Migration from VMware - vSAN or Ceph?

Hunduster

Hunduster

Member
Proxmox Subscriber
May 13, 2023
57
11
8
Düsseldorf, Germany
Hello everyone,

We are also about to migrate from VMware to PVE.

We currently have VMware vSphere Essentials Plus with 3 host servers which are directly attached to a Dell Powerstore 1000T. Usable storage ~60TB.

Now our 3 host servers will run out of support in May this year, while the Power Store would still have 2 years of support.

We are currently considering which path to take now and in the future. Do we stay with a SAN as central storage or do we go straight to HCI and Ceph when we buy new servers?

My biggest concern at the moment is not the cost, our Powerstore prematurely and generally against HCI, but the presumably steep learning curve with Ceph as there are only three of us in IT with a total of 300 employees and 98% on-premise services.
 
Last edited:
Note that currently, you don't have thin provisioning && snapshot with SAN.

(I'm still working to add snapshot for SAN).

Ceph have a learning curve, but from my experience, if the setup is done correctly and hardware is ok, the only need to replace dead drives.

But if you really need to migrate in 2 months, maybe keep your SAN for now, and take time to evaluate ceph.
 
We don't have to migrate in 2 months. Only the on-site support for our servers is coming to an end.

The VMware licence runs until November this year. So we still have some time. The question is whether we are doing it right because I think it will end up with an HCI setup one way or another.
 
Hunduster said:
We don't have to migrate in 2 months. Only the on-site support for our servers is coming to an end.

The VMware licence runs until November this year. So we still have some time. The question is whether we are doing it right because I think it will end up with an HCI setup one way or another.
Click to expand...

It’s difficult to advise whether HCI is the right choice for you without a deeper understanding of your requirements. Moving to HCI means investing in new servers with different storage configurations (internal disks), potentially upgrading your network, and acquiring new skill sets.

Your approach to capacity planning, data protection, and expansion will change, as will performance considerations.

Can PVE & Ceph run in a “set it and forget it” mode? Absolutely - just like a traditional SAN such as PowerStore. That said, the real question is: Who takes that midnight call on 12/31 when something goes wrong? And who do they call if additional help is needed? You may want to establish a relationship with a trusted Proxmox support partner.

Ultimately, the decision depends on your company’s risk tolerance and budget.

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
@bbgeek17

Thank you very much for your detailed answer.

We have to invest in new servers this year anyway. New storage would then be due in two years at the latest. That's why we're currently considering replacing the Powerstore early and selling it for a small amount of money. If we go down the HCI route now, we would have an investment of around 180k for three new servers, each with 2 sockets and 1TB RAM with 60TB full SSD per host.

Our management is aware of the costs and has not fallen behind. We don't have the licence to print money either, but at the moment it's really less about the costs and more about what is a sensible investment and what we can handle.

100G switches are available from us.

We are certainly willing to acquire all of this and are also on fire, but we also have to ensure that operations run smoothly.

On the subject of backups, we use Veeam to back up to a TrueNAS system, which in turn mirrors the data to another, identical offsite system using 20G dark fibre.

We always answer the calls in the middle of the night ourselves anyway. In the past as well as in the future :-D
 
Last edited:
Hunduster said:
We would have an investment of around 180k for three new servers, each with 2 sockets and 1TB RAM with 60TB full SSD per host.
Click to expand...
That certainly is a good budget. Just be mindful that “two is better than one” isn’t always true. Dual-socket systems introduce complexities like cross-NUMA communication, which can impact performance and require more careful VM placement.

You may find this article interesting as you do your research: https://kb.blockbridge.com/technote/proxmox-tuning-low-latency-storage/index.html#processor-topology

Cheers

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
You must log in or register to reply here.
Top Bottom