Comprehensive data protection for all workloads
Post Reply
Bockelbrink
Novice
Posts: 8
Liked: never
Joined: Dec 08, 2009 1:21 pm
Full Name: Marco Bockelbrink

Veeam Backup - Vstorage API / Virtual Appliance Mode

Post by Bockelbrink »

Hello

we uses the vStorage API in Virtual Appliance Mode. The Backup Server run in the VMWare Essentials Plattform and uses a Windows 2008 SMB Share for the Backup. The Backup Server has a 2 GB Network Link to the VMWare Infrastructure. We make the Backup Jobs in the Night where no one is online on the system.

we have transfer rates between 30-50mb/s at a fullback. Its okay. One Server a Fileserver only has 12-15MB/s. Why is the transferrate so differend between the machines? did the backup do an block device backup oder file backup? I mean did the the server backup the files in the vmware and now we have the problem that a file server has many small files we cant backup faster?

Hope some can help

bye
Marco
Gostev
Chief Product Officer
Posts: 31455
Liked: 6646 times
Joined: Jan 01, 2006 1:01 am
Location: Baar, Switzerland
Contact:

Re: Veeam Backup - Vstorage API / Virtual Appliance Mode

Post by Gostev »

Marco, what you see in session history is not a transfer rate, by instead VM processing rate. So it is also affected by the time it takes to create and delete snapshot, process small VM configuration files etc.

Backup processing speed mostly depends on the amount of changes since last backup. The more changes VM had, the more data needs to be processed and injected into backup file, so the slower backup processing speed is.

Yes, backup is at virtual disk block level, so it does not matter if VM has many small files. You just need to double check if the session history reflects that changed block tracking is leveraged to backup this particular VM, otherwise backup may be quite slow.

Thanks.
Bockelbrink
Novice
Posts: 8
Liked: never
Joined: Dec 08, 2009 1:21 pm
Full Name: Marco Bockelbrink

Re: Veeam Backup - Vstorage API / Virtual Appliance Mode

Post by Bockelbrink »

Thank you.

We don't understand why the windows 2003 mailserver with exchange 2003 if fast with 50mb/s and the windows 2003 fileserver only backup with 12 and 13 mb/s. It is the same sharedstorage and the same virtual server. All VM's uses Changed block tracking. it is activated for each virtual disc.

anybody an idea?

thank you & bye
marco
Gostev
Chief Product Officer
Posts: 31455
Liked: 6646 times
Joined: Jan 01, 2006 1:01 am
Location: Baar, Switzerland
Contact:

Re: Veeam Backup - Vstorage API / Virtual Appliance Mode

Post by Gostev »

Marco, as I've said above, backup processing speed mostly depends on the amount of changes since last backup. You can try to backup these 2 VMs in separate jobs and compare the VRB file size produced during the incremental backups. This will show you how much changed data blocks is generated for each VM every day.
Bockelbrink
Novice
Posts: 8
Liked: never
Joined: Dec 08, 2009 1:21 pm
Full Name: Marco Bockelbrink

Re: Veeam Backup - Vstorage API / Virtual Appliance Mode

Post by Bockelbrink »

Hello

sorry. I have forget to say we do a full backup each day! the customer want the full back each day. So the data changes is not the problem of the performance? or looks a fullback on the vm changes?
Gostev
Chief Product Officer
Posts: 31455
Liked: 6646 times
Joined: Jan 01, 2006 1:01 am
Location: Baar, Switzerland
Contact:

Re: Veeam Backup - Vstorage API / Virtual Appliance Mode

Post by Gostev »

Ah, OK. I did not realize you are talking about full backup - in this case CBT does not even matter. Does your customer realize that they will get full backup each day by doing regular incrementals? Because Veeam uses synthetic backup, the output file (VBK) always represents full backup file of the latest state. And backup speed will be many times faster this way to, compared to doing complete full backups daily. Not even talking about backup storage requirements.

Anyhow, as for full backup, speed can be affected significantly depending on these factors:
1. For VMs with thick disks: depending on amount of zeroed blocks in VMDK. Note that I am not talking about "free disk space", but actual zeroed VMDK block (where data was never written before). If VM has lotsa free disk space, "wiping" the disk will increase backup performance.
2. For VMs with thin disks: depending on provisioned vs. actually used space (processing rate is calculated off provisioned disk space).

Thanks!
Post Reply

Who is online

Users browsing this forum: orb and 188 guests