Different backup speed for same VM

[ta_yoshizumi]

Hello,

Case number is 03291316.

I have a question about backup speed.

When we conducted the test in two groups according to the backup application, the following speed result was obtained.
- Group Job1:Backup 120MB/sec, Restore 120MB/sec
- Group Job2:Backup 60MB/sec, Restore 120MB/sec

These job's information is below.
- Repository is not same each jobs
- Specification of proxy server is different
　 - Group Job1:8vCPU, 24GB memory
　 - Group Job2:4vCPU, 12GB memory

I can understand that if proxy1 upper specification is the faster than proxy2, using proxy1 backup speed is fast but the reason is not known because the restore has the same speed. Has the same phenomenon been confirmed in the past?

Thank you.

[Andreas Neufert]

No, but there could be several root causes for this.

Let´s start with the Job bottleneck analyse, what do the %tage show in each job?
What is the backup mode per each disk at each job? Go to the Backup Job statistics and select the VM. Per VM disk you see the backup mode (NBD/hotadd/directSAN/...)

[ta_yoshizumi]

Hello, Thank you for comment.

All backup mode for jobs is hotadd.

The same VM targeted for backup, the listed proxy server, the repository are different, the backup speed is only half the speed compared with one job, but the restore speed is the same
What are the causes of these issues?

In fact, Group1 and Group2 has each 3jobs.
See details about bottleneck is below:
Group1:Job A - Bottleneck is Source, Job B - Bottleneck is Network, Job C - Bottleneck is Network
Group2:Job A - Bottleneck is Network, Job B - Bottleneck is Source, Job C - Bottleneck is Source

After contacting the Veeam, there was a request to upload a copy of Veeam Database (SQL Server) to FTP.
However, because this case environmentI is not mine, it is our customer environment, so I can not get a database copy for no reason. Why does Veeam support need Veeam Database for this issues?

[Andreas Neufert]

It is the fastest way to analyse what is happening here.

What is the backup mode for the disks ? Do one of the Jobs run in NBD mode while the other run in network mode?

Do you have configured Antivirus exceptions according to this KB ? https://www.veeam.com/kb1999

[ta_yoshizumi]

No, All backup mode for disk is hotadd.

Is antivirus affected the data reading speed at the time of backup, or execution slows down?

[Andreas Neufert]

yes, sometimes. So you need to make the KB1999 mentioned changes on the Veeam Servers and backup target servers.

[ta_yoshizumi]

If antivirus does not installed on VM?

I have additional question for this issues.

I received the following inquiries from our customer.
- It seems that there is a speed difference according to the job execution source (B & R / Enterprise Manager).
Are jobs created with Veeam B & R and Enterprise Manager reading data slower for jobs created with Enterprise Manager?

I am inquiring at case number 03291316, but there is no response at all and the assignment is not assigned.

[foggy]

I can see that the case is being worked on by our engineers. That said, while in your case performance difference is most likely explained by environmental reasons (bottlenecks are different), it is quite expected to see various processing speeds even in case the same proxies/repositories are involved in processing of different VMs due to the reasons explained here.

[ta_yoshizumi]

Thank you, finally the case were worked.

In the linked forum, the possibility that CBT was not validated was shown, but all jobs tested and verified in my customer's environment are all undertaken with full backup.

We created and executed jobs on the Veeam console and Enterprise Manager with the same settings for the proxy server, repository server, and job, but the results of slow reading of data for all jobs created with Enterprise Manager are displayed It was.

Even when a job created with Enterprise Manager was executed from the Veeam console and a job created by the Veeam console was executed from Enterprise Manager, the data reading speed of only jobs created by Enterprise Manager was slow.

Have you confirmed the phenomenon that the data reading speed is slower for jobs created with Enterprise Manager than jobs created with Veeam console?

Thank you for reading.

[foggy]

This shouldn't be the case, since EM is not involved in actual data processing. Please continue investigating with support, there must be some difference in jobs setup.

[ejenner]

Surely there are too many variables in a case like this to be able to say for certain that the speeds you are seeing at the console at any given moment are the actually representative of the average speed of the backup. I know if you're backing up to ReFS formatted volumes as an example... you'll see varying times depending how ReFS is receiving and processing your files. Does the disk you're writing to perform any kind of optimsation at all? That would be unusual as most storage methods will have a variable write speed for one reason or another. i.e. trying to write to a heavily fragmented disk is going to take longer than the same disk if it was blank.

The best check would be to use a stopwatch and to always start with blank disks and blank databases.

What issue is this causing anyway?