Thoughts on Veeam with hyper-converged appliances

[Frosty]

Hi all,

I'm about to embark on a full infrastructure refresh; something to last for 5+ years. Will be planning to support systems which will be on-premises, datacentre and public/private cloud. Recently upgraded our comms links to fibre (currently 100Mb/sec, but thinking about gigabit).

Is anyone here using hyper-converged appliances, such as the Dell VxRail, or similar? Interested in your thoughts if you are. We currently have a more traditional VMware setup, with 3 x ESXi host servers, and some Dell MD3200+MD1220+MD1200 shared SAS storage. It has been very reliable and performed at a high level. With a hyper-converged approach, I wonder about whether I am giving up control such that I can't managed performance/workloads in the same way.

We love Veeam software, so we will continue to use that for backup/restore/DR and I expect that I will put some gear into a rack in our ISPs datacentre for DR purposes. We've always used it in very simple fashion; basically just backups to a .VBK and then copy to portable HDDs for offsite storage. I know it can do a lot more than that, so I will be looking to learn as much as I can in the next month or so in order to design a really good backup/recovery/DR setup.

Frosty

[Pat490]

as far as I understand all HCI (like VxRail) you will not give up control but only get additional simplicity if you want! You get a new easier UI but you can always switch to regular VC/VCSA. And also all support related questions can go to one company because you bought everything all-together!

[Gostev]

Patrick is spot on, that's my understanding as well.

[DaveWatkins]

The biggest downside, and the reason we didn't go for it for our last refresh, is that you can't upgrade just one component. Say you have plenty of compute but just need to double your storage, the only way to do that is buy more nodes which also increases your compute capacity unnecessarily. If your business is fairly static or has predictable growth then that's fine, if you are planning to perhaps buy another company and have no idea how much data they have/store then it can quickly become an issue when trying to integrate systems

[Frosty]

I might be in the sweet spot for HCI. Pretty predictable looking forward (I hope!). So I would back myself to pick capacity for CPU, RAM and Storage and get it fairly right. I'm wondering whether to move my old infrastructure to datacentre for DR, or whether to add some HCI nodes down there. Anyone have any thoughts on that? I suppose it comes down to cost/benefit/risk analysis.

[Mike Resseler]

Stephen,

It indeed comes down to cost / benefit / risk . I used to do this all the time. My 'older' gear became used for DR purposes, or became part of the infrastructure where I tested recoveries on a regular basis (Yes, I am one of those guys that actually did that, although not enough ). By saving your company money by reusing the older equipment and move it to DR you can make them happy but it comes with risks that you need to be aware off and explain those risks very well to your managers.

- If you fall back on your DR side, be aware you won't have a support contract on it anymore and if something happens at that moment (and yes, I was so lucky at a certain point in time )
- What is the cost of resources for that older HW now (electricity / cooling / ...) and how much will it cost for newer generation HCI nodes? Might look stupid but it can make a difference (unless you leave your DR site off all the time but I assume that is not the idea of a DR site )
- Study / Experience: You probably know the tools / management for the old infra very well, but what if a new IT person starts at your company? Then he / she needs to learn two types of infra (with the older one being forgotten until too late ). Having two same type of environments saves on education and probably also less management time. But is that enough to justify the cost of other nodes

And I am certain I am forgetting a few more. Maybe they come back when I have my morning coffee

[Pat490]

DaveWatkins has a valid point about upgrades but of course you could also always add traditional SAN.
But I think the main purpose of HCI to keep it simple, add a new node and get more cpu, ram and storage. All in one package.

[chrisdearden]

The biggest downside, and the reason we didn't go for it for our last refresh, is that you can't upgrade just one component. Say you have plenty of compute but just need to double your storage, the only way to do that is buy more nodes which also increases your compute capacity unnecessarily. If your business is fairly static or has predictable growth then that's fine, if you are planning to perhaps buy another company and have no idea how much data they have/store then it can quickly become an issue when trying to integrate systems

That would depend on your choice of HCI solution - some are easier to scale up ( as well as out ) than others.