DDBoost, about network load balancing

[ita-tomi]

I have case where I have DD2200 with 4x 1Gbit interfaces and one Veeam server with one 10Gbit interface. I was wondering if I configure 3 interfaces into ifgroup for DDBoost in DataDomain (leave one for management): Is Veeam able to split one backup session traffic over multiple links and achieve >1Gbit network speeds towards DD?

[dellock6]

Veeam relies on the underlying network stack of the windows machine where is executed, so you need to trunk the three network using solutions like lacp to make them appear like a single network.

[foggy]

Actually, LACP is not required, DD Boost client is able to use Load Balancing/Link Failover configured on the storage itself. Keep in mind, however, that a single stream will use a single link, so in v8 you need to have multiple jobs to be able to use multiple links (in v9, per-VM backup chain will allow to achieve that with a single job).

[ita-tomi]

Thanks, I suppose single session load balancing was DDBOOST v3 feature and thats coming in v9 with Veeam?

What comes to LACP in my understaning and experience it will load balance per session, so with single server we are again using only one 1Gbit for session. Another session may or may not go another 1Gbit link. And its more complicated since we need to have matching switch configuration.

[foggy]

DD Boost 3.0 is indeed coming in v9, however what I'm talking about is not related. Veeam B&R v9 will also provide an ability to have a separate backup chain for each VM in a job, so each VM will use a separate write stream to the target storage and a single job will be able to saturate all available links.

[VladV]

Just saw that the per-vm backup chain is an option on the repository not the backup job itself. I was hoping that this function was available on a per backup job basis so I can limit the vbk total size to under 1 TB and conform with Server 2012R2 deduplication requirements. Applying this to all jobs targeting the repository isn't something I wanna do due to relying on robocopy scripts for rotating drives.

[veremin]

Just create an additional repository (typically, a folder), enable corresponding option for it and point the required job to the newly-created repository with per-VM chain option enabled.