IT infrastructure
A strategy for hardware, software and networks that actually work together.
Nobody notices infrastructure while it works. It gets noticed when a site is unreachable, a restore takes longer than promised, or the licence invoice doubles. All three have the same cause: an environment that grew rather than being designed.
Virtualisation and replacing VMware
Since the change in VMware licensing, many mid-sized companies are evaluating alternatives. We have taken both major open-source routes into production:
Proxmox VE with Ceph as distributed storage – for environments meant to work without a SAN. At a special machinery manufacturer in Baden-Württemberg we consolidated four standalone hypervisors into a three-node cluster and migrated 45 machines with no unplanned downtime.
XCP-ng with Xen Orchestra – for environments with an existing SAN. At an agricultural wholesaler in Lower Saxony we replaced eight vSphere hosts, moved 60 machines and folded in 20 previously physical servers.
Which route is right depends on storage, operating model and what your team wants to maintain. We tell you after the assessment, not before.
Networks
Segmentation instead of flat networks, next-generation firewalls, WAN connectivity across sites, VPN and Wi-Fi with access control. We work with pfSense, OPNsense and commercial platforms – the choice follows the requirement.
At a laboratory services provider in Bavaria we redesigned the network architecture across twelve sites, introduced centralised endpoint management and cut infrastructure licensing costs considerably.
Storage and backup
A backup only counts as a backup once the restore has been rehearsed. We set up backups with immutable retention, add an off-site copy and test recovery regularly. At one client this reduced the recovery time objective for business-critical systems from twelve hours to under two.
Typical engagements
Assessment and target architecture, migration from VMware to Proxmox VE or XCP-ng, network segmentation, firewall concepts, backup and recovery planning, monitoring with Zabbix, connecting branch offices.
On firewalls in detail: pfSense, OPNsense and Zenarmor.
Frequently asked questions
What does replacing VMware cost?
For small to mid-sized environments the one-off project effort is three to ten person-days for planning, build, testing and migration. A standard case with two to four hosts, 15 to 50 machines, high availability, Active Directory and a new backup is more like seven to fifteen days. Add vSAN, several sites, critical line-of-business applications or compliance requirements and it becomes fifteen days and up.
On top of that comes what does not appear on the service provider’s invoice: replacement or parallel hardware where the old cluster cannot be reused directly, shared storage, rebuilding the backup including restore tests, and adjustments to monitoring, firewall rules and documentation.
Hypervisor licence costs are not the large item with either open-source platform – neither Proxmox VE nor XCP-ng. The biggest mistake when replacing VMware is to compare only those costs.
How long is the business down during a migration?
Not in one stretch. You do not migrate “the business”, you migrate machine by machine, or application by application, by criticality and in separate maintenance windows.
Per machine that typically means: ten to thirty minutes for a non-critical Linux service, twenty to sixty minutes for a Windows server including driver changes and function tests, thirty to a hundred and twenty minutes for a file server depending on rate of change. Databases and ERP systems need thirty minutes to several hours depending on the consistency method. Directory services we prefer to rebuild in parallel rather than move across hard.
For critical systems we plan a warm migration: the new environment is built entirely in parallel, large machines are replicated in advance, and the maintenance window holds only the final sync and the switchover. The source machine stays powered off – as a defined way back, not as a hope.
Proxmox VE or XCP-ng?
Both are mature platforms; the question is what fits your operations.
Proxmox VE is our more frequent recommendation. It brings a unified Linux and KVM stack, LXC containers, ZFS and Ceph as storage options, and Proxmox Backup Server with deduplication and verification. Control through API and command line suits automation with Ansible or Terraform. Anyone running several environments on a reusable standard design does well with it.
XCP-ng with Xen Orchestra is the better choice when the exit from VMware has to happen quickly – the direct migration path through Xen Orchestra takes a lot of project risk out. Likewise when your team knows Xen historically, or Xen Orchestra is already set as the management and backup layer.
We run both in production and decide after the assessment, not before.