS volumes - but for those who use Storage Spaces, even more opportunities open up. After all, such systems allow you to take measures to repair damaged blocks, using a Mirror or Parity-copy of the blocks. Thus, it is possible to ensure the integrity of the data in an infinite-incremental backup chain.
How to achieve all this beautiful future? Very simple: configure the backup repository on Windows Server 2016. (Many use the same server as for the Veeam backup server.) As a result, the ReFS API will be available, and, accordingly, all the functionality described above.
What else to read and see
» Article on Habré with an overview of new features and products Veeam
» An article about backup methods Veeam
» Veeam Availability Suite 9.5 Announcement
Source: https://habr.com/ru/post/311034/
All Articles