Capacity and architecture
We design host count, storage type, networking and resource split between teams based on current load and expected growth.
Start here
Free Audits 5Urgent help
Emergency 4Ongoing support
Managed IT 5Projects
Servers & Hosting 6 Network & Virtualization 9 DevOps 5 Security & Recovery 5 Hardware & Licensing 2Find out where your servers, security, backups and performance stand, at no cost.
Server down, network out, site hacked or data lost? Call us now.
Monthly network and server support with response times written into the contract.
Setup, configuration, management and migration of Linux and Windows servers, panels and mail.
Network design and cabling, MikroTik, VoIP, branch links, virtualization and private cloud.
Networking & communications
Virtualization & cloud
Containers, Kubernetes, automated delivery, infrastructure as code and observability.
Server hardening, firewalls, backup and DR, ransomware recovery and incident forensics.
Advice, supply and installation of servers and network gear, plus genuine enterprise licenses.
Start here
Urgent help
Ongoing support
Projects
When developers file a ticket and wait days for a test server, or your data isn't allowed to leave the building, a private cloud makes sense. We build it on your own hardware so teams can create VMs and networks from a web panel within set quotas. For most organizations a Proxmox cluster is enough, and we recommend OpenStack only at larger scale.
Private cloud setup is building an environment on your own hardware where teams create virtual machines and networks themselves from a web panel, within quotas you define. It makes sense when developers wait days for every test server, company data may not be kept on a public cloud, rented cloud server bills keep growing, or each department has built VMs with no standard. PikoSystem designs capacity and architecture, installs a Proxmox VE cluster with Ceph, or OpenStack at larger scale, sets CPU, RAM and disk quotas per team, builds Cloud-Init VM templates, isolates team networks with VLANs or SDN, and ties sign-in to Active Directory or LDAP so each team sees only its own resources. You get an architecture document, role and quota definitions, ready VM templates, test restore results and a panel guide.
We design host count, storage type, networking and resource split between teams based on current load and expected growth.
We install a Proxmox VE cluster with Ceph, or OpenStack. OpenStack gives each team a project with CPU, RAM and disk quotas; on Proxmox, each team's access and resources are controlled with pools, roles and VM templates.
VM templates built with Cloud-Init let teams create standard machines without filing a ticket.
Each team or project gets its own network through VLANs or SDN, with firewall rules controlling traffic between them.
Sign-in is tied to Active Directory or LDAP, and roles are set so each team sees only its own resources.
We set up VM backups and resource monitoring so capacity shortfalls show up before the cluster fills.
A private cloud suits steady, predictable workloads: servers that run for months with little change in demand. Temporary or seasonal loads, such as a large test environment needed for a few weeks, fit a public cloud better because you switch them off afterwards.
Other factors are data residency rules, public cloud egress and link costs, your hardware refresh cycle and whether a team exists to run the platform. Without someone handling updates, monitoring and capacity, a private cloud slowly turns back into an untidy pile of VMs.
Ceph offers three interfaces. RBD provides VM disks, CephFS is a shared file system for several servers, and RGW gives S3-compatible object storage for applications and backups.
With three-way replication, usable space is roughly a third of raw capacity. Erasure coding saves space but costs CPU and is slower for VM disks. Ceph raises a nearfull warning at 85 percent by default and rebalances data when a node fails, so fill the cluster only to a level that stays below that after losing a node.
OpenStack enforces hard CPU, RAM and disk quotas per project. Proxmox VE has no equivalent hard quota, so resources are governed through pools, roles such as PVEVMUser and VM creation only from approved templates. Provisioning with Terraform and a Proxmox provider also keeps every VM defined in code and reproducible.
To stop abandoned VMs piling up, tag each one with an owner and project, give test environments an expiry date and review usage per pool every month.
Corosync, which handles Proxmox cluster membership, is sensitive to latency. Share its link with Ceph and VM traffic and nodes start dropping out of the cluster at busy hours. Give corosync a dedicated link and preferably a redundant one.
Another mistake is upgrading Proxmox without following the Ceph version order; the official upgrade guide requires moving Ceph to a compatible release first. Buying hardware with no headroom for a failed node also leaves the cluster short of space at the first failure.
We look at teams, workload types, security requirements and existing hardware.
We propose the architecture, hardware list and resource model and finalize them with you.
Platform, networking, storage, identity and backups are set up, and templates are created.
One team moves onto the private cloud first. We apply their feedback, then open the environment to the other teams.
With regular virtualization a sysadmin creates VMs for everyone else. In a private cloud teams provision their own resources within quotas, networks are isolated and usage is reported per team. The underlying hardware can be exactly the same.
For most organizations with a few teams and a few dozen VMs, Proxmox VE with Ceph is enough and easier to maintain. OpenStack fits large environments with several data centers, full API requirements and a dedicated operations team.
Yes. Sensitive workloads can stay on the private cloud while public-facing services or temporary loads run on rented cloud servers. We also set up the secure link between the two and cross-site backups.
Related searches
Quote
A few lines on where things stand and what you want is enough. An engineer calls you back, not a sales rep.