Minimal recommended hardware for agent use

[tacioandrade]

Tonight I went to do a test with the Veeam Agent simulating instances that I have in Digital Ocean and Amazon of 512MB and 1GB. For simulation instead of generating the backup of the whole system, I only selected the / etc and / home directories, because I believe that for the reduced number of files consumption would be lower.

When running the backup on a Ubuntu Server 14.04 VM with 512MB of RAM, the Veeam Agent consumed at least 80% of the available RAM, leaving the services virtually inoperable due to lack of resources.

In the same machine with 1GB ram the result was a little better, but even so the memory consumption made the access to only 1 page opening get quite compromised.

I know that generating incremental backup, compressing files, etc., consume many resources, but is there any way to compensate for this for using the Agent in small instances to be feasible?

Has anyone done the same tests and got the same or different results from mine?

[tacioandrade]

I think I've got a better understanding of how the different options of the Veeam Agent work.

When the backup is performed using the file level instead of disk inodes, the use of RAM increases dramatically, meanwhile when I modify the same job to backup the baremetal, the use of processor that goes to the top and that of RAM Remains low.

At least in my tests, the use of Entery Machine backup is the best option, because by the same token my services continued to run at a slower rate than when I ran the file-level backup.

[nielsengelen]

With the current status there is no option to modify it directly in regards of resource usage however you can take a look at external tools like cgroups or others.

[PTide]

Hi,

The module performance has been improved in the latest versions, I gave you a link via PM. What is you current kernel btw?

Thanks

[tacioandrade]

With the current status there is no option to modify it directly in regards of resource usage however you can take a look at external tools like cgroups or others.

I've heard of this functionality but never used it until today, maybe I do not test during the weekend (yesterday was done due to an insomnia crisis) hehehe

The good thing is that due to this small setback I discovered something interesting. =)

Hi,

The module performance has been improved in the latest versions, I gave you a link via PM. What is you current kernel btw?

Thanks

Great news that are already working with the part of optimization of performance, in general the companies only focus on this type of optimization after the application is ready, which turns out to be sad for the end user.

The VM I am running is a Ubuntu Server 14.04 that I had already installed here at home with the Kernel 3.13.0-63.

I have seen DM here and I will try to upgrade and run Job just to check how it is doing, but for the work you have done, we will certainly have good results.