All-in-one appliance

[ejenner]

Hi guys,

Apologies if this has already been asked.

With the new V11 which is coming out shortly is it advisable, assuming your infrastructure supports it... to dispense with a single B&R server and instead create multiple B&R servers with direct attached storage? i.e. to use the all-in-one enhancements.

Reason I ask is that we didn't ever move to a system of NAS or SAN repositories. Most of our backup data is hosted on repositories which are built out of servers with direct-attached disk. I've updated to 2019, evacuated and repopulated the repositories so have 2019 ReFS as well.

At the moment all of our backups are orchestrated from a single B&R server... is there a case with the availability of v11 to move the B&R processing to the server directly attached to the disks or have I misunderstood the intended use for the new features.

Thankyou.

[HannesK]

Hello,

or have I misunderstood the intended use for the new features.

which feature do you mean?

I don't understand the question. But I try to answer...

1) direct attached storage is fine.
2) all-in-one servers are fine until you reach the hardware limit
3) during upgrade, all services connected to a VBR server are unavailable.
4) if you have multiple VBR servers and you upgrade them one-by-one, then only one server (plus connected components) is effected.

Best regards,
Hannes

[ejenner]

This was about Gostev's latest email update.

"One area where v11 really changes the game is all-in-one backup appliances: deployments where all Veeam components are installed on a single server with lots of disks. Those who follow this digest closely know I've been promoting the usage of general-purpose servers for Veeam deployment for many years now, and every release we kept enhancing the engine for this deployment scenario. But if it was more of an evolution before, v11 is the revolution because it more than doubles the backup performance per appliance. To get there, we had to make many changes even to the most basic stuff – for example, how we write backup files content to storage! We also had to revisit our shared memory transport engine that passes data between source and target data movers running on the same box – incredibly, at these performance levels, even modern RAM speed can become a bottleneck if you don't work with it optimally. Finally, once this was all behind us, we ran into compute becoming the primary bottleneck on multi-CPU servers! So we had to implement full NUMA awareness and enhance our data movers placement logic to ensure they never end up on different CPUs."

[Gostev]

With the new V11 which is coming out shortly is it advisable, assuming your infrastructure supports it... to dispense with a single B&R server and instead create multiple B&R servers with direct attached storage? i.e. to use the all-in-one enhancements.

Actually, there's rarely a need to have more than one backup server except for scalability reasons (more than a few thousands machines to protect). Well, perhaps also for geographical (proximity of vCenter server) and political reasons (separation). But vast majority of Veeam customers should be fine with the single backup server.

On the other hand, it is quite common for very large customers to have more than one repository server. For those, to reduce network traffic it is always a good idea to have backup proxy and backup repository running on the same box, united with the proxy affinity rule.

So in case you have multiple storage servers dedicated to Veeam Backup & Replication, your installations should ideally look like this:

Code: Select all

    [Server 1: Backup proxy + Backup repository + Backup server
SOBR[Server 2: Backup proxy + Backup repository
    [Server 3: Backup proxy + Backup repository

The whole idea of "all-in-one" Veeam deployments (and why they are becoming so popular among our customers these days) is that modern hard drives and server hardware provides so much capacity and performance, that the most customers won't even need more than one such box on-prem. Because the single server can deliver 500TB+ of usable storage for the on-prem "landing zone" to store most recent backups, plus the infinite capacity of cloud object storage for older backups.

Now, of course if your production environment is a few PB in size, then you will have to have more than one such server. In this case, SOBR allows you to very easily scale your deployment horizontally with additional "all-in-one" proxy+repository boxes.

[ejenner]

We have a couple of proxy repositories for backing up from 3PAR SAN.

What I'm wondering from our perspective is if we have a couple of repositories like those you describe, in our case with 100tb+ of disk direct-attached to a server... is there some advantage to installing the B&R component directly on those servers so the storage and backup orchestration happens within the same unit?

[Gostev]

Having backup proxy and backup repository on the same box (and united with the proxy affinity rule) will certainly improve backup performance for any version, but especially with v11.

[ejenner]

So we're saying with just the proxy and repository on a single machine we might see a performance increase with V11? It's already fast, we get our nightly backups done before midnight anyway. Always exciting to improvements either way.

[Gostev]

It depends where the bottleneck is. For example, you may not see any improvement if you current setup is already able to cap your primary storage array IOPS and/or your SAN fabric throughput.

But with setups like the one it was tested, where two arrays can throw throw 100Gpbs over two separate SAN fabrics, it really helps to have proxy and repository on the same server to remove the bottleneck in the form of the network connection between them.