Solved

Azure DR in depth


Userlevel 1

Hi,

I have been trying to understand how the “Restore to MS Azure Iaas” works. I need to know hos this works so i can bring up this feature with our customers.

I have an Azure Storageaccount as a SOBR + my local NAS.

I hade some failures to achieve a Azure to Azure restore before i found out that restores to Azure wont work when my Performance Tier is online. The process didnt restore FROM Azure as i wanted to test out.

When i put the Perfomance Tier in maintenence mode i got it to work! (Thanks Veeam Foums for this!)

What i furthermoore found out is that the restore speed from the storage account is far away from the miximum perfomance of 60Gbps / 7500 MB/s.

Restore Speed:

My restore speed was 169 MB/s,. This can ofcourse be a “OK” speed but i cant find out what the maximum speed can be or what the bottleneck could be for restoring.

 

Conversion Speed:

After the data is restored from the Azure storageaccount the data is then converted into a VM.

 

 

My concern is really about 2 things - “Documentation” and “How to increase Restore Speed” but i will explain my quesstions in moore depth here:

  1. Please update guides on your Help Center (that is really good with good pictures) but lacking things explaining the restore process as described above, Restore from Capacity Tier to Azure and what “Links-in-the-Chain” can be improved or make the chain work better.
  2. I would like to get moore information on how to increase the Restore Speed.
    1. How is different Storagaeaccounts and settings for this in Azure involved in the restore speed?
    2. What can be done do improve restore speed and what are the bottlenecks?
    3. Can the restoreproxy in Azure be optimized for this?
    4. What is the theoretical maximum restore speed in Azure?
  3. How can the Conversion Speed be improved?
    1. What can be optimized for making a VM conversion faster?
    2. What is the bottleneck for the conversion?
    3. Is the restoreproxy VM involved in this to?
    4. If so, is there any information on how moore vCpus/Ram will increase the conversion process or is there any other tweaking that can be done here?
  4. What would the restore speed be if the data was in a Archive Tier?
    1. Will Different Tiers and Storageaccounts give different performances for restore?
  5. There should be some moore flesh-to-the-bones here i think.
    1. Even a comparison between Azure, Goggle, Amazon?
    2. Comparision focus on Speed vs Cost?
    3. Will a Hot Storagaaccount be better in any way or is that just overkill when thinking of a DR solution? Because the restore speed is somehow limited do factors i would like to know moore about...
  6. Can the Azure Appliance be of any benefit in this restore scenario?
    1. I did this restore without Azure Aplliance, just using the Backup Proxy in Azure
  7. What will happen if I restore several VMs as in parallell restore jobs?
    1. Is it possible to make multi restores/parallell restores with 1 azurerpoxy or do i need a restoreproxy per restore job?
    2. What will happen with the restore speed and conversions with mutliple restores jobs ?
  8. How will the same process and performance be in “Amazon EC2” and  “Google CE”
    1. Is other clouds better/faster/cheaper for restore?
  9. Therse are my biggest concerns and questions after doing some tests with restore to “MS Azure Iaas” and I would like to have moore indepth information on this matter.

Ofcourse, I understand that Azure maybe a “blackbox” but surely there should some moore information, tips and tricks, or maybe someone could bring some light on this to optimize the restore process a bit.

I have also been in contact with Microsoft on this question and they say, “Veeam is a ‘Black-Box’ and we dont know how they are doing “there-thing” and cant really explain anything but maximal transfer speed for your storageaccount is 7500MB/s and you cant optimize anything for the storageaccount, please contact Veeam for explinations”

 

Thanks in advance (and sorry for any misspellings)

Jonny Arowell

 

 

 

icon

Best answer by wesmrt 21 March 2023, 15:35

View original

3 comments

Userlevel 7
Badge +20

I would suggest to put these talking points and suggestions in the Veeam Forum where they will be seen by management.  They are good here for discussion but if you are trying to get things updated/changed then the forums is the place - https://forums.veeam.com/

Userlevel 7
Badge +4

Hello Jonny,

I see you have a lot of questions.
I will try to answer what it's possible without knowing the envioprement and without see any logs. For a deep investigation you should open a case at Veeam Support. Or, as Chris said, you can post the same question in our Veeam Forum.

 

1) We have this article in our User Guide: https://helpcenter.veeam.com/docs/backup/vsphere/restore_azure_hiw.html?ver=120
It's explain step-by-step the restore process.

 

2) The size of the Azure Restore proxy could impact the restore speed: https://helpcenter.veeam.com/docs/backup/vsphere/restore_azure_proxy_size.html?ver=120
I suggest increasing the size of the machine on Azure and see if you have any speed increased.

 

3) During the conversion process we need to convert the disk to VHD format supported by Azure. The time for convert the disk will depends of the size of the disk. This process is maded by VBR server.

 

4) The process will be the same, but with Archive Tier you have to retrieve the data first and this can take some time.

 

5) From my experience, all the public clouds will have similar results.
You won't have any difference if the backups are located in a hot or cool storage account.

 

6) Azure appliance is used only to backup/restore Azure VMs, so it's not applied in this situation.

 

7) It's possible to restore in parallel, but the Azure restore porxy will always follow the 1 CPU = 1 task = 1 disk been processed.
More than 1 azure restore proxy can definitely improove if you are restoring multiple VMs.


 

Userlevel 1

Thanks Wesley for your inputs and feedback!

In point (3) you state that the VBR server is involved in the conversion of the VM, thats interesting!

I thought the proxy was involved in this and made all the conversion.

This means that if we during conversion have a OnPremise “outage”/Problem the restore will fail.

I thought it was like:

  1. Start Restore Process from VBR Server
  2. Azure Proxy VM takes over and do the restoring and conversion.
  3. Proxy Reports back to VB

Good to know!

 

From the link and Veeam Help Center i can see this image on “How Restore to Microsoft Azure Works”

It could be improved with also adding the process for Restore from “Capacity Tier/Archive Tier” to Azure or maybe i have missed this process from anywhere. 

Thanks for the feedback again

Comment