Discussions related to using object storage as a backup target.
Post Reply
KarmaKuma
Enthusiast
Posts: 32
Liked: 6 times
Joined: Feb 05, 2022 11:16 am
Contact:

Question w.r.t v12 performance tier Object Storage

Post by KarmaKuma »

Hi

Not sure if this is the right place/subforum to ask, so mods please feel free to move this thread if it's inappropriate in here, or, change to pm if neccessary...

I successfully applied for the private v12 Beta and am trying to prepare myself and the PoC environment as good as possible. Can someone with insight to v12 shed some light into the inner workings of Object Storage as performance tier, specifically with regards to data flow.

How is the data transferred from the backup proxies to the Repository? Can each backup-proxy in a multi-proxy setup send the backup-data directly to the repository or has the data to flow via a "repository data mover", the same way it has to if one is using SMB/NAS repositories? Can a direct "many to one" data flow be achieved without possible repository data mover bottleneck or crosstalks between backup-proxies and repository data movers if one does not define an explicit repository data mover (where in worst case proxy a is the data mover for a task running on proxy b while proxy a itself is running backup tasks that send the data to the repository via proxy c)?

I am specifically asking for use cases where on-prem Object Storage is available and no (cloud) gateway or cache server is neccessary.

In our case it's two Dell PowerScale Clusters with S3 support (situated in two L2 connected datacenters), where it would be nice, if each backup-proxy could talk to the storage clusters (scale-out cluster) directly without having to pass via a specific or "per-task dynamic" data mover... ...kind of the same way it works with local repositories which recieve the backup data directly from the backup-proxies since their data mover logic is built-in.

How about copy jobs from one Object Storage to the second one?

We'll be running daily per-machine forever forward incremental chains with 30 days retention (so 30 restore points) and proper health check schedule, no GFS...
HannesK
Product Manager
Posts: 14287
Liked: 2877 times
Joined: Sep 01, 2014 11:46 am
Full Name: Hannes Kasparick
Location: Austria
Contact:

Re: Question w.r.t v12 performance tier Object Storage

Post by HannesK »

Hello,
the Veeam Systems Engineer who handed out the beta to you should be able to cover your questions (at least that's the official process for beta tests :-))

Correct, the easiest way to visualize data transfer to object storage is imagine it was a NAS repository. Automatic gateway selection works the same as today. For dedicated gateway selection, there is a new option to select multiple gateways (instead of only one up to V11)
In our case it's two Dell PowerScale Clusters with S3 support
I expect that to fail. Please use only "Veeam Ready" systems for production workloads. Isilon / PowerScale has limitations around maximum objects per bucket. So that might be okay for small tests, but not "production grade"

Backup copy jobs: same like NAS repositories today. See gateway server section in the V11 user guide. V12 improvement: multiple dedicated gateways can be selected

Best regards,
Hannes
PS: I moved your question to the object storage sub-forum, because it was an object-storage related question
KarmaKuma
Enthusiast
Posts: 32
Liked: 6 times
Joined: Feb 05, 2022 11:16 am
Contact:

Re: Question w.r.t v12 performance tier Object Storage

Post by KarmaKuma »

I will get access to the private Beta (the one thats close to release candidate) that will come out in a few weeks - not the current Beta 2...

Could you tell me how the multiple dedicated gateways/data movers functionality is organized in v12? A simple pool of possible workers that get to be chosen "at random" by proxies? An ordered list with a preferred/primary worker and a list of alternate workers? Completely different?

Is there a difference in how it works for SMB/NFS/Object data movers or is the logic the same for all Repository-types that need a data mover?
HannesK
Product Manager
Posts: 14287
Liked: 2877 times
Joined: Sep 01, 2014 11:46 am
Full Name: Hannes Kasparick
Location: Austria
Contact:

Re: Question w.r.t v12 performance tier Object Storage

Post by HannesK »

yes, it's a simple pool (round robin). once you have the software, it will be self explaining.

all logic around gateway servers is the same for SMB / NFS / Object / ddboost / catalyst
KarmaKuma
Enthusiast
Posts: 32
Liked: 6 times
Joined: Feb 05, 2022 11:16 am
Contact:

Re: Question w.r.t v12 performance tier Object Storage

Post by KarmaKuma »

thanks a lot for your answers so far. now hopefully last question:

Round robin on what level? Per job, per proxy, per vm or per task? I assume and hope there is some sort of load distribution ;)
veremin
Product Manager
Posts: 20270
Liked: 2252 times
Joined: Oct 26, 2012 3:28 pm
Full Name: Vladimir Eremin
Contact:

Re: Question w.r.t v12 performance tier Object Storage

Post by veremin »

It should be VM in this case, but let's postpone the detailed discussions about the v12 version until it's released. Thanks!
KarmaKuma
Enthusiast
Posts: 32
Liked: 6 times
Joined: Feb 05, 2022 11:16 am
Contact:

Re: Question w.r.t v12 performance tier Object Storage

Post by KarmaKuma » 1 person likes this post

thnx for the clarification, and understood ;-)
Post Reply

Who is online

Users browsing this forum: No registered users and 25 guests