How to back up virtual machines with shared disks?

Question

System administration

Input data:
There are ESXi servers, there are VCenter (v5.1), virtual machines are spinning on them. Several storage systems are connected to ESXi, on which LUNs are cut and given to virtual machines. Inside virtual machines, disks from different storage systems are connected. Acronis Backup & Recovery 11.5 is used as a backup system , the agents of which are installed inside all VMs. The start time of backup of virtual machines is tied to the services that are located on them. Each machine is backed up 7 times a week (1 full and 6 differential). The operation of virtual machines outside of backup time is optimal, and disks are not a bottleneck. It is impossible to allocate an entire LUN to each VM, because not enough space for everyone.
Problem:
Backing up more than two VMs at the same time with shared disks is fraught with a performance failure of both machines and the services they host.
Here is a diagram for clarity:

The diagram shows 4 VMs and 3 LUNs, but in reality there are about 100 virtual machines, and about 15 LUNs. LUNs of different sizes and performance.
On the hypervisor side, the disks are connected like this:

• VM1 -> DS1, DS2
• VM2 -> DS1, DS2
• VM3 -> DS1, DS2, DS3
• VM4 -> DS2, DS3

From the storage side, virtual machines are connected like this:

• DS1 -> VM1, VM2, VM3
• DS2 -> VM1, VM2, VM3
• DS3 -> VM3, VM4

What do you want:

1. Make a backup plan that would take into account the load on the disks, the start time of each machine.
2. Find a tool to document the backup procedure. A kind of timeline, opening which you can determine at what time, for how long and which machine will be backed up, where there is a window to insert another task.

Share your experience, who and how solves such problems? Now it is almost a chaotic movement of copying plans and divination on coffee grounds.
UPD: Updated the version of Acronis Backup & Recovery (ABR) - version 11.5 is used, not 10.

Answer 1

t_q_l : Sorry, I didn't see it right away.
ABR doesn't work well with virtual machines. It works with them at the file level, so it has all the disadvantages of this approach. I myself used ABR 10 on Standalone ESXi with internal storage - I was satisfied. But after switching to full-fledged cluster solutions and shared storage, I was very upset. In the end, I chose Veeam, as it uses a different approach - pulling snapshots at the block level. The advantage of this approach is especially noticeable when several machines are backed up at the same time. If the data of these machines are located side by side, then the system simply performs almost linear reading from the storage. In addition, block-level deduplication is available, which further affects the occupied space, reducing the cost of storage.
In general, if you use shared storage, then the first thing you should have thought about is the SAN architecture, and as a result, the implementation of this approach opens the door for you to the world of products that provide redundancy at the SAN level, and this is a completely different approach, and as a result, completely different opportunities and efficiency.
I recommend trying the trial version of Veeam - this solution works well in hybrid environments, the so-called paraSAN, which will allow you to evaluate new approaches to redundancy and evaluate their effectiveness for your specific tasks. And then choose what suits you best.
Unfortunately, I am not familiar with your topology and needs, for this reason I will not advise anything, but only recommend that you familiarize yourself, as a starting point, with other methods and technologies for implementing and reserving virtual environments

Answer 2

At the request of the workers, I am clarifying the virtualization and backup scheme.
Virtual machine disks (from 2 or more) are located on different LUNs of different performance and size:
Backup using Acronis Backup & Recovery v.11.5 involves taking snapshots of all disks and copying them to separate NAS.
Problem: Backing up more than two VMs at the same time with shared disks is fraught with a performance failure of both machines and services located on them.
Backing up machines one at a time - we do not fit into the allowable "window". If we start backing up several VMs at the same time, then it's hard to guess that there is no plugging. An example in the picture is VM1 VM2 cannot be backed up together, but there are only 4 machines in the picture, but in reality there are 100 VMs and 15 LUNs.

1. How to build a backup scheme correctly?
2. How / in what way to keep documentation in order to correctly allocate the start time, end time and free windows for other machines?