Hello @Farmer92 Farmer92

@Onno van den Berg - do you have any general advise or anything to share? 

As a person in Commvault support i am not really able to answer this one haha.

Kind regards

Albert Williams

I am huge advocate for Object Storage, having used offerings from NetApp, Scality and VAST.

Across multiple environments, I have seen much better performance than our traditional SMB-based libraries, and the ability to use features such as Object Lock allow us to better secure our backup data.

If I would to be pick new storage for on-premise I would always go for a solution that supports object and it preferably is also designed for object. The ability to easily segment it into a separate security domain with external user/authentication is so much safer than local storage that is attached to the system. In additional it is easier to for example deploy features like storage accelerator to improve performance even further as you do not have to process all the data by a pre-defined set of MAs. 

So as ​​​​​​@Erase4ndReuseMedia already mentioned some solutions: I would look into solutions like NetApp StorageGRID, Pure Storage Flashblade, Scality etc. The first two mentioned offer great solutions based on QLC SSDs these days that deliver so much performance from a small footprint which is really nice. Yes, it is initially a bit more expansive than a disk-based solution but the RTO is so much better not to forget the much lower power consumption and better durability. 

   Thank you folks.  I am surprised by the object store recommendation.  From things I have been reading there is quite a performance hit for object(S3) vs NFS or SMB.
   Any concerns with deduplication and data aging with object stores?   There would be no micropruing right?  Do you think it would consume more space on Object storage vs block attached SMB?

Thank you again for the great feedback.  I am hoping to come up with a new local MAG library option pretty soon.

@Erase4ndReuseMedia  FYI.  I love your job title.   I am a huge fan as well.  Long live the Java console!  Or at least make a web gui that is just like it.

Hey @Farmer92 

I am surprised by the object store recommendation.  From things I have been reading there is quite a performance hit for object(S3) vs NFS or SMB.

I had expected a performance hit when we moved to StorageGRID, especially as there were known performance issues at the time, but backup and restore performance ended up being significantly better than our NetApp FAS-based SMB Disk Libraries - there were definitely design choices and environmental factors that contributed to that outcome, though.

There were issues with DASH Copy performance to and from non Object-based Copies on account of the mismatch in DDB Default Block Sizes, but that was easily overcome. 

 Any concerns with deduplication and data aging with object stores?   There would be no micropruing right?  Do you think it would consume more space on Object storage vs block attached SMB?

I can't say I noticed any meaningful negative impacts to deduplication, data aging, or even capacity - but at petabyte scale, it would have needed to be significant for me to really notice. As for micro pruning, that will be dependent on the platform - you can check out the Supported Cloud Storage Products page to see which platforms support micro pruning.

I love your job title.   I am a huge fan as well.  Long live the Java console!  Or at least make a web gui that is just like it.

I genuinely believe they should have followed the VMware route and developed a HTML5 version of the CommCell Console (whilst also incorporating some of the great features in Command Center), rather than alienating long-term customers with what feels like cheap imitation of Commvault.

Been working with commvault now for 10 years, have installed multiple different solutions. alltough all based on servers/mediaagents and just attached disk shelfs as the solution grows for the most of it.

A simple solution that works fine.

Have also experiended commvault with storeonce witch was a nightmare, so i would not reccommend a dedupe appliance with commvault. My conclution is that you will loose much of the already built in features for data aging, job priotity etc going down the route with a dedupe appliance.

Now we are running HyperScale X, for us it works really well, we have over 7pb of stored data devided on several clusters. We are running hpe servers and not commvault appliances.