Running out of repository space, what's the best option?

[ROSA]

We are running out of space on the Veeam Backup repository and I'm trying to figure out the best way to expand the repository.

Our VBR server acts as all in one (repository, proxy).
It's a HP ProLiant DL380e Gen8 server, running Windows Server 2012 R2 Standard (x64).
Actually there is a 500 GB mirror for the OS and VBR and a 40 TB array used as Veeam Backup repository, all HDD bays of the server are occupied.

Ideas so far:
1) replace the current drives with bigger ones.
Problem: Actually there are 4 TB drives, the max the controller supports is 6 TB- that isn't really worth it.

2) add another RAID controller card and an external storage box with HDDs connected to this new controller.
Problem: it seems there is no free slot in the server (4 slots, 3x FC card, 1x Smart Array P822)
Replace the actual RAID controller with a new better one?

3) A new server which will act only as a repository.

4) Any other idea?

What would you recommend?

[DGrinev]

Hi Rosa,

Can you provide details about the backup jobs, retention values, backup methods, that should help us to give an advise not only from a hardware perspective. Thanks!

[ROSA]

18 Jobs, 141 VMs.
The original retention was 28 restore points, I had to lower it to 21 (with this setting I have 13.8 TB free (40 TB total).
Backup is running once a day, everyday, incremental, with an active full backup once a week (on Friday).
Secondary backup target is a LTO5 tape library.
The server has 3 FC cards in it - LAN, SAN, iSCSI
21 VMs are Hyper-V running on a HP StoreVirtual 4730 (iSCSI)
The rest of the VMs are VMWare running on NetApp FAS 8020 (FC)

If you need any additional information, please let me know.

[ASG]

Hi,

the DL380e G8 should have 5 or 6 PCIe slots - if you install a second CPU (if not installed) and install the second PCIe riser (quick google shows 661628-B21 as P/N). That way you can have additional cards (external SAS for example) - but the second riser will only work if the second CPU is installed.

Bye

Edit:
That is if you don't have the rear drive cage installed... If it's installed than there is no space in the G8 iirc

[ROSA]

Good idea.

Sadly I just realized, there are 2 drive bays, where the additional cards would be.
There are 14 HDDs total, 12 on the front, 2 in the back.
2x500 GB SAS (mirror) - OS, VBR
12x4 TB SATA (RAID) - Veeam Backup repository

[SimonS]

2) add another RAID controller card and an external storage box with HDDs connected to this new controller.
Problem: it seems there is no free slot in the server (4 slots, 3x FC card, 1x Smart Array P822)
Replace the actual RAID controller with a new better one?

SmarArray P822 (https://h20195.www2.hpe.com/v2/getdocum ... =c04111513) have 4 external SAS ports, so you can add external disk enclusure (https://h20195.www2.hpe.com/v2/GetPDF.a ... 111697.pdf) put in disk and create new or extend existing array.

[ROSA]

That's a very interesting idea, thank you.

[TaylorB]

If you compare costs, it is often not a whole lot more expensive to just buy another server with 8 or 12 LFF slots vs buying an expensive external RAID controller and disk shelf. Then you also get the benefit of additional processing power and an additional proxy to go along with your extra disk space. The actual disks themselves will cost the same, regardless, which is the bulk of the cost. Veeam is designed to be distributed, so having multiple proxies and multiple repositories is not a problem and can provide better performance vs stacking it all on one massively crowded server.

[ROSA]

That's a good point.
It gets a little complicated because the FC.
I have to admin I have no idea, how to connect the new server, so it doesn't slow down the backups ....

[Stokkolm]

The best option is to get a deduplicating storage appliance. Here are the current stats from my HPE StoreOnce 5100:

User Data Stored 414.94 TB
Size On Disk 58.72 TB
Deduplication Ratio 7.07:1
Capacity Saved 85.85 %

It might be useful to note that the dedup rate is about 33% better than I was getting with the built-in VBR de-duplication. Also, restores are much faster in my experience.

[bulletlynn]

If you upgraded to Server 2016, reformat drive to ReFS and switch to synthetic fulls you would free up quite a bit of disk space.

[ROSA]

The actual situation looks like we are buying a whole new server.
It'll be a HP Apollo 4200 Gen9 with 12x8TB HDDs and later we will buy another 12x8TB.
There will be Server 2016 on it, so I'll definitively try ReFS.