Hi!

Welcome to the forum!

Lots of questions there so let’s try and break these out and help, the main issue we’ll have is there are a fair few manufacturers out there so we can only advise with the ones we’re familiar with.

I couldn’t see a mention of Hyper-V vs VMware for this, so I’ll be giving generic advice that applies to both.

I primarily deal with Dell for infrastructure and the best option for something like this IMO is the Dell PowerEdge R740XD, but it gives you a good idea of what’s possible and what to look for in your preferred vendor.

Most Dell servers have the option of a BOSS (Boot Optimised Storage Solution). This is a RAID 1 two drive system used for the OS of the physical device. Now Dell don’t support VMware on SD cards, BOSS makes the most sense here. Especially if you’re using Hyper-V the ability to boot from SSD will help maintain performance.

The Dell R740XD has multiple chassis configurations, you haven’t mentioned actual required performance numbers here so the odds are you’ll have some clients with more demands than others, or some that have a higher standard of performance expectation for DR. The chassis configurations available allow you to use either sole use of or a mixture of NVMe, 2.5” SSD/HDD or 3.5” SSD/HDD. This will allow you to potentially consolidate replication and backup to the same device. If so you could look at a creating a RAID 60 volume based out of traditional spinning disks (or if the data set was small enough you could look at flash storage) whilst using faster SSD/NVMe RAID array for the replication target, with spare disk capacity remaining within the chassis.

If you’ll require more storage than a single server would require but want to keep it as close to a single unit as possible, you could use a single server + a storage shelf via the use of a SAS card to something such as a Dell PowerVault MD1400/1420 to tier out as necessary.

The Dell R740XD supports 1 or 2 CPUs, so you can give yourself expansion room by spec’ing a decent single CPU if that fits your requirements and know you can simply add a second should the customer require it. RAM is of course also configurable…

The R740XD supports up to 100GbE networking so regardless of the switching your clients are using you can find a NIC that will ensure this device won’t be the bottleneck.

Now that the server spec discussion is out of the way, onto your remaining questions:

If you want to start a VM that has been replicated(via traditional or CDP), the storage that the VM was replicated to is the storage that will be used for all of the change tracking. If you’re spinning up a backup however you need to specify either a vPower NFS share (VMware only) to keep the changes on, or a target datastore that the changes can be written to (this datastore needs to be visible within ESXi if you’re using VMware as your platform).

Replication will give you better performance than backup recovery and it can start immediately, that is because the VMs are registered in your target Hypervisor server and are stored uncompressed, ready to start.

Backups on the other hand are compressed and deduplicated to save space, so when you want to restore the entire VM to your DR server you have two options, either a full VM restore which will take potentially a long time but the VM will run faster once restored because its then running natively on the Hypervisor. Alternatively you can use Instant VM recovery which will run the VM from the backup, provided you have a datastore it can save the changes to (see comments above about this or vPower NFS for VMware), this will be slower performance and incur more resource penalties as a result (CPU/RAM requirements to read from the backup etc).

When sizing this DR host don’t forget to size your Veeam components accordingly, the Veeam help Center documentation will tell you the minimum you’ll need for Veeam B&R itself, if you’re using monitoring such as Veeam One, the specs for that, the specs required for proxy (data processing) /repository (data read/writing) roles.

Hopefully this has been helpful, even if a long post! Some of the things I’ve mentioned may have triggered more questions so please don’t hesitate to follow up with further questions as you think of them!

As noted by @MicoolPaul knowing the hypervisor will help to make recommendations.  I also work for a large MSP as well as we are starting to move towards the Lenovo SR550 2U server for our Veeam Appliances.  We previously used HPE Apollo servers with varying configurations but the Lenovo now allows us to standardize this.

If you are using Hyper-V with Windows 2019 on the backup server itself these SR550 units would be more than plenty to run the VMs you have noted as the smallest model has 25TB with 8 core CPU and 192GB of RAM.

We have 25TB, 50TB and 100TB models that we deploy depending on the customer requirements and needs.

Another option to maybe save money could be to have a second Veeam server installed and then you run a Backup Copy job over to that one or even just another server to act as a secondary repository server for this purpose.  Then if the primary server fails you could install Veeam and restore things from the secondary.

There will be many varying opinions on this subject and it will come down to your customers needs vs what they require or can spend.

Always here to chat within the MSP realm of things. 

Seeing as Michael and Chris gave you great long answers, can I provide a short answer ….”it depends”?

+1 Best Answer 

Short answer 

Thanks everyone.

Thanks heaps for the detailed response @MicoolPaul 

Sorry - we mainly use ESXI.

@Chris.Childerhose funny you say that, i determined the SR250 was not good enough and designed a system around the Lenovo SR550.

• 2x NVME mirror for the ESXI installation.
• 2x SSD Mirror for Server 2019 OS install for Veeam B&R Console
• 6x 3.5” Whatever drive size needed in RAID 6 for the repository location.

Sounds like a good build there. Would be interesting to see.

Happy to help, let us know how it goes!