Cloud fundamentals

What is the cloud? Build your own on dedicated servers.

How cloud computing works, what makes a cloud private, and how we can help you build one on dedicated servers in Gloman’s infrastructure.

What does cloud computing mean?

Cloud computing gives you access to computing resources over a network. Processing and storage happen on servers in a data centre, while you access applications from your own devices.

A cloud pools resources and lets authorised users provision and release them on demand. Automation and usage measurement help manage that pool. A server connected to the Internet is only one building block.

Definition: NIST, The NIST Definition of Cloud Computing

Public, private and hybrid cloud

These terms describe who uses the infrastructure and how environments connect.

  • Public cloud: a provider offers resources to multiple organisations, often on shared hardware.
  • Private cloud: the cloud environment is reserved for one organisation. It can run at your premises or in a provider’s data centre.
  • Hybrid cloud: connected private and public environments support different parts of your workload.

Further reading: Cloudflare, What is the cloud?

Where dedicated servers fit

A dedicated server gives your organisation exclusive use of a physical machine. A hypervisor can divide its CPU, memory and storage between virtual machines, each running its own operating system and applications.

To build a private cloud, we plan the management and automation around those machines: templates, access controls, network allocation, monitoring and a way to provision resources consistently. Installing a hypervisor is the starting point.

Dedicated capacity is worth evaluating when you have steady workloads, large data transfers or specific hardware requirements. Compare the full cost, including software, backup capacity and engineering time. Public cloud can suit variable demand or applications that depend on managed services.

How we would design your cloud

We start with your applications and their dependencies. The design should answer these questions before hardware is ordered:

  • Compute: how much CPU and RAM do the applications need at peak load? On multi-socket servers, check NUMA placement and IRQ affinity so memory access and network processing do not become bottlenecks.
  • Storage: what capacity, latency and write rate do you need? Choose local or shared storage according to the workload and recovery plan; allow room for growth and replication overhead.
  • Networking: how will application traffic, storage traffic and administration be separated? Define private networks, firewall rules and protected administrator access.
  • Provisioning: who can create virtual machines, from which templates and within what resource limits? Choose a management platform that your operations team can support.
  • Operations: who handles hardware incidents, hypervisor updates, guest operating systems, application patches and out-of-hours alerts? Put those responsibilities in the service scope.

Plan for failures and recovery

A single host can be a useful development environment, but its failure takes all of its virtual machines offline. Production systems need a design based on the downtime and data loss your business can tolerate.

For a cluster, account for spare compute capacity, storage availability, redundant network paths and quorum. Test what happens when a host or link fails. A second location requires its own replication and recovery design.

Keep independent backups with separate access controls and test restores. Replication can copy an accidental deletion, and snapshots on the same storage can disappear with that storage. Define how much recent data you can lose (RPO) and how quickly service must return (RTO).

An example for a media business

Consider a publisher running websites, a content management system and video processing jobs. A proposed private cloud could separate the web applications, database and processing workers into different virtual machines, with internal traffic kept on private networks.

CPU and memory limits would stop processing jobs consuming all the resources needed by the websites. Backup storage would be planned separately, and CDN delivery could serve viewers without exposing the administration network.

This is a design example. The number of hosts, storage layout and failover behaviour depend on measured demand and the recovery requirements agreed for the project.

Building on Gloman’s infrastructure

Gloman provides dedicated infrastructure and operates its own network in Athens and Amsterdam. Dedicated servers can form the foundation for your private cloud, with the location and configuration selected around your users, applications and data flows.

Bring an inventory of your current systems, peak resource usage, storage growth and connectivity needs. Together we can define a target design, a pilot workload and a migration plan with a rollback path.

Capacity, connectivity, licences, backups and ongoing management should be specified in the proposal. Cloud capacity grows within the resources available to the cluster; adding physical servers requires planning.