Performance Degraded After v8 Upgrade

[EricinIT]

Hi,

Prior to using Veeam Backup for M365 v8, my backup proxies were at times using close to 100% CPU during our busiest backup times late evening. However, after the v8 upgrade, the CPU never reaches 100% anymore and is basically under 20% even with many jobs running. In the VBM365 console, most if not all jobs show "Objects waiting for processing" with a count. It's like the jobs are just queuing and processing only a set number of objects which no longer uses the full amount of resources available to it anymore. I did put all the repositories into a Proxy Pool and added a second proxy to it to see if that'd help. However, that didn't make any difference from what I can tell and objects still go to Objects waiting for processing. It no longer seems to be using the full resources available anymore. Also, the jobs seem to run way longer as well into the day from what used to be completing by early morning.

Eric

[bobbylwest]

I would tend to agree. I have a new install of v8, with proxy pool and its like watching paint dry most of the time. From what I can tell the Postgres DB is busy doing a bunch of stuff but the actual fetching of data from M365 is taking forever... Especially Onedrive/SharePoint data.

[EricinIT]

Yes, SharePoint and OneDrive have always been throttled. The workaround was to use auxillary applications to increase throughput on that. However, I've noticed that v8 goes into this "Organization throttling" mode quite frequently which I've not noticed it doing in v7. This seems to reduce processing performance considerably. Is there any method to prevent this?

Eric

[EricinIT]

After doing a much testing with the Veeam Backup for M365 v8 product, what helps is to scale the proxy pool to have more smaller proxies instead of a few large proxies in it. This seems to get the 64 threads per proxy and use more CPU than with a couple large proxies where the CPU usage stays quite low.

Eric

[Polina]

Hi Eric,

The following considerations can help you with sizing your proxies in v8:
- CPU/RAM ratio should be planned as 4 GB of RAM per each 1 CPU core (i.e. 8 CPU/32 GB RAM)
- Threads number changes dynamically, with a minimum of 4 threads allocated per each 1 CPU core and a maximum of 12 threads per core (i.e. for an 8 CPU proxy you get a minimum of 32 threads which can scale to up to 96 threads)

Hope it helps

[benoit.rey]

Hi,

The performances since our transition to v8 have been also very degraded (expecially on Exchange and Sharepoint jobs for one specific customer).

We have jobs that used to take 2 hours that now take more than 20 hours.

We also have a big increase in copy job execution time (Azure to Azure).

Two tickets opened without real response to date.

Benoît

[Polina]

@benoit.rey

Please post the IDs of your support cases.

[benoit.rey]

@Polina 07505156 & 07509455