Some organizations maintain the stance of “it’s just a backup server, this old hardware will suffice,” then complain when backups run slowly, and recoveries take a long time. Some organizations take the opposite approach and use high-performance SANs for backup targets/repositories to write—and recover—backups as quickly as possible, keeping in mind that features like Instant VM Recovery are limited to the speed of the disks on the backup storage.
Whichever approach you embrace, backup servers are an important aspect of any organization’s IT realm.
Backup servers often have many components, such as a console, a data mover or proxy of some sort, and a place to store the backup files called a repository. Depending on the size of your organization, you may install all the necessary components on one server or spread them across multiple servers for greater scalability and performance.
Below, you’ll find our guide to help you make informed decisions about your sizing backup servers and infrastructure considerations. Many of the requirements in this series come straight from Veeam and Microsoft, albeit our recommendations for RAM sizing on a Windows server generally come from experience. Different implementations and scenarios require different configurations.
If terms are unfamiliar to you, head over to our Veeam vocabulary webinar replay for a refresher.
Physical vs. Virtual
We prefer physical backup servers. Virtualization is excellent, and companies often have a “virtual first” policy to prevent server sprawl. Still, we look at performance and the goal we are trying to achieve when making these decisions. Physical servers provide an added layer of recoverability in virtual environments, meaning they operate independently of the virtual infrastructure and can tolerate failures that a virtual machine could not.
The 3-2-1 backup rule dictates storing backup files on different media (storage) than production data, so you will need storage and a method to have a repository outside of your primary storage. Technologies such as Veeam’s Instant VM Recovery allow organizations to tolerate primary storage failures as long as the backup repository is not a part of that media.
If you have a large virtual environment, splitting the roles across multiple virtual machines may be beneficial. As data grows, adding additional proxy servers or increasing resources is simple, but you should not keep the backup management server on the same cluster it is protecting. Virtual backup servers can skew performance metrics, be overprovisioned, and cause issues with VMware’s Distributed Resource Scheduling (DRS) while backups are running.
For this example, we will use a single, physical server for ease of calculations.
Windows Server OS Considerations
Veeam components should, and in some cases must, reside on a 64-bit Windows operating system. Microsoft’s requirements to run Windows Server 2019 (512MB of RAM?!) are laughable; any IT admin with experience will tell you the OS, required security agents, other third-party apps, and remote connections bump that number significantly. We recommend 16GB of RAM for the operating system alone. The OS requires 32 GB of disk space; with additional roles, components, updates, and other applications, this can easily climb over 100GB. Since we will use one server for all Veeam Backup & Replication roles in this example, we will cover additional resources below.
Veeam Roles: Backup Server, Console, Proxy, Repository, SQL Server
We recommend a minimum of two physical CPUs with at least four cores each for the base server. For disk space, you’ll need:
- 5GB for product installation
- 4.5GB for .NET Framework
- 10GB for every 100 virtual machines protected for cataloging
- 100GB+ for Instant VM Recovery cache
- Sufficient space for the repository – see below
Recommended RAM is 4GB + 500MB for every enabled job. Memory consumption varies based on the number of VMs in the job, the size of the virtual infrastructure, etc.
On-premises network speeds should always be at 1GbE or higher. For large data sets or shorter backup windows, 10GbE may be necessary; however, if the server components are older or slower, a faster network will not help.
- Microsoft SQL Server 2008 to 2019 (2016 SP1 Express included in the setup)
- System Center Virtual Machine Manager 2012 SP1 to 2019 Admin UI (optional, to register SCVMM server with Backup & Replication infrastructure)
- Microsoft .NET Framework 4.7.2 (included in the setup)
- Windows Installer 4.5 (included in the setup)
- Microsoft Windows PowerShell 5.1 (included in the setup)
- Firefox, Google Chrome, Microsoft Edge, or Microsoft Internet Explorer 11.0 or later
Veeam Backup & Replication Console
The Veeam Console provides the user interface to schedule and manage jobs. It does not have significant requirements, although, in larger deployments, under-provisioned instances may cause slow responses.
The console requires:
- CPU: x86-64 processor
- Memory: 2 GB RAM
- Disk Space: 500 MB for product installation and 4.5 GB for Microsoft .NET Framework 4.7.2 installation (only if installing on a separate machine)
Proxies are Veeam’s data movers and also are where deduplication, compression, and encryption happen. MoreCPU and RAM resources mean faster data processing. Veeam requires:
- CPU: x86-64 processor (minimum 2 cores or vCPUs). Using multi-core processors improves data processing performance and allows for more tasks to be processed concurrently.
- Memory: 2 GB RAM plus 200MB for each concurrent task. Using faster memory improves data processing performance.
- Disk Space: 300 MB plus space for log files.
To scale large jobs, a large number of jobs, or jobs spread across clusters, consider using multiple proxies.
Veeam Backup Repository
The repository is where the backup files reside. It can be direct-attached storage, network-attached storage, deduplication appliances, or object storage.
The repository requires:
- CPU: x86 processor (x86-64 recommended)
- Memory: 4 GB RAM, plus up to 4 GB RAM (64-bit OS) for each concurrently-processed machine or file share
Here is where most questions arise: how much disk space do you buy or allocate for the backup repository?
The amount of data you are protecting, the change rate, how frequently you back up that data, and your retention policies determine the amount of backup repository space you need. In our October webinar, we demonstrated free tools that help make informed decisions, and the restore point simulator is valuable in this situation. If you have 20TB of data, do backups once per day, and keep 30 restore points, you’ll need around 50TB of repository space. Adding weekly synthetic full backups doubles the requirement, and increasing the frequency of backups requires more storage.
In all cases, these calculations should serve as a minimum requirement for the amount of storage needed for your backup repository.
Admins often overlook disk performance for backups, measured per disk, per RAID set, with the controller in mind, etc. If you need 20TB of space, 6 x 4TB drives will generally net you better performance than 4 x 6TB drives due to having more spindles to use.
Veeam Backup & Replication has several components with individual resource requirements. When combining the elements onto a single server, it is essential to understand the resources needed to perform timely backups and restores or recoveries. When deploying physical servers, it is always better to slightly over-provision than be unable to meet business needs due to poor performance.