Hi everyone,
I’m putting together a dedicated Tier-1 backup appliance (primarily Veeam) and would greatly appreciate your real-world experience and suggestions on the proposed config.
**Target:** ≥ 400 TB usable capacity, good ingest speed, solid reliability.
**Planned configuration – HPE Alletra Storage Server 4120 24LFF CTO**
- CPU: 2 × Intel Xeon-Gold 6542Y 2.9 GHz 24-core (48c/96t total)
- RAM: 512 GB (16 × 32 GB) DDR5-5600
- Boot: 2 × 960 GB NS204i-u (RAID-1 via onboard controller)
- Cache/Tiering: 2 × HPE 7.68 TB SAS 24G Read-Intensive LFF SSD (P49040-K21)
- Capacity drives: 22 × HPE 24 TB SAS 12G BC 7.2K LFF HDDs
- Storage controller: HPE Alletra Storage Server 41X0 SR 32-port PCIe (P61479-B21 = SR932i-p Gen11 equivalent, 8 GB cache)
- Networking: 3 × Broadcom BCM57414 10/25 GbE SFP28 2-port NICs (6 ports total)
- Redundant PSUs, iLO Advanced, etc.
**Planned RAID layout on the SR932i-p**
- 2 × 11-disk RAID-6 arrays striped together → **RAID-60**
- Usable capacity: 18 × 24 TB = **432 TB** (meets the 400 TB+ goal nicely)
- The 2 × 7.68 TB SSDs will be used either as ReFS FastFile cache (Windows) or as a Veeam SOBR Performance Tier / cache extent
**OS options**
- Windows Server 2022/2025 + ReFS + Veeam 12.3.2
- or Veeam Software Appliance (VSA Linux) v13.0.1
**Questions to the community**
1. **Throughput expectations**
With 22 × 7.2K SAS drives in RAID-60 and 25 GbE networking, what kind of real-world ingest/restore speeds have you seen on similar Alletra 41xx / SR932i-p builds with Veeam?
(Looking for synthetic backups, instant recovery, copy jobs, etc.)
2. **Optimisation suggestions**
- Is RAID-60 still the sweet spot with 24 TB drives, or would you go a different route (e.g., single big RAID-6 + hot spares, RAID-50, etc.)?
- For the 2 × 7.68 TB SSDs: ReFS FastFile (Windows) or SOBR Performance Tier/extent cache (Veeam) – which gives better results in practice?
- NIC bonding – RDMA/iSER worth it for Veeam, or just plain MPIO + SMB multichannel?
- Anything else I’m missing (firmware versions, specific BIOS tunings, etc.)?
Budget is pretty much locked to this chassis and drive count, so no room for all-flash or more CPUs.
Thanks in advance for any feedback or war stories!
Cheers
Marius
-
marius402
- Service Provider
- Posts: 65
- Liked: 9 times
- Joined: Oct 06, 2014 10:46 am
- Full Name: Marius Redelinghuys
- Contact:
-
pybfr
- Veeam Software
- Posts: 251
- Liked: 47 times
- Joined: Sep 26, 2022 9:54 am
- Full Name: Pierre-Yves B.
- Contact:
Re: Seeking feedback on HPE Alletra 4120 24LFF build – 400 TB+ usable for Tier-1 Veeam backup repository
There are no reason in this day and age to deploy ReFS repositories, using Veeam Hardened Repo ISO is a simpler, safer, cheaper option.
There are lingering issues with ReFS+Windows 2025:
veeam-backup-replication-f2/server-2022 ... 96912.html
Please note you will not be able to use your SSD as cache, except with specific RAID controller caching features I am not aware of,but it is not mandatory anyway to get excellent performance (see further down...) so you could spare this expense.
You dont need that much RAM, 256GB would already be plenty, neither do you need Xeon Gold class CPU.
RAID 60 is the best disk layout to use, since 20TB+ are extremely long to rebuild, with 2 or 4 arrays depending on your priorities
You can find quite a bit of info here on repositories design principles:
https://bp.veeam.com/vbr/2_Design_Struc ... ositories/
I would say this BOM is probably bit overkill, so if you need to, you probably can make it quite a bit cheaper.
For performance figure you can read this somewhat old thread (performance would be even better with current release), but full of interesting info:
veeam-backup-replication-f2/veeam-v11-n ... 71923.html
There is this HPe reference architecture whitepaper you may find useful as well and it is a bit more recent with performance figures:
https://www.hpe.com/psnow/doc/a50000150enw
There are lingering issues with ReFS+Windows 2025:
veeam-backup-replication-f2/server-2022 ... 96912.html
Please note you will not be able to use your SSD as cache, except with specific RAID controller caching features I am not aware of,but it is not mandatory anyway to get excellent performance (see further down...) so you could spare this expense.
You dont need that much RAM, 256GB would already be plenty, neither do you need Xeon Gold class CPU.
RAID 60 is the best disk layout to use, since 20TB+ are extremely long to rebuild, with 2 or 4 arrays depending on your priorities
You can find quite a bit of info here on repositories design principles:
https://bp.veeam.com/vbr/2_Design_Struc ... ositories/
I would say this BOM is probably bit overkill, so if you need to, you probably can make it quite a bit cheaper.
For performance figure you can read this somewhat old thread (performance would be even better with current release), but full of interesting info:
veeam-backup-replication-f2/veeam-v11-n ... 71923.html
There is this HPe reference architecture whitepaper you may find useful as well and it is a bit more recent with performance figures:
https://www.hpe.com/psnow/doc/a50000150enw
Who is online
Users browsing this forum: Semrush [Bot] and 17 guests