Hello everyone,
I'm currently working/planing a new system to replace (multiple) existing storage servers.
I reached out for commercial consulting, but I only received one reply within a week, they can help me by the end of the year...so I'll see if someone around here can help me
I'll try to sum up most important fact & decisions so far:
The main question I'm struggeling with right now is what VM I'm supposed to choose and related to that, which storage system (file system itself + "smart" options like self-healing, deduplication, snapshots etc) it'll be in the end: ZFS, Gluster, Ceph, ...
Raid5/6 (software or even hardware) are probably not the best solutions, especially with rebuild times.
ZFS might be an option, although checksumming needs quite some CPU performance (which I was told to considerate by a colleague). I'm not sure about the SSD caching or how to do it in this matter.
As for the VM OS itself: depending on the solution either a blank linux (debian, ubuntu, centos) system and samba configured properly or a more out of the box experience with a NAS distro or something like Univention Corporate Server (which can smoothly be combined with the 2-3 Windows VMs) to manage access rights, quota etc.
Any help with designing the best solution would be appreciated.
I'm currently working/planing a new system to replace (multiple) existing storage servers.
I reached out for commercial consulting, but I only received one reply within a week, they can help me by the end of the year...so I'll see if someone around here can help me
I'll try to sum up most important fact & decisions so far:
- we currently have about 350TB of data stored on several servers, mostly Windows (without virtualization), some NAS systems
- we need to start with 250TB of fresh storage
- it's a almost pure file server system: no databases, just a lot of different dirs and storage locations
- the storage plan will include 4 shares in future (3x 50TB, 1x 100TB) as the main storage options available for our team
- storage content: mainly medium and big pictures and video content (working on pictures and video from Windows workstations with Adobe Software; different steps are done on different systems, that's why we need a file sharing solution and iSCSI/SAN etc. to the workstations themselves will not do)
- besides the storage server there will be 4-5 additional VMs (2-3 Win, 2-3 Linux) with internal tools (monitoring solutions and similar)
- the storage needs to be fast: it should be able to write and read with 1Gbyte/sec alsmost constantly (10Gbit SFP+ as limiting factor in our network between storage+workstations)
- I guess we write about 50TB of new data each month (copying projects, deleting old projects, beginning new ones; yes, that will wear down hardware, especially if we use SSDs for caching, but that is ok as long as monitoring works fine and replacing is easy)
- I'm still not sure about the backup solution: I'd love to test the new proxmox backup, but I'm still not sure about the physical backup options which we need in addition of the backup of VMs (containers will most likely be not used or rarely used)
- we do not need a HA environment (if host is down, it is down; but we should be able to get it back up and running within 1-2 days)
- we need an external backup (meaning physical external backup; looks like LTO is our only option)
- planned hardware so far:
- 36x LFF Chassis
- Single CPU Xeon Scalable Silve (e.g. 4214R)
- 192-256GB RAM
- 2x 2TB SSD for "internal" storage for the Windows VMs etc.
- approx. 20-24x 14TB SAS HDD (depending on the storage solution)
- RAID / HBA Controller (most likely HBA directly attached to VM if no one disagrees)
- SAS/NVMe SSD for Caching (depending on the storage design for the storage VM)
The main question I'm struggeling with right now is what VM I'm supposed to choose and related to that, which storage system (file system itself + "smart" options like self-healing, deduplication, snapshots etc) it'll be in the end: ZFS, Gluster, Ceph, ...
Raid5/6 (software or even hardware) are probably not the best solutions, especially with rebuild times.
ZFS might be an option, although checksumming needs quite some CPU performance (which I was told to considerate by a colleague). I'm not sure about the SSD caching or how to do it in this matter.
As for the VM OS itself: depending on the solution either a blank linux (debian, ubuntu, centos) system and samba configured properly or a more out of the box experience with a NAS distro or something like Univention Corporate Server (which can smoothly be combined with the 2-3 Windows VMs) to manage access rights, quota etc.
Any help with designing the best solution would be appreciated.