Extremly poor performance/transfer rate while backup

Just tested the same agent/client storage config and reached 50MB/s as soon as I use a vault without deduplication.

The shown speed is definitely incorrect >1GB/s measure myself is between 10-20MB/s

Deduplication will try and save you allot of bandwidth by performing the first level of deduplication from the agent, so any files that exist in the vault already will not be transferred to it again. Any new data that does not exist in the vault already will be transferred and then indexed for future backups.

Best practice is to backup a single server into your deduplication vault and wait for it to finish indexing/processing. Once that is done then backup the next machine and then the next one.. once the deduplication database is setup with a few backups you should notice your network traffic has been reduced during backups significantly.

Also ensure you have a fresh install of ABR11 and new/fresh dedupe vault created with ABR11 to get maximum benefits of some of the new features with ABR11, if you are using an older ABR10 vault I would suggest creating a new ABR11 vault as there is new back end databases being used, ability for ABR11 dedupe to use/access more RAM which is very important.

Do not worry too much about the network transfer speeds but rather concentrate more on total amount of data backed up and final archive size as this is where you are going to see significant reductions when using deduplication.

Also ensure your storage node that is running the dedupe vault has good I/O available and a decent amount of RAM for the server to perform the deduplication processing.

I would also strongly sugest reading through some of the documentation at the following location which does a great job at explaining the process, requirements and fine tuning your backups:
http://www.acronis.com/support/documentation/ABR11/#3349.html

All the best and please let us know how you go.

* vault was new ABR11 compatible only
* initial backups failed (just with an message stoppped)
* follow up backups were running endlessly

StorageServer:
* Storage is currently a RAID5 with 5 discs but will be expanded on a Adaptec 6805 controller - no disc load (<<10MB/s)
* A second (source) SSD R5 with 3 discs
* 2x 1GBIT (LAN + iSCSI traffic)
* storage server has 16GB RAM & 4 core yeon X3430 2.4GHZ Proc - CPU load was never more than 10%
* ABR11
* Windows StorageServer 2008 R2 SP1 as OS

ManagementServer & SQLServer:
* Windows Enterprise on virtual machine
* ABR 11
* SQLServer 2008 (for small DBs like Acronis)
* 4GB of Mem
* i7 4 cores
* connected to RAID5 RAID with 3 SSDs (storage server) (via separate LAN)

Clients:
* Windows 7 SO1
* ABR 11
* 1-2 GB RAM
* ABR11
* connected to RAID5 with 3 SSD (storage server) (via seperate LAN)

Question: What is the role of the SQLServer in that game - must it be highperformance - will it give me a performance gain if I seperate Management and SQLServer?

This morning I tried to stop the backups started yesterday night - no success. Stopp command was not executed.
Chekcing the activity log there was no activity logged
BUT checking the event log I get: A LOT network disconnected messages which I have no explanation for (the network as such is stable)

NOTE: even after restart of management, agent & storage server the task are shown as running!! I assume this status is out of the SQL DB. Another 15min later after pressing manual synchronize on the machine view the status was going to failure...

After changing the dedup settings from storage server only back to dedup also on agent it works faster.
The combination file based backup and dedup only on server seems not to work.

-> Lets wait for the next backup jobs to be executed before I post the final result

They have an excuse to _not_ work on Windows Storage Server - http://kb.acronis.com/content/5403 . But it's just an excuse.

will it give me a performance gain if I seperate Management and SQLServer?

I think it will just create more network traffic between AMS and SQL server.

All file based full backup worked after going to default dedeup instead of using pure central storage based dedup

dev-anon wrote:

They have an excuse to _not_ work on Windows Storage Server - http://kb.acronis.com/content/5403 . But it's just an excuse.

Yepp that is really only an excuse since storage server and non storage enterprise server are based on the same code. They just want avoid the testing since expensive to setup further testenv.

Update:
As statet the backup seems now to run without serious issues. But it is randomly very slow. Some backups take 30min some 5hours, transferring almost the same amount of data (3-10GB).
CPU, DISC AND network utilisation is almost not visible (on agent, AMS and storage server) so it is def. NOT an hardware performance issue.

Where do I find the exact tasks the management server is executing (expect central storage of backup plans and distribution). I am asking since it needs ~450MB RAM - for what?

Performance of new version is better than the the first ABR 11 version. GUI on AMS (backupplans) is MUCH better. Backupplans are now executed faster and more important stable (I get almost the same time for each repetition and no longer sometimes execution times of half a day)