Avamar NDMP slow Level 0 backups and performance best practices
Summary: Avamar NDMP slow Level 0 backups and general performance best practices
This article applies to
This article does not apply to
This article is not tied to any specific product.
Not all product versions are identified in this article.
Symptoms
When using an NDMP Accelerator to backup NAS devices to Avamar, the initial backup is a Level 0 or full backup.
Cause
Various causes are possible.
Resolution
The L0 backup causes the NAS to send every file and directory on the volume being backed up. This can take a long time.
As the number of directories, files and data increases, the time for the NAS to analyze the data to send increases.
The current best practice indicates 10,000,000 files or less in a backup.
The Full or L0 backup must complete. Until it does, Avamar continues to run a full backup, and never moves on to the quicker incremental backup.
Cases have been observed where:
Case A
Best Practices for performance:
Never kill a level 0 backup if it can be allowed to continue running.
Ensure that "overtime" is set for the backup so Avamar will not kill it when the schedule time is reached.
Also note, that when starting a manual backup, the schedule limits will not affect the runtime, but if the backup takes longer than 24 hours, it is killed.
To avoid this, set the backup to use overtime to run until completion.
If a volume is moved to a different client to back it up another level 0 backup of that volume will be taken.
The new backup might complete a bit more quickly than the original L0.
This is because most data has already been sent to Avamar. The NAS will still send every byte for the volume to the NDMP Accelerator node. This is so that it can check if every byte of data is already backed up.
Isilon
If using Isilon there is only a single file system, /ifs.
When backing up at the /ifs directory level only 1 stream will be used.
If a directory subtree is moved between directories, the complete tree is traversed for changes. All the files are backed up as if it is a level 0 because it was not previously seen in that path.
The backup performance depends various factors including;
Usually, the NAS spends most of its time scanning the directory tree for data to send, than sending the data.
A wide tree with many, shallow directories, allows quicker completion than a deep tree.
An NDMP backup can run eight streams at one time.
Each entry in the dataset uses 1 stream up to the max-streams set in the dataset.
Using more streams allows the backups to be run in parallel and complete more quickly than a single volume was used.
The largest volume starts first to give it the best change of completing and the next largest and so on.
As the number of directories, files and data increases, the time for the NAS to analyze the data to send increases.
The current best practice indicates 10,000,000 files or less in a backup.
The Full or L0 backup must complete. Until it does, Avamar continues to run a full backup, and never moves on to the quicker incremental backup.
Cases have been observed where:
Case A
- A 7TB level 0 took 121 hrs but the next took only 4 hrs.
- Subsequent level 1 backups completed within ~45 minutes.
- A 40TB volume took close to 300 hrs to complete the level 0 backup.
- Subsequent Level 1 backups completed within around 4 hours.
Best Practices for performance:
Never kill a level 0 backup if it can be allowed to continue running.
Ensure that "overtime" is set for the backup so Avamar will not kill it when the schedule time is reached.
Also note, that when starting a manual backup, the schedule limits will not affect the runtime, but if the backup takes longer than 24 hours, it is killed.
To avoid this, set the backup to use overtime to run until completion.
If a volume is moved to a different client to back it up another level 0 backup of that volume will be taken.
The new backup might complete a bit more quickly than the original L0.
This is because most data has already been sent to Avamar. The NAS will still send every byte for the volume to the NDMP Accelerator node. This is so that it can check if every byte of data is already backed up.
Isilon
If using Isilon there is only a single file system, /ifs.
When backing up at the /ifs directory level only 1 stream will be used.
If a directory subtree is moved between directories, the complete tree is traversed for changes. All the files are backed up as if it is a level 0 because it was not previously seen in that path.
The backup performance depends various factors including;
- Number of directories in the tree
- The depth of the tree
- File size
- Type of file
- How busy the NAS disks are
Usually, the NAS spends most of its time scanning the directory tree for data to send, than sending the data.
A wide tree with many, shallow directories, allows quicker completion than a deep tree.
An NDMP backup can run eight streams at one time.
Each entry in the dataset uses 1 stream up to the max-streams set in the dataset.
Using more streams allows the backups to be run in parallel and complete more quickly than a single volume was used.
The largest volume starts first to give it the best change of completing and the next largest and so on.
Affected Products
AvamarProducts
Avamar, Avamar Plug-in for NDMPArticle Properties
Article Number: 000054368
Article Type: Solution
Last Modified: 17 Dec 2025
Version: 5
Find answers to your questions from other Dell users
Support Services
Check if your device is covered by Support Services.